Monitoring electrical wiring connection configurations for power systems

Abstract

The disclosure relates to monitoring a power system including a power distribution unit (PDU) and a powered electrical equipment unit. The disclosure includes receiving PDU data including time series data indicating power usage of each PDU receptacle during a first time period; receiving activity data including time series data indicative of one or more activity metrics for each electrical equipment unit during the first time period; an event is detected indicating a change in an electrical wiring connection configuration between a socket of the PDU and an electrical equipment unit during a first time period. The event is detected by determining, based on the received PDU data and activity data, a first association group between the PDU socket and the electrical equipment unit indicating an electrical wiring connection configuration during a first time period; and comparing the first association group with a reference association group to identify altered associations indicating altered electrical wiring connections between the respective paired PDU sockets and electrical equipment units during the first time period. The disclosure also includes, upon detecting the event, estimating a change point of the electrical wiring connection configuration based at least in part on: a determined confidence score related to a changed association; and one or more end points of the first time period.

Description

Technical Field

[0001] The present disclosure relates to monitoring an electric power system including a plurality of power distribution units (PDUs) and a plurality of electrical equipment units powered therefrom. Specifically, the present disclosure relates to detecting changes in the configuration of electrical wiring connections between power outlets of the PDUs and the electrical equipment units. Background Art

[0002] The power system controls the delivery of power to various electrical equipment units or users that require such power. For example, in a data center, power is delivered from a power distribution unit (PDU) to various units of electrical equipment in a server room, such as various server machines. Specifically, power is delivered via electrical wiring connections between the outlets of the PDU and the server machines.

[0003] It is desirable to understand the topology or configuration of the electrical wiring connections connecting PDUs and electrical equipment units (e.g., servers), i.e., to understand which PDU outlets are connected to which electrical equipment units. This can help, for example, ensure that adequate redundancy is provided for server machines or other equipment that provide critical services, identify security vulnerabilities, or understand the impact of, for example, withdrawing or shutting down a particular power line for maintenance. This electrical wiring configuration can change relatively frequently over time; for example, when server machines are switched in and out of service, or when maintenance is to be performed on certain parts of the power system.

[0004] It is known to perform manual mapping of the topology or configuration of the electrical wiring connections of an electrical power system. That is, the physical wiring links may be checked manually by service personnel. However, this approach has the disadvantage of being prone to error and relatively slow and expensive to perform. In practice, a relatively long period of time may elapse between a change or a change occurring in the wiring topology and the change being reflected in the records, since the records may be updated only during relatively infrequent updates that are performed manually. This may create the problem that the knowledge of the wiring configuration may be time-sensitive, for example in the event of an unplanned server outage requiring the replacement of a set of PDUs.

[0005] The present disclosure is made in this context. Summary of the invention

[0006] According to one aspect of the present disclosure, a computer-implemented method for monitoring an electric power system is provided, the electric power system including a plurality of power distribution units (PDUs) and a plurality of electrical equipment units powered. The method includes: receiving PDU data, the PDU data including time series data indicating power usage of each PDU socket during a first time period; receiving activity data, the activity data including time series data indicating one or more activity metrics of each electrical equipment unit during the first time period; detecting an event indicating a change in an electrical wiring connection configuration between a socket of the PDU and the electrical equipment unit during the first time period, detecting the event by: determining a first association group between the PDU socket and the electrical equipment unit based on the received PDU data and the activity data (related to the first time period), the first association group indicating the electrical wiring connection configuration during the first time period; and comparing the first association group with a reference association group to identify a changed association, the changed association indicating a changed electrical wiring connection between a corresponding pair of PDU sockets and the electrical equipment unit during the first time period; and when the event is detected, estimating a change point of the electrical wiring connection configuration based at least in part on: a determined confidence score associated with the changed association; and one or more end points of the first time period.

[0007] In this way, the method allows changes or alterations in the wiring configuration (which may occur relatively frequently) to be identified in real time, quasi real time, or at any other desired frequency and the point of change can be accurately identified, for example, to notify an operator responsible for the change, or to associate the change with subsequent distribution changes.

[0008] It should be understood that the first time period may, for example, correspond to a most recent time period of the acquired time series data, for example, for the current analysis time period.

[0009] Optionally, estimating the change point includes estimating a proportion of the first time period before or after the change in the electrical wiring connection based on the determined confidence score. For example, estimating the change point may include: determining a duration of the first time period; and estimating the change point based on: one or more end points of the first time period; the determined duration of the first time period; and an estimated proportion of the first time period before or after the change in the electrical wiring connection.

[0010] In an example, the reference association group is a historical association group between PDU outlets and electrical equipment units, indicating an electrical wiring connection configuration during a second time period prior to a first time period. For example, the second time period may be a non-overlapping time period of the time series data immediately prior to the first time period.

[0011] Optionally, the received PDU data also includes time series data indicating power usage of each PDU outlet during the second time period. The received activity data may also include time series data indicating one or more activity metrics of each electrical device unit during the second time period. In an example, detecting an event may also include determining a reference association group based on the received PDU data and the activity data associated with the second time period.

[0012] In an example, estimating change points also includes iteratively adjusting the estimated change points by: determining corresponding associations associated with the changed electrical wiring connections based on received PDU data and activity data during: a third time period before a previously estimated change point; and / or a fourth time period after a previously estimated change point; and applying a function for adjusting a previously estimated change point based on a comparison of the determined associations and corresponding associations determined during a previous iteration.

[0013] Optionally, the function is configured to perform at least one of the following: if the determined association of the fourth time period does not match the corresponding association determined during the previous iteration, increase the previously estimated change point; if the determined association of the third time period does not match the corresponding association determined during the previous iteration, decrease the previously estimated change point; and / or if the determined association of the fourth time period matches the corresponding association determined during the previous iteration and the determined association of the third time period does not match the corresponding association determined during the previous iteration, decrease the previously estimated change point.

[0014] Optionally, during each iteration, adjusting the estimated change points may also include: determining a confidence score associated with each determined association; and applying a function for adjusting the previously estimated change points based on a comparison of the determined confidence score or total confidence score of the current iteration with the corresponding confidence score or total confidence score determined during the previous iteration.

[0015] Optionally, the function is configured to perform at least one of: if the determined confidence score or the total confidence score increases relative to the previous iteration, then adjust the change point in a manner of the previous iteration; and / or if the determined confidence score or the total confidence score decreases relative to the previous iteration, then adjust the change point in a manner opposite to the previous iteration. In an example, the estimated change point may be adjusted until the estimated change point is the same for consecutive iterations, or until the difference between the estimated change points of consecutive iterations is less than a threshold.

[0016] The third time period may, for example, start during the second time period. Alternatively, the third time period may start at the starting point of the second time period. Alternatively, the third time period may, for example, start during the first time period. Alternatively, the third time period may start at the starting point of the first time period. The fourth time period may, for example, end during the first time period. Alternatively, the fourth time period may end at the ending point of the first time period.

[0017] In an example, consecutive non-overlapping time periods of received PDU data and activity data are analyzed for event detection. The duration of each time period is determined by an analysis frequency. The method may also include adjusting the analysis frequency based on the estimated change point.

[0018] Optionally, the analysis frequency is adjusted based on the estimated change point and one or more historical change points indicating corresponding historical electrical wiring connection changes.

[0019] Optionally, adjusting the analysis frequency includes determining corresponding interval time periods between consecutive change points, and modeling the interval time periods as a function. For example, the function can be a probability distribution of the interval time periods. In an example, the function can be an exponential distribution.

[0020] Optionally, the analysis frequency is determined based on a function. For example, the analysis frequency can be determined based on an average value of the function.

[0021] Optionally, the sampling rate of the time series data is determined as a function of the analysis frequency. For example, the sampling rate may be determined as a scalar function of the time period of the analysis frequency.

[0022] Optionally, determining a first group of associations between PDU sockets and electrical equipment units includes: for each electrical equipment unit, estimating a model describing the activity of the corresponding electrical equipment unit based on the power usage of each PDU socket; and selecting which PDU sockets are associated with the corresponding electrical equipment units based on the estimated model.

[0023] Optionally, determining a first association group between PDU sockets and electrical equipment units includes: for each PDU socket, estimating a model describing power usage of the corresponding PDU socket based on the activity of each electrical equipment unit; and selecting which electrical equipment units are associated with the corresponding PDU socket based on the estimated model.

[0024] In an example, determining a first set of associations between PDUs and electrical equipment units includes: calculating a distance metric between power usage of each PDU outlet and one or more activity metrics of each electrical equipment unit; and determining which PDU outlets are associated with each respective electrical equipment unit based on the calculated distance metric.

[0025] Optionally, the method further comprises analyzing the determined first set of associations against one or more defined constraints of the electrical wiring configuration to be satisfied, and outputting a remedial action if the determined first set of associations does not satisfy each constraint.

[0026] Optionally, the power system is a data center power system.The one or more electrical equipment units may be, for example, server machines.

[0027] Optionally, the one or more activity metrics for each electrical equipment unit include one or more of: central processing unit (CPU) utilization of the electrical equipment unit; memory utilization of the electrical equipment unit; the number of bytes transferred in input / output operations generated by processes of the electrical equipment unit; the number of disk accesses per second; and graphics processing unit (GPU) activity of the electrical equipment unit.

[0028] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium is provided, on which instructions are stored. When the instructions are executed by a processor, the processor executes the method as described in the previous aspect of the present disclosure.

[0029] According to another aspect of the present disclosure, a controller for monitoring an electric power system is provided, the electric power system including a plurality of power distribution units (PDUs) and a plurality of electrical equipment units powered. The controller includes one or more processors, the processors being configured to: receive PDU data, the PDU data including time series data indicating power usage of each PDU socket during a first time period; receive activity data, the activity data including time series data indicating one or more activity metrics of each electrical equipment unit during the first time period; detect an event, the event indicating a change in an electrical wiring connection configuration between a socket of the PDU and the electrical equipment unit during the first time period, detecting the event in the following manner: based on the received PDU data and activity data, determining a first association group between the PDU socket and the electrical equipment unit, indicating the electrical wiring connection configuration during the first time period; and comparing the first association group with a reference association group to identify a changed association, the changed association indicating a changed electrical wiring connection between a corresponding pair of PDU sockets and the electrical equipment unit during the first time period; and when the event is detected, estimating a change point of the electrical wiring connection configuration based at least in part on: a determined confidence score associated with the changed association; and one or more end points of the first time period.

[0030] It will be appreciated that preferred and / or optional features of each aspect of the present disclosure may be incorporated into other aspects of the present disclosure alone or in appropriate combinations. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Examples of the present disclosure will now be described with reference to the accompanying drawings, in which:

[0032] Figure 1 Schematically illustrates a power system of a data center according to an example of the present disclosure;

[0033] Figure 2 An example of a method for monitoring the Figure 1 The steps of the method for a power system;

[0034] Figure 3 An example according to the present disclosure is shown in Figure 2 The event detection sub-step in the method shown;

[0035] Figure 4 An example according to the present disclosure is shown. Figure 3 Sub-steps of the method of event detection shown;

[0036] Figure 5 An example for estimating Figure 4 Sub-steps at the change point in the method shown; and

[0037] Figure 6 Another example of the present disclosure is shown for estimating Figure 4 Sub-steps at the change point in the method shown; and

[0038] Figure 7 An example according to the present disclosure is shown. Figure 2 Sub-steps in the method shown for adjusting the analysis frequency. DETAILED DESCRIPTION

[0039] Figure 1 1 is a schematic diagram of a data center 10 for housing computer systems and related components. For example, data center 10 may be in the form of a building or a dedicated space within a building.

[0040] Figure 1A power system 12 is schematically shown in which power is supplied to systems and components in a data center 10. The power system 12 includes a plurality of power distribution units (PDUs) 121 in the form of devices that distribute power from an input to a plurality of outlets of each PDU 121. PDUs are typically used to distribute power to devices such as computer racks and / or network devices in a data center. The input of each PDU 121 can receive power from any suitable power source 124 (e.g., an uninterruptible power supply (UPS), a (backup) generator, or other utility power source). Different PDUs in the PDU 121 can receive power from different power sources 124. For example, a first group 121a of PDUs 121 can receive power from a first UPS 124a, and a second group 121b of PDUs 121 can receive power from a second UPS 124b that is different from the first UPS 124a.

[0041] The power system 12 includes a plurality of electrical equipment units or components 122 that require power to operate or function. In the depicted example, the PDU 121 may provide power to electrical equipment in a server room or space 101 located in the data center 10. The electrical equipment units 122 in the server room 101 may primarily include server machines (or simply servers) that provide services (e.g., processing or storage / storage services) to various client stations (e.g., computers). The electrical equipment units 122 may also include other server room equipment that requires power, such as peripherals or hardware.

[0042] PDU 121 supplies power to server 122 via physical link 123 therebetween. Specifically, the link is in the form of electrical wiring 123, each of which connects a socket of one of PDU 121 to one of servers 122. Figure 1 As shown, each server 122 may be connected to more than one PDU 121. In the context of a data center, this provides redundancy in the power system, as the failure of one PDU does not necessarily mean that the operation of the associated server ceases, thereby preventing unplanned downtime of service critical equipment.

[0043] The specific wiring configuration of the power system 12 - i.e., which PDU outlets 121 are connected to the servers 122 via the electrical wiring 123 - can change relatively frequently over time. In a data center, servers and related equipment may be taken out of service relatively regularly for maintenance or upgrades, e.g., a particular power line may be shut down for a period of time. MAC (Move, Add, Change) operations may be performed to install, relocate, and / or upgrade various electrical equipment, e.g., servers.

[0044] Manually monitoring the mapping of electrical wiring connections 123 between the outlets of the PDU 121 and the servers 122 would be expensive, time consuming, and prone to error. Furthermore, when performed manually, updates to the mapping may be performed relatively infrequently, which means that a relatively long time may elapse between a change in the wiring configuration and the change being reflected in the record.

[0045] Although the power system is described in the context of providing power to equipment in a data center, it should be understood that the described power system can be used in different contexts where the PDU provides power to various electrical equipment units and components, for example, in a home or office setting, at a manufacturing site, etc.

[0046] Figure 1 Also included is a system or controller 14 for monitoring the power system 12. Specifically, the system 14 is provided for determining the configuration of the physical wiring or links 123 between the outlets of the PDU 121 and the server 122 and determining changes in the configuration, as will be discussed in more detail below. The controller 14 includes an input configured to receive data indicative of the operation of the power system 12, for example, data from the PDU 121, the server 122, and / or another source (e.g., a storage device) storing data indicative of the operation of the power system 12. The controller 14 includes an output that can send an alarm or control signal based on the determined wiring configuration and / or detected changes.

[0047] The controller 14 may be in the form of or include any suitable computing device, for example, one or more functional units or modules implemented on one or more computer processors. Such functional units may be provided by suitable software running on any suitable computing substrate using conventional or custom processors and memory. One or more functional units may use a common computing substrate (e.g., the functional units may run on the same server) or a separate substrate, or one or both functional units may themselves be distributed among multiple computing devices. The computer memory may store instructions for executing the methods to be performed by the controller 14, and the processor may execute the stored instructions to perform the methods.

[0048] Although indicated as being separate from the power system 12 in the illustrated example, in different examples, the system or controller 14 may be considered part of the power system 12. The controller may be located in any suitable location.

[0049] For example, the controller may be located near one or more other components of the power system, such as in a server room 101 with server machines of the data center 10, or in a different location within the data center 10. Alternatively, the controller may be located remotely from other components of the power system and / or remotely from the data center. Indeed, in some examples, the controller may be considered to be one of the electrical equipment units powered by the PDU 121 and monitor itself as part of the method described below to automatically determine and monitor the wiring topology between the PDU outlets and the electrical equipment units.

[0050] An advantage of the present disclosure is that it provides automatic determination and monitoring of the configuration or topology of physical links or wiring connections between sockets of a PDU of an electric power system and electrical equipment units to which the PDU provides power (e.g., server machines in a data center). Specifically, an advantage of the present disclosure is that automatic monitoring allows changes or alterations in the wiring configuration (which may occur relatively frequently) to be identified in real time, quasi-real time, or at any other desired frequency. In addition, an advantage of the present disclosure is that change points can be accurately identified, providing additional information related to the electric power system 12. For example, the change point can be used to identify the operator responsible for the change, or to associate the change with a subsequent distribution change.

[0051] This means that actions in response to identified configuration changes can be performed in a timely manner. For example, in the case of a security breach where a server is disconnected from a power source, the breach is detected immediately, meaning that actions can be taken quickly to contain the breach. As another example, in the case of an unexpected outage of one or more of the electrical equipment units, the associated PDUs can be immediately identified and replaced or repaired if necessary, minimizing unplanned downtime.

[0052] The automated monitoring of the present disclosure also provides for an inexpensive and accurate determination of the wiring configuration and eliminates the risk of error and expense involved when such tasks are performed manually.

[0053] The present disclosure achieves these benefits by determining a mapping between a PDU's sockets and electrical equipment units (e.g., servers) that represents a physical link between the PDU sockets and the servers. Specifically, the mapping is determined based on analyzing the power usage of each of the PDU sockets in conjunction with the (processing) activity of the servers so as to determine a correlation or pattern indicating a physical wiring connection between a specific PDU socket in the PDU sockets and a specific server in the servers. This will be described in more detail below. Beneficially, the present disclosure uses readily available data to automatically map wiring configurations.

[0054] Figure 2Steps of a method 20 are shown that is performed by a system or controller 14 to determine and monitor the configuration of electrical wiring connections 123 between power outlets of a PDU 121 and electrical equipment units 122 (eg, servers and / or other server room equipment).

[0055] The method 20 includes receiving PDU data indicating power usage over time for each of the PDU outlets at step 201. Specifically, the PDU data is received at an input of the controller 14. The PDU data may be in the form of time series data indicating power usage of the PDU outlets over a defined historical time period, i.e., historical time series data in the form of a power consumption signature indicating the temporal power consumption of each PDU outlet. The time series data may be sampled at regular intervals according to a specified sampling rate.

[0056] The PDU data may be received or obtained directly from each PDU outlet (or from each PDU 121). Alternatively, the PDU data may be obtained from a central platform that receives and stores power consumption data for each PDU outlet. The controller 14 may receive the PDU data substantially continuously, meaning that the power consumption data is received in real time or near real time, or the controller 14 may receive the PDU data at regular intervals, wherein the data covers a specified operating period.

[0057] In addition, the method 20 includes receiving activity or performance data indicating one or more activity or performance metrics of each electrical device unit 122 (e.g., server) over time at step 201. Similar to the PDU data described above, the activity data is received at an input of the controller 14. The activity data may be in the form of time series data indicating one or more metrics of server activity or performance over a defined historical time period, i.e., in the form of historical time series data in the form of a server activity signature indicating server activity or performance over time for each server 122. The time series data may be sampled at regular intervals according to a specified sampling rate.

[0058] The activity data may be received or obtained directly from each server 122, for example, via a standard monitoring interface commonly available on the server, such as VMware, vCenter, Windows Sysinternals, SolarWinds IT monitoring software, HPE OneView, etc. That is, the activity data may be retrieved from each server 122 by connecting to a management server or a dedicated API (application programming interface) on each server. The activity data may be received from the server's (PMS) platform management system.

[0059] The activity data may include any suitable data indicating the activity or performance over time of each server 122. For example, the activity data may include processor usage (central processing unit (CPU) percentage), memory usage, bytes read / written on disk (e.g., disk accesses per second), bytes sent / received on a network interface, graphics processing unit (GPU) activity of the server, etc.

[0060] Method 20 may optionally include: in step 202, realigning the received PDU and activity time series data so that samples of the PDU data and the activity data relate to the same time frame. Specifically, the sampling period of the received data may not be constant over time. For example, even if the sampling period should be five seconds, in fact, it may actually be between four and six seconds. Each server may also have a different sampling rate and / or be sampled at different times, for example, the first server is sampled at 0, 5, 10... seconds, and the second server is sampled at 2, 12, 22... seconds. Therefore, the received data can be manipulated to have the same sampling period and the same sampling instance. This can be performed by interpolation (e.g., linear interpolation) of the received data.

[0061] In one example, data realignment can involve upsampling the received PDU data and / or activity data and interpolating the upsampled data to a defined sampling period. The upsampled interpolated data can then be downsampled to a desired sampling period (typically the same sampling period as the original data), for example, one second, where the sampling of the PDU data and the activity data involves the same time step, i.e., the resulting data has a common sampling instant between the PDU socket and the server. Downsampling means retaining the desired data while discarding the remaining data. This data realignment can advantageously allow more accurate analysis and comparison of the PDU data and the server data to identify patterns and associations in the following steps.

[0062] The method 20 includes, at step 203, detecting an event that indicates a change in the configuration of the electrical wiring connections between the PDU outlets and the electrical equipment units 122. The method steps performed to detect such events can be scheduled at regular intervals using data generated in the intermediate time series. In other words, the method 20 can include analyzing consecutive non-overlapping intervals or cycles of received PDU data and activity data according to a specified analysis frequency.

[0063] Thus, in step 203, method 20 includes analyzing a first time period (between time T1 and T2) to detect any event indicating a change or alteration of the electrical wiring connection configuration. In this case, the first time period typically corresponds to the most recent time period because the analysis is performed over consecutive time intervals.

[0064] In order to detect such an event, method 20 includes sub-steps 301 to 305, such as Figure 3 shown.

[0065] In sub-step 301 , method 20 includes determining a first set of associations between PDU outlets 121 and electrical equipment units 122 based on PDU data and activity data associated with a first time period (ie, between T1 and T2 ).

[0066] The first association group determined in this manner indicates an electrical wiring connection configuration between the PDU outlets and the electrical equipment units 122 during the first time period, and may be determined according to one or more methods, which will be described in more detail below.

[0067] In one example, the association of each server (or other electrical equipment unit) 122 may be determined or derived by estimating a model that describes the activity of the corresponding server 122 during a first time period according to the power usage of each of the PDU outlets. The estimated model may then be used to determine which of the PDU outlets are associated with the corresponding server 122. In a different example, a model that describes the power usage of the corresponding PDU outlets according to the activity of the servers 122 during the analyzed time period may be estimated, and the estimated model may then be used to determine which of the servers 122 are associated with the corresponding PDU outlets.

[0068] In more detail, consider the first server in servers 122. The activity data of server 122 (e.g., the time series of the analyzed time period between T1 and T2) and the power usage data of each PDU outlet are extracted. The power signature time series of each PDU in PDU 121 are then used to fit a model to predict or estimate the server activity. For example, the fitted model can be a linear model of the following form:

[0069] s=a0+a1p1+a2p2+a3p3+…

[0070] Where s is the server 122 under consideration, p1, p2, p3, ... are the outlets of the PDU 121 of the system 12, and a1, a2, a3, ... are coefficients representing the proportion of activity on the server 122 under consideration that is related to the power consumption of the corresponding PDU outlet. a0 can be considered as an intercept term representing a baseline level of activity of the server 122 under consideration that is not explained by the power consumption of any PDU outlet.

[0071] It is assumed that an increase in power consumption at the PDU outlets is highly unlikely to correspond to a decrease in server activity. Therefore, a constraint can be imposed that the coefficients take non-negative values, i.e., a0, a1, a2, ... ≥ 0.

[0072] Once the model has been fit to the server 122 under consideration, a step is performed to discard from the model those PDU outlets that are not associated with the respective server 122. This may be referred to as a feature selection step. Specifically, the feature selection step examines or analyzes the derived coefficients in the estimated model, and specifically examines or analyzes the strength of the relationship between the time series of each of the PDU outlets and the server 122. PDU outlets whose derived coefficients are considered not to be significantly different from zero are discarded, and the remaining PDU outlets are considered to be connected to the server 122 under consideration.

[0073] The feature selection step can be performed step by step. For example, one of the PDU sockets can be considered for removal from the estimated model. A comparison of model metrics with and without the one of the PDU sockets can be performed. For example, this can involve estimating another model in the absence of data associated with the PDU socket considered for removal, and comparing the model to the other model. If there is no statistically significant degradation in the performance of the server 122 under consideration, it can be considered that the one PDU socket is not connected to the server 122, and the one PDU socket is removed from the model. Otherwise, the one PDU socket remains in the model. The process can be repeated for each PDU socket. The feature selection step can be performed using a linear regression method.

[0074] In addition, a bootstrap method can be used to improve the accuracy of feature selection. Specifically, the time series data in the first time period (T1 to T2) can be decomposed into smaller data sub-parts and then concatenated together to minimize the impact of unusual instances in the data when estimating the model.

[0075] The above steps are repeated in sequence for each of the servers 122 until it has been derived which of the PDU outlets are associated with and therefore connected to which of the servers 122 .

[0076] Although the steps of estimating the (linear) model and performing feature selection are described as separate steps above, wherein feature selection is after model estimation, these steps may alternatively be performed simultaneously. Specifically, this can be performed using an elastic net regularization algorithm. As known to those skilled in the art, the elastic net is a penalized regularized regression method that linearly combines the lasso and ridge methods, and is also known to those skilled in the art. For a description of the elastic net algorithm, see, for example, "Regularization and Variable Selection via the Elastic Net", Zou et al., JR Statist. Soc. B (2005), 67, Part 2, pp. 301-320. For a description of the lasso (minimum absolute shrinkage and selection operator) method or algorithm, see, for example, "Regression shrinkage and selection via the lasso", Tibshirani, JR Statist. Soc. B (1996), 58, No. 1, pp. 267-288. A description of the ridge regression algorithm is found, for example, in “Ridge Regression: Biased Estimation for Nonorthogonal Problems”, Hoerl et al., Technometrics (1970), Vol. 12, No. 1, pp. 55-67.

[0077] As mentioned before, the Elastic Net algorithm is a combination of the Lasso model and the Ridge Regression model. In both cases, these models aim to fit a linear model between the outcome (in this case, the server time series over the analyzed time period (T1 to T2)) and the predictor (in this case, the PDU time series over the analyzed time period (T1 to T2)), while aiming to minimize the complexity of the resulting model. "Complexity" in this context refers to the number of variables used in the model. The Lasso model does this by discarding predictors by setting their coefficients to zero, while the Ridge Regression does this by shrinking the coefficients to zero. In both cases, the coefficients can be estimated using coordinate descent, which aims to minimize a loss function that penalizes the complexity of the model.

[0078] When used alone, the lasso model may discard PDU time series that are highly correlated with one another in the PDU time series, for example, in the case of balancing power usage between two PDU outlets. In addition, using ridge regression alone will fail to discard any PDU outlets. The elastic net algorithm allows for a combination of these methods, and in particular allows for the discarding of uncorrelated PDU outlets while retaining relevant but highly correlated PDU outlets.

[0079] In a further modification of the example of the estimation model, the model can be a nonlinear model instead of a linear model. For example, a random forest can be used, which can also simultaneously derive or estimate the relationship between the server and the PDU socket while discarding irrelevant PDU sockets.

[0080] Once the model has been estimated for each server, i.e., once the association between each server and PDU socket has been derived, the determined association for each server can be updated in a repository, memory, or other data storage device of server-PDU associations, which can be part of the controller or system 14 or separate therefrom.

[0081] In another example, the step of determining the association between the PDU outlets and the servers 122 (sub-step 301) can be performed based on the calculated distance metric. Specifically, the distance between the power usage or consumption time series of each PDU outlet and the activity or performance metric time series of each server 122 is calculated over the analyzed time period (T1 to T2). The calculated distance is a measure of similarity, i.e., the correlation between the two time series over the analyzed time period (T1 to T2). The larger the distance, the less similar the two time series are. On the other hand, a smaller distance indicates a greater similarity between the time series signals.

[0082] The distance metric may be calculated using any suitable method (eg, mean square error), wherein a correlation coefficient (eg, Pearson correlation coefficient, Kendall coefficient, Spearman coefficient, etc.) is used as a measure of the linear correlation between two sets of data (ie, two time series). In practice, the distance between two time series can be calculated in different ways, for example: the multiplicative inverse of the correlation; using the matrix profile algorithm, which is known to technicians and whose description can be found, for example, in "Matrix Profile XII: MPdist: A Novel Time Series Distance Measure to Allow DataMining in More Challenging Scenarios", Gharghabi et al., 2018 IEEE International Conference on Data Mining, pp. 965-970; calculating the structural similarity index measure (SSIM), which is known to technicians and whose description can be found, for example, in "Image Quality Assessment: From Error Visibility to Structural Similarity", Wang et al., IEEE Transactions on Image Processing (2004), Vol. 13, No. 4, pp. 600-612; dynamic time warping; transform-based similarity methods, including discrete Fourier transform (DFT) or discrete wavelet transform (DWT).

[0083] Once all distances between pairs of time series are calculated, the PDU outlets are assigned to the server 122 such that the sum of the distances between two assigned time series is minimized over all assignments. That is, the sum of all distances between the selected pairs of time series (or other forms of received data) is minimized. This is known as the linear sum assignment problem, as known to those skilled in the art, and it can be solved, for example, as described in "On Implementing 2DRectangular Assignment Algorithms" Crouse, IEEE Transactions on Aerospace and Electronic Systems (2016), Vol. 52, No. 4, pp. 1679-1696.

[0084] In this context, each server machine 122 may need to have a redundant power supply. Therefore, multiple PDU outlets may be associated with each server. Therefore, an assumption for the linear and allocation problem may be that each server has two power supplies (PDU outlets), and the problem is solved based on this constraint or assumption to minimize the sum of distances. The resulting / determined allocations or associations are stored in a repository or data storage device (part of the controller 14 or separate from the controller 14). Similar to the above, the process of determining allocations or associations can be scheduled to be repeated at fixed intervals using data generated in intermediate timestamps. A set of determined associations together constitute a determined configuration of wiring connections between the PDU outlets and the servers 122.

[0085] In sub-step 302, method 20 may optionally include checking the server-PDU associations determined in the previous step. In one example, this may include checking the determined electrical wiring configuration against one or more defined constraints to be satisfied by the electrical wiring configuration. These constraints may, for example, include that a particular server 122 needs to be connected to a particular group of PDUs 121 or be located in a particular rack of PDUs 121, and / or that each server 122 (or some subset of servers 122) needs to be connected to at least two different PDUs 121 (for redundancy capabilities). The constraints may additionally or alternatively include that at most a predetermined number of servers 122 must be connected to a PDU 121, for example, because of power limitations of a power supply 124 to which the PDU 121 is connected. Such a check against constraints may be performed regardless of how the server-PDU associations are performed in step 203, but may be particularly used in examples where a model is estimated to determine associations.

[0086] In sub-step 303, method 20 includes comparing the first set of server-PDU associations (determined for the first time period) to a reference set of server-PDU associations to detect changed associations. In an example, the reference set of server-PDU associations may be a historical set of associations between PDU outlets and electrical equipment units, indicating an electrical wiring connection configuration during a second time period. For example, the second time period may be immediately prior to the first time period and extend between time T0 and T1.

[0087] Specifically, sub-step 303 may include comparing a first set or list of associations determined in step 301 (for the most recently analyzed time period - T1 to T2) with a second set or list of server-PDU associations determined for a previous time period (T0 to T1). For the purposes of the current analysis, i.e., during sub-step 303, the second set of associations may be determined, or generated during a previous iteration or run of the process and retrieved from a memory or data storage device to perform the comparison. In each case, the second set of associations may have been determined based on the received PDU data and activity associated with the second time period (i.e., the previous time period T0 to T1) substantially as described in step 301.

[0088] The comparison step is intended to identify differences between the current association group and the previous association group to identify changes or modifications that have occurred to the wiring configuration.

[0089] One way to do this is to first identify associations that exist in the first association group (i.e., the current group) but did not exist previously (i.e., did not exist in the second association group). For example, the first element (association) can be picked from the current association group. If that element was in the previous group, the next element in the current group is considered. If the first element was not in the previous group, it can be determined whether such a change or alteration is expected. For example, a particular server can be marked before determining the current association group to indicate that the particular server is about to be moved, added, etc. In this case, the change can be considered to be expected. On the other hand, if no such label or other information is available, the change can be considered unexpected and the particular element can be marked as such. Repeat this operation for each element (i.e., each entry or row of the current group).

[0090] After the above step of considering each element of the new association group, each element of the previous association group can be considered to identify associations that were previously present but are now gone (i.e., do not appear in the current group). Likewise, where a change from the previous group to the new group is identified, a check can be performed to determine whether the change was expected, e.g., information may be used to indicate that a particular server is about to be removed from the system before determining the current association list.

[0091] This analysis comparing current and previous association groups may be performed regardless of how the server-PDU association is performed in step 301, but may be particularly useful in examples where association is determined based on minimizing a distance metric between time series. In some examples, a time-stamped log of previous association lists may be stored for further analysis, e.g., to track how changes in topology affect the overall efficiency of the system.

[0092] If the first association group and the second association group match, the controller 14 can determine that no change in the electrical wiring configuration has occurred between the two time periods, and when the next interval of server / PDU data is received, the method can return to step 201 to repeat the analysis for subsequent cycles.

[0093] In this regard, it should be understood that the analysis frequency is typically set so that each analysis period includes one event, namely one change in the electrical wiring configuration. However, due to various reasons for wire changes, changes may not occur during the analysis period.

[0094] When a changed association is detected, in sub-step 303, the changed association indicates a changed electrical wiring connection 123 between the corresponding paired PDU outlets and the electrical equipment unit 122 during the first time period. Therefore, if the controller 14 detects a changed association, i.e., a change in the electrical wiring configuration between two time periods, the method 20 further includes a sub-step 304 for estimating a change point according to one or more methods.

[0095] In the example, Figure 4 As shown, method 20 includes sub-steps 401 and 402 for estimating change points.

[0096] In sub-step 401, method 20 includes determining confidence scores for the first set of associations, each confidence score indicating a weight of evidence supporting a corresponding determined association. When the associations are estimated based on data correlations and a system model in sub-step 301, it should be understood that the relative correlation strengths can, for example, reflect the confidence of the determined associations.

[0097] It will be appreciated that the confidence score may be determined as part of, or in conjunction with, the method for determining the first group of associations in sub-step 301. Thus, although the steps of determining the first group of associations and the corresponding confidence score are described above as separate steps, it will be appreciated that the confidence score may typically be determined simultaneously with the corresponding association. Thus, in sub-step 401, the confidence score may be determined by calling from a memory of the controller 14, for example, having been previously determined in sub-step 301. Furthermore, when a bootstrap approach is used to determine the associations in sub-step 301, it will be appreciated that the corresponding associations may be estimated for the corresponding sub-sampling of the first time period, and in sub-step 401, the confidence score of the altered association may be calculated as a proportion of such sub-sampling corresponding to the association.

[0098] In sub-step 402, method 20 includes estimating a change point based on the determined confidence score and one or more end points of the first time period. Specifically, the confidence score can be used as an indicator of the proportion of the first time period (T1 to T2) before or after the electrical wiring connection is changed.

[0099] For example, in sub-step 402, the change point may be estimated by determining a duration of the first time period and estimating the change point based on a start point of the first time period and an estimated proportion of the first time period before the electrical wiring connection is changed.

[0100] In other words, the change point T' can be estimated according to the following equation:

[0101] T'=T1+(1-C1)×(T2-T1)

[0102] Wherein, T1 is the starting point of the first time period, T2 is the ending point of the first time period, and C1 is the confidence score determined for the changed association during the time period from T1 to T2.

[0103] Since the association has changed during the time period from T1 to T2, and the confidence score C1 indicates the weight of the evidence supporting the determined association during that time period, the confidence score can be understood as providing an approximation of the proportion of the time period after the changed association. It should be understood that in this example, the confidence score C1 is a value between 0 and 1, where a value of 0 represents a minimum confidence and a value of 1 represents a maximum confidence. However, this is not intended to limit the scope of the present invention, and in other examples, for the purposes of the above equation, the determined confidence score C1 can be scaled and / or standardized (i.e., providing a value between 0 and 1), or an alternative formula can be applied that uses the confidence score C1 as an indicator of the proportion of the first time period (T1 to T2) before or after the change in the electrical wiring connection.

[0104] In another example, the method 20 further includes an iterative process 403 for improving the estimated change point T′ determined in the sub-step 402 .

[0105] Specifically, Figure 5 Example sub-steps 501 to 503 of an optional iterative process 403 of method 20 are shown for further improving the estimated change point T′ after the initial estimate in sub-step 402 .

[0106] Specifically, during each iteration (ie, i=1...n, and n is a positive integer), in sub-step 501, method 20 may include changing point T' i-1 The third time period and / or change point T' before the previous estimate i-1 The corresponding association associated with the changed electrical wiring connection 123 is determined during a fourth time period after or following the previous estimation of the change. For example, the third time period may start during the second time period, for example, at time T0, and at the change point T' i-1 The fourth time period can end at the change point T' i-1The previous estimate starts with and ends during a first time period, for example, at time T2.

[0107] In this context, it should be understood that for the first iteration, the change point T' i-1 The previous estimate of corresponds to the estimate produced in sub-step 402. However, in subsequent iterations, the change point T' i-1 The previous estimate of Figure 5 The estimates produced in the previous iteration of the method shown.

[0108] In each case, substantially as described in sub-step 301, the corresponding time period (ie, the third time period (T0 to T') received may be used to determine the time period. i-1 ) or the fourth time period (T' i-1 The PDU data and activity data from T2) to T3)) are used to determine the association associated with the changed electrical wiring connection 123. In this way, the method 20 can determine the first association A1′ associated with the changed electrical wiring connection 123 during the third time period. i , and / or determining a second association A2′ associated with the modified electrical wiring connection 123 during a fourth time period i .

[0109] In this example, method 20 also includes sub-step 502 during which controller 14 also determines a confidence score for the association determined in sub-step 501 , substantially as described in sub-step 401 .

[0110] As previously mentioned, it will be appreciated that the confidence score may be determined as part of, or in conjunction with, the method for determining the association in sub-step 501. Thus, although the steps of determining the association and the corresponding confidence score are described above as separate steps, it will be appreciated that the confidence score may generally be determined simultaneously with the corresponding association.

[0111] It should also be appreciated that the method 20 may determine a confidence score for each association and disregard any changes in those confidence scores that do not identify a change for the corresponding electrical wiring connection 123 .

[0112] In this way, the method 20 can be used for the first association A1 ′ i Determine the first confidence score C1' i , and / or for the second association A2' i Determine the second confidence score C2' i .

[0113] Thereafter, in sub-step 503, method 20 determines the association (A1′) based on the association (A1′) determined in sub-step 501. i , A2' i) and / or the confidence score (C1′) determined in sub-step 502 i , C2' i ) with the corresponding association (A1′) determined previously (ie, in sub-steps 301 and 401 or during the previous iteration (i-1)) i-1 , A2' i-1 ) and / or confidence score (C1' i-1 , C2' i-1 ), applying one or more rules, schemes and / or functions to adjust the previously estimated change point T' i-1 .

[0114] Figure 6 An example set of functions / rules for adjusting previously estimated change points is shown.

[0115] In sub-step 601, method 20 checks the second association A2′ determined in sub-step 501. i Is it equal to the previously determined second association A2' i-1 (ie, in sub-step 301 or during a previous iteration).

[0116] If A2' i Not equal to A2' i-1 , then in sub-step 602, the previously estimated change point T' is increased i-1 For example, the controller 14 may apply a specified time increment δT1 to a previously estimated change point T' i-1 , to determine the new estimated change point T' i .

[0117] However, if A2' i Equal to A2' i-1 , then method 20 continues to check the first association A1′ determined in sub-step 501 in sub-step 603 i Is it equal to the first association A1' determined previously? i-1 .

[0118] In this case, if A1' i Not equal to A1' i-1 , then the previously estimated change point T' i-1 In sub-step 604, the controller 14 may, for example, apply a prescribed time decrement δT2 to the previously estimated change point T'. i-1 , to determine the new estimated change point T' i .

[0119] However, if A1' i-1 Equal to A1' i-1 , then method 20 continues to check each confidence score C1' in sub-step 605i and C2' i or the total confidence score C1′ determined in sub-step 502 i +C2' i Is it greater than the previously determined confidence score C1' i-1 and C2' i-1 Or the total confidence score C1' i-1 +C2' i-1 (ie, in sub-step 401 or during a previous iteration).

[0120] If the confidence has increased, that is, if (C1' i +C2' i )>(C1' i-1 +C2' i-1 ), the method 20 comprises, in sub-step 606, adjusting the previously estimated change point T′ in the same manner as during the previous iteration (i-1) i-1 That is, if the confidence has increased and the estimated change point T' i-2 Increased during the previous iteration (i-1), then in sub-step 606, the estimated change point T' i-1 Increase again to determine the new change point T' i Similarly, if the confidence has been increased and the estimated change point T' i-2 decreased during the previous iteration, then in sub-step 606, the estimated change point T' i-1 Reduce again.

[0121] Alternatively, if it is determined in sub-step 607 that the confidence has been reduced, that is, if (C1′ i +C2' i )<(C1' i-1 +C2' i-1 If the confidence has decreased, then method 20 includes adjusting the previously estimated change point T′ in the opposite manner to the previous iteration in sub-step 608. i-1 That is, if the confidence has been reduced and the estimated change point T' i-2 Increased during the previous iteration, the estimated change point T' i-1 In sub-step 608, the value is reduced to determine the new change point T' i Similarly, if the confidence has been reduced and the estimated change point T' i-2 decreased during the previous iteration, then in sub-step 608, the estimated change point T' i-1 Increase.

[0122] However, if the confidence remains the same during consecutive iterations, that is, if (C1' i +C2'i )=(C1' i-1 +C2' i-1 ) or the difference between consecutive iterations is less than the threshold ε, the method 20 completes the iteration process in sub-step 609 and outputs the estimated change point T' i-1 .

[0123] In other examples, it will be appreciated that alternative rules, schemes or functions may be used to adjust the estimated change point T′, which may, for example, include any one or more of the above-described sub-steps 601 to 608 .

[0124] In each case, the estimated change point T' and / or the association of the change may also be recorded in memory, for example, the controller 14 stores a database of historical electrical wiring connection configuration changes.

[0125] return Figure 2 , the estimated change point T′ has been improved, the method 20 may optionally include, in step 204 , outputting one or more actions via the controller 14 in response to the results of the analysis.

[0126] In one example, when an event is detected in step 203, one such action output in step 204 may include adjusting the analysis frequency based on the estimated change point T'. That is, adjusting the frequency of analyzing consecutive intervals of activity data and PDU data (according to method 20).

[0127] The analysis frequency is typically set to a frequency that balances operating costs and accuracy parameters (e.g., event detection accuracy). A higher analysis frequency typically improves event detection accuracy, but increases operating costs. Striking a balance between these two goals is not easy and depends on the entropy of a particular application, or, for example, a particular data center 10 in which the method 20 is deployed. For example, a data center where the electrical wiring connection configuration changes every hour will benefit from a higher analysis frequency than a data center where the electrical wiring connection configuration changes every month.

[0128] Thus, in sub-step 304, the controller 14 may identify the precise timing of each change in the electrical wiring connection configuration and use such information as a surrogate for the entropy of the connectivity model.

[0129] To this end, method 20 may include Figure 7 Sub-steps 701 to 703 are shown for determining the analysis frequency as an output action in step 204 .

[0130] In sub-step 701, method 20 includes determining an interval time period between consecutive historical change points, including the most recent change point T' determined in sub-step 304. The interval time period may be determined independently of the association of the respective changes (i.e., independently of which electrical wiring connections 123 have been changed), thereby taking into account each detected change in the configuration of the electrical wiring connections.

[0131] Thereafter, method 20 processes the determined interval time periods to identify an analysis frequency that processes corresponding intervals of activity and PDU data of appropriate duration to capture one event per interval. Thus, the analysis frequency may be optimized in this manner according to one or more methods.

[0132] For example, in sub-step 702, method 20 includes modeling the interval time period as a function of the time between events, such as a probability distribution. For example, the interval time period can be modeled as an exponential distribution, assuming that changes in the electrical wiring connection configuration occur continuously and independently at a constant average rate.

[0133] In sub-step 703, method 20 determines an analysis frequency based on a function that models the interval time period. For example, the analysis frequency can be determined based on the average interval time period of the exponential distribution determined in sub-step 702. Specifically, the analysis frequency can be determined as the inverse of the average interval time period. Therefore, the determined analysis frequency should be suitable for detecting one event per interval, i.e., a change in the electrical wiring connection configuration.

[0134] The determined analysis frequency is used for subsequent monitoring of the electrical system 12 according to the method 20 to provide an optimized balance between operating costs and event detection accuracy. In this case, one event is expected to occur within each interval or analysis period.

[0135] In an example, one or more actions of step 204 may also include determining a sampling rate for PDU data and / or activity data collection. For example, the controller 14 may further determine each sampling rate as a function of the analysis frequency to provide maximum event detection accuracy with a minimum sampling rate. In an example, a continuous monotonic function may be extracted to define an optimal sampling rate based on the analysis frequency. The function may be a linear function, for example, where the sampling rate R is determined as:

[0136] R=K×F

[0137] Wherein, K is a predetermined constant, and F is the analysis frequency. It should be understood that in other examples, other suitable methods for determining the sampling rate based on the analysis frequency may be used.

[0138] In any case, the determined sampling rate R is then transmitted to the data acquisition portion of the controller 14 and used for subsequent monitoring of the electrical system 12 .

[0139] It will be appreciated that the analysis frequency and / or sampling rate may be updated in this manner after each event detection, or, for example, at a prescribed update frequency or after a prescribed number of events have been detected.

[0140] return Figure 2 In other examples, based on the events detected in step 203, remedial measures can be output in step 204, particularly in the event that an unexpected change in the electrical wiring connection configuration is detected, in the event that one or more constraints are not satisfied, and / or in the event that one or more redundant measures are no longer satisfied as a result of the change.

[0141] In one example, if one or more constraints are deemed not satisfied, an action may be output. For example, an audible and / or visual alarm (or other suitable alarm) may be generated, for example, near a server room. Alternatively or in addition, a notification may be sent to maintenance personnel and / or system administrators, for example, via email, telephone notification, audible or visual indicators in a control room of the data center 10.

[0142] In another example, if one or more unexpected changes are detected in the wiring configuration (in step 203), an action can be output. For example, an alarm can be sent to a system administrator to provide information related to the unexpected change. If it is still accessible via the network, a possible action can be to trigger a secure erase operation of the server 122 associated with the unexpected change. A further action can be to prevent access to the relevant server, for example, by automatically locking the door of the server room of the data storage 10 where the server is located, thereby preventing the removal of the device from the server room. Other actions can also be performed based on the required security level of the particular data center under consideration. For example, in the case of relatively low security, the action after sending an alarm to the system administrator (or other relevant personnel) can be performed only after the system administrator confirms that the alarm is not a false alarm. Although this may increase the delay of applying security measures, destructive server downtime can be avoided in the case of false alarms. It should be understood that these actions in response to unexpected changes may be particularly useful in the context of detecting and taking action to contain security vulnerabilities or sabotage in the data center (for example, individuals maliciously disconnect the server from the power line in a way that changes the power topology of the system, for example, unauthorized replacement of the server or theft of the server).

[0143] As described above, in a specific example, the determined association between the outlets of PDU 121 and server 122 can be used to ensure that the server 122 of system 12 has sufficient and necessary redundancy, for example, to avoid a disaster. The described method can also be used to restore redundancy to each server in server 122 as needed.

[0144] In more detail, redundancy refers to designing a system to replicate certain components so that a failure in one of the components (e.g., causing a disruption in normal power) does not affect the operation and service of critical IT infrastructure. In this context, redundant power supplies may be provided so that servers can continue to operate in the event of a power outage or failure. As described above, servers and / or PDUs in a data center may be added to, moved from, or removed from a power distribution system relatively frequently, for example, to perform routine work on computer equipment, such as installation, relocation, or upgrades. Therefore, ensuring that redundancy (e.g., power supply redundancy) is maintained in such a power system may be challenging.

[0145] The redundant power supply requirement or constraint may be that the critical equipment must be connected to at least two different PDU outlets, and / or the PDU outlets to which the critical equipment components are connected receive power from different power supplies 124. In this case, each electrical equipment unit 122 is a server machine. The operation of each server 122 may be critical, such that each server 122 requires redundant power, i.e., each server 122 needs to be connected to at least two of the PDU outlets 121. In different cases, only some of the servers 122 may provide services that are considered critical, in which case only a critical subset of the servers 122 may need to have redundant power supplies. In still different cases, multiple electrical equipment units may provide multiple different types of devices (e.g., peripheral devices as well as servers), in which case only a subset of the electrical equipment units may be considered critical and require redundant power supplies.

[0146] When analyzing the server-PDU associations at step 203 of method 20, it may be determined whether any constraints related to the necessary redundancy of system 12 are satisfied. This may first involve determining which electrical equipment units are considered critical, i.e., they require redundant power supplies. For example, this may be performed via a lookup in an equipment inventory repository. It may be that certain types of electrical equipment units (e.g., servers) are considered critical, while other types (e.g., peripherals) are not critical.

[0147] For each key electrical equipment unit in the identified key electrical equipment units, it can be first determined whether the corresponding unit is connected to at least two different PDU sockets. This ensures that the failure of one of the connected PDU sockets does not mean that the operation of the key unit is affected. If the condition that the key equipment unit is connected to two PDU sockets is met, it can be determined whether the corresponding key equipment unit is connected to at least two different power supplies 124a, 124b. In other words, the different PDU sockets to which the key electrical equipment unit is connected may need to be provided with electricity by different power supplies. For example, one PDU in the connected PDU 121 can receive electricity from the first power supply 124a, and another PDU in the connected PDU 121 can receive electricity from the second power supply 124b. This ensures that the failure of one of the power supplies 124a, 124b does not mean that the operation of the key equipment unit is impaired.

[0148] If it is determined that one of the critical electrical equipment units 122 does not satisfy the redundancy constraints, action can be taken to restore the required redundancy to the system 12. This can involve the controller 14 identifying PDU sockets to which the critical unit 122 can be connected to restore redundancy. A list of available PDU sockets can be obtained in the first instance, i.e., a list of PDU sockets that are not in use (e.g., because they are already connected to the electrical equipment unit 121). Such a list can be obtained from the wiring configuration of the determined associations (from step 203). Based on the determined associations, it is known which PDU sockets are connected to which electrical equipment units 122, and therefore, which PDU sockets have available sockets that are not currently in use, i.e., not currently connected to another component.

[0149] In one example, the output action at step 204 may simply provide an indication of which critical equipment unit 122 does not meet redundancy requirements and a list of available PDU outlets so that a user or operator can select which available PDU outlet to connect to the identified critical equipment unit 122 to restore redundancy.

[0150] Alternatively, the associated steps determined for the redundancy constraint analysis may further include selecting a specific one (or more) of the available PDU outlets, and the output action may then be providing a specific recommendation to the user to connect the selected PDU outlet to the identified critical equipment unit 122 to restore redundancy.

[0151] A list or specific recommendation of the identified critical equipment units 122 and available PDU outlets may be provided in any suitable manner. For example, this may be performed via an alert sent to management software of the system 12, a text message or call to a mobile phone, or a visual alert in a control room of the data center 10.

[0152] The selection of a particular one of the available PDU outlets can be based on a number of different factors, and the selection can be performed to optimize one or more aspects of the wiring configuration and system operation. The physical layout or arrangement of the various components in the data center 10 can be stored in a memory and can be available to the controller 14. The selection of a particular available PDU outlet can be based on the relative physical proximity of different components of the system 12. For example, in one example, a particular one of the available PDU outlets that is closest to the identified critical unit 122 (or an available outlet of a particular one of the PDUs 121) can be selected, which can help maintain a simple wiring configuration. In another example, a particular one of the available PDU outlets that is closest to / adjacent to another (or another) PDU outlet connected to the identified critical unit 122 can be selected, again for reasons such as simplicity of configuration, but optionally, the available PDU outlet receives power from a different power source 124.

[0153] The selection of a particular one of the available PDU outlets may optionally be based on the load of the different power sources 124 providing power to the PDU 121 at a given time. The current load of the different power sources 124 may be obtained in any suitable manner. For example, the current load of each power source may be derived from the determined server-PDU associations. In an example, the selection of the available PDU outlets may be made to improve load balancing in the system 12, for example, the selected PDU outlet may be the portion of the PDU 121 that receives power from the power source 124 in the system 12 having the lowest current load.

[0154] The selection of a particular one of the available PDU outlets may be based on a combination of different factors, for example, according to an optimization algorithm that optimizes across a number of different factors. For example, the selected PDU outlet may be identified based on one or more of: maximizing the use of adjacent PDU outlets or adjacent PDUs 121; proximity to the electrical equipment unit in question; improved load balancing of the system 12; and consideration of the entire power chain of the identified PDU or PDU outlet.

[0155] Many modifications may be made to the examples described without departing from the scope of the appended claims.

Claims

1. A computer-implemented method for monitoring an electric power system, the electric power system comprising a plurality of power distribution units (PDUs) and a plurality of electrical equipment units powered by the power distribution units, the method comprising: receiving PDU data including time series data indicating power usage of each PDU outlet during a first time period; receiving activity data comprising time series data indicative of one or more activity metrics of each electrical equipment unit during the first time period; detecting an event, the event indicating a change in a configuration of electrical wiring connections between outlets of the PDU and the electrical equipment unit during the first time period, by: determining, based on the received PDU data and the activity data, a first set of associations between the PDU outlets and the electrical equipment units, the first set of associations indicating a configuration of the electrical wiring connections during the first time period; as well as comparing the first set of associations to a reference set of associations to identify changed associations indicating changed electrical wiring connections between corresponding pairs of the PDU outlets and the electrical equipment units during the first time period; as well as Upon detecting the event, estimating a point of change in the electrical wiring connection configuration based at least in part on: a determined confidence score associated with the association of the change; and One or more end points of the first time period.

2. The method according to claim 1, wherein: Estimating the change point includes estimating a proportion of the first time period before or after the change of the electrical wiring connection based on the determined confidence score.

3. The method according to claim 1 or claim 2, wherein: The reference association group is a historical association group between the PDU outlets and the electrical equipment units, the historical association group indicating the electrical wiring connection configuration during a second time period prior to the first time period.

4. The method according to claim 3, wherein: The received PDU data also includes time series data indicating power usage of each PDU outlet during the second time period; wherein the received activity data further comprises time series data indicating one or more activity metrics of each electrical equipment unit during the second time period; and Wherein, detecting the event further comprises determining the reference association group based on the received PDU data and activity data related to the second time period.

5. The method of any one of the preceding claims, further comprising iteratively adjusting the estimated change point by: Determining a corresponding association related to the changed electrical wiring connection based on the received PDU data and the activity data during: a third time period prior to the previously estimated change point; and / or a fourth time period after the previously estimated change point; and Based on a comparison of the determined associations and corresponding associations determined during previous iterations, a function for adjusting the previously estimated change points is applied.

6. The method according to claim 5, wherein: The function is configured to perform at least one of the following: if the determined association for the fourth time period does not match a corresponding association determined during a previous iteration, increasing the previously estimated change point; If the determined association for the third time period does not match a corresponding association determined during a previous iteration, reducing the previously estimated change point; and / or If the determined association of the fourth time period matches a corresponding association determined during a previous iteration, and the determined association of the third time period does not match a corresponding association determined during a previous iteration, then the previously estimated change point is decreased.

7. The method according to claim 5 or claim 6, wherein: During each iteration, the estimated change points include: determining a confidence score associated with each determined association; and Based on a comparison of the determined confidence score or total confidence score of the current iteration with the corresponding confidence score or total confidence score determined during the previous iteration, a function for adjusting the previously estimated change point is applied.

8. The method according to claim 7, wherein: The function is configured to perform at least one of the following: If the determined confidence score or the total confidence score increases relative to the previous iteration, adjusting the change point in the manner of the previous iteration; and / or If the determined confidence score or the total confidence score decreases relative to the previous iteration, the change point is adjusted in an opposite manner to the previous iteration.

9. The method according to any one of claims 5 to 8, wherein: The third time period starts during the second time period, and optionally, starts at the beginning of the second time period; The third time period starts during the first time period, optionally, starts at the beginning of the first time period; and / or The fourth time period ends during the first time period, and optionally, ends at an end point of the first time period.

10. A method according to any one of the preceding claims, wherein: Consecutive non-overlapping time periods of the received PDU data and activity data are analyzed for event detection, the duration of each time period being determined by an analysis frequency, and wherein the method further comprises adjusting the analysis frequency based on the estimated change point.

11. The method according to claim 10, wherein: The analysis frequency is adjusted based on the estimated change point and one or more historical change points indicative of corresponding historical electrical wiring connection changes.

12. The method according to claim 11, wherein: Adjusting the analysis frequency includes determining corresponding interval time periods between consecutive change points, and modeling the interval time periods as a function, optionally, the function is a probability distribution of the interval time periods, optionally, the function is an exponential distribution.

13. The method according to claim 12, wherein: The analysis frequency is determined based on the function, optionally wherein the analysis frequency is determined based on an average value of the function.

14. The method according to any one of claims 10 to 13, wherein: The sampling rate of the time series data is determined as a function of the analysis frequency, and optionally, as a scalar function of the time period of the analysis frequency.

15. A method according to any one of the preceding claims, wherein: Determining the first association group between the PDU socket and the electrical equipment unit includes: For each electrical equipment unit, estimating a model describing the activity of a corresponding electrical equipment unit according to the power usage of each PDU outlet; and selecting which PDU outlets are associated with the corresponding electrical equipment units based on the estimated model, or For each PDU outlet, estimating a model describing the power usage of corresponding PDU outlets according to the activity of each electrical equipment unit; and Which electrical equipment units are associated with the corresponding PDU outlets are selected based on the estimated model.

16. The method according to any one of claims 1 to 14, wherein: Determining the first association group between the PDU and the electrical equipment unit includes: calculating a distance metric between the power usage of each PDU outlet and the one or more activity metrics of each electrical equipment unit; and A determination is made based on the calculated distance metric as to which PDU outlets are associated with each respective electrical equipment unit.

17. The method of any preceding claim, further comprising analyzing the determined first set of associations against one or more defined constraints of the electrical wiring configuration to be satisfied, and outputting a remedial action if the determined first set of associations does not satisfy each constraint.

18. A method according to any one of the preceding claims, wherein: The power system is a data center power system, and wherein one or more of the electrical equipment units are server machines.

19. A method according to any one of the preceding claims, wherein: The one or more activity metrics for each electrical equipment unit include one or more of the following: a utilization rate of a central processing unit (CPU) of the electrical equipment unit; memory usage of the electrical equipment unit; the number of bytes transferred in input / output operations generated by processes of said electrical equipment unit; Disk accesses per second; as well as A graphics processing unit (GPU) activity of the electrical equipment unit.

20. A non-transitory computer readable storage medium having stored thereon instructions which, when executed by a processor, cause the processor to perform the method according to any one of the preceding claims.

21. A controller for monitoring an electric power system, the electric power system comprising a plurality of power distribution units (PDUs) and a plurality of electric equipment units powered by the power, the controller comprising one or more processors, the processors being configured to: receiving PDU data including time series data indicating power usage of each PDU outlet during a first time period; receiving activity data comprising time series data indicative of one or more activity metrics of each electrical equipment unit during the first time period; detecting an event, the event indicating a change in a configuration of electrical wiring connections between outlets of the PDU and the electrical equipment unit during the first time period, by: determining, based on the received PDU data and the activity data, a first set of associations between the PDU outlets and the electrical equipment units, the first set of associations indicating a configuration of the electrical wiring connections during the first time period; as well as comparing the first set of associations to a reference set of associations to identify changed associations indicating changed electrical wiring connections between corresponding pairs of PDU outlets and electrical equipment units during the first time period; as well as Upon detecting the event, estimating a point of change in the electrical wiring connection configuration based at least in part on: a determined confidence score associated with the association of the change; and One or more end points of the first time period.

Images