Disaggregated compute system and method

The disaggregated compute system addresses data security challenges by using a fabric manager and memory controller to protect data within the memory pool, ensuring secure and efficient data protection without host involvement, supporting any CPU and maintaining performance.

WO2024245536A1PCT designated stage expired Publication Date: 2024-12-05HUAWEI TECH CO LTD +1

Patent Information

Application Number
PCT/EP2023/064390
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-30
Publication Date
2024-12-05

AI Technical Summary

Technical Problem

Conventional disaggregated compute systems face challenges in ensuring data security, as they rely on hosts for memory protection, require specific CPU support for memory security features, and can experience performance degradation and increased user burden for data protection.

Method used

A disaggregated compute system with a fabric manager that enables and disables data paths between hosts and memory devices, using a memory controller to protect data by cleaning or encrypting it, and employing a fabric mesh for connectivity, allowing the datacenter to manage memory isolation without host involvement, supporting any CPU and maintaining performance.

Benefits of technology

Ensures secure and efficient data protection without burdening users or hosts, maintaining performance and flexibility across all CPUs, and enabling seamless memory resource switching between hosts.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure EP2023064390_05122024_PF_FP_ABST
    Figure EP2023064390_05122024_PF_FP_ABST
Patent Text Reader

Abstract

A disaggregated compute system and method are provided. The disaggregated compute system comprises a fabric manager configured to communicate with a plurality of hosts and with a memory pool of the system. The memory pool comprises a memory controller and one or more memory devices. The fabric manager enables a data path between any host and at least one of the memory devices of the memory pool. After the host writes data in the at least one memory device, the fabric manager disables the data path between the host and the at least one memory device of the memory pool. The memory controller configures the memory pool to protect the data of the host in the at least one memory device after the data path between the host and the at least one memory device of the memory pool is disabled.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] DISAGGREGATED COMPUTE SYSTEM AND METHOD

[0002] TECHNICAL FIELD

[0003] The present disclosure relates to a disaggregated compute system. The disclosure further provides a method for a disaggregated compute system and a computer program product to perform the method.

[0004] BACKGROUND

[0005] Current computing architectures include a compute node with a local memory. Only a single host can access the memory which is located next to a central processing unit (CPU) in the same enclosure.

[0006] Disaggregation architectures enable the use of a memory pool outside of the host enclosure. The memory is on a separate power domain, so that when the host losses power or disconnects from the memory, the data on the memory will be intact. In these systems the host connection to the memory is controlled by a fabric manager, which can accesses both the host and the memory pool and can also connects them both.

[0007] FIG. 16 schematically depicts the operation of a conventional disaggregation architecture. In FIG. 16(a), a host (for example Host A) may write data to a memory module. After the host has written the data, no host is connected to the memory module, as depicted in FIG. 16(b). Unlike a traditional architecture, which depends on power of a local CPU, the memory in the disaggregated system may be connected to a separate power system and, as such, the data in memory is not deleted.

[0008] Then, the fabric manager may connect another host, for example Host B, to the memory module, as depicted in FIG. 16(c). Then, Host B can read the data written by Host A. Hence, security of the data cannot be ensured.

[0009] In the Hard Disk Drive (HDD) / Solid State Drive (SSD) domain, conventional solutions addressed data security by using a self encrypting drive (SED) feature on the computing device itself. The host changes the key on a drive so that without the key the data is encrypted. The memory encrypting feature is also controlled by the host connected to the memory module.

[0010] However, these solutions, depicted schematically in FIG. 17, require a CPU that supports the memory security feature. Moreover, they require the host user to enable and configure the memory security feature in the CPU (if available). Furhter, the convenional solutions may exhibit performance degradation when using the memory security feature in the CPU. Further, in the current solutions the burden and the responsivity for the data protection in on the user rather than on the datacenter owner.

[0011] SUMMARY

[0012] In view of the above, this disclosure aims to improve conventional solutions for a disaggregate compute system. An objective is to enable the memory pool to provide a secure service by protecting the data of each user (host) in an efficient and simple manner, without burdening the users and hosts with additional requirements and additional management.

[0013] This and other objectives are achieved by the solutions of this disclosure as described in the independent claims. Advantageous implementations are further defined in the dependent claims.

[0014] According to a first aspect, a disaggregated compute system is provided. The disaggregated compute system includes a fabric manager configured to communicate with a plurality of hosts and with a memory pool of the system, and the memory pool includes a memory controller and one or more memory devices, wherein the one or more memory devices are configured to hold data of the hosts. The fabric manager is further configured to enable a data path between any host and at least one of the memory devices of the memory pool, and disable the data path between the host and the at least one memory device of the memory pool after the host writes data in the at least one memory device. The memory controller is configured to configure the memory pool to protect the data of the host in the at least one memory device after the data path between the host and the at least one memory device of the memory pool is disabled. This provides the advantage of ensuring that the rack or datacenter that provides the memory pooling service does not need to rely on the hosts for protecting the data in the one or more memory devices.

[0015] Furthermore, the disaggregated compute system can use any CPU and is not bound to products that provide the specific memory security features.

[0016] The memory pool may be implemented in a separate module located outside of an enclosure of the host. Further, the memory pool may be common for all the hosts.

[0017] In an implementation form of the first aspect, the fabric manager is further configured to control the memory pool to allocate a memory space for the host in the at least one memory device.

[0018] The fabric manager may be a combined hardware-software system that may be configured to enable allocating the resources of the common memory pool and to enable connectivity of any host to the memory pool.

[0019] In an implementation form of the first aspect, after disabling the data path between the host and the at least one memory device of the memory pool after the host writes data in the at least one memory device, the fabric manager is further configured to enable a local data path between the memory controller and the at least one memory device of the memory pool.

[0020] This enables protection of the user's data without burdening the users and hosts with additional requirements and additional management.

[0021] In an implementation form of the first aspect, protecting the data of the host in the at least one memory device includes cleaning the memory space for the host allocated in the at least one memory device of the memory pool.

[0022] When switching a memory resource between different hosts, after detaching a host from the memory device of the memory pool and prior to connecting another host to the memory device, the host's data residing on the memory device may be cleaned, thus preventing the host's data to be subsequently accessed by any other host. In an implementation form of the first aspect, cleaning the memory space for the host allocated in the at least one memory device of the memory pool includes erasing the memory space for the host allocated in the at least one memory device of the memory pool, alternatively overwriting the memory space for the host in the at least one memory device of the memory pool using random data.

[0023] This provides the advantage of enabling different manners to protect the host's data, enhancing the efficiency and flexibility of the device.

[0024] In an implementation form of the first aspect, the disaggregated compute system further includes a fabric mesh, the fabric mesh being configured to communicate with the fabric manager, with the plurality of hosts, and with the memory pool. The fabric mesh is further configured to enable the data path between the host and the at least one memory device of the memory pool through the fabric mesh, and disable the data path between the host and the at least one memory device of the memory pool through the fabric mesh.

[0025] That is, the connectivity of the hosts to the at least one memory device of the memory pool may be provided by the fabric mesh. The fabric mesh may comprise one or more physical implementations.

[0026] In an implementation form of the first aspect, the memory controller includes a memory host device and, after disabling the data path between the host and at least one memory device of the memory pool after the host writes data in the at least one memory device, the fabric manager is further configured to enable a data path between the memory host device and the at least one memory device through the fabric mesh.

[0027] The fabric mesh, thus, may further enable the connectivity of the memory host device to the at least one memory device within the memory pool.

[0028] In an implementation form of the first aspect, the memory controller includes a reset module host, and after disabling the data path between the host and the at least one memory device of the memory pool after the host writes data in the at least one memory device, the fabric manager is further configured to enable a data path between the reset module host and the at least one memory device through the fabric mesh. The fabric mesh, thus, may further provide the connectivity of the reset module host to the at least one memory device within the memory pool.

[0029] In an implementation form of the first aspect, protecting the data of the host in the at least one memory device after the data path between the host and the at least one memory device of the memory pool is disabled includes performing a low level reset of the at least one memory device of the memory pool.

[0030] This provides a further manner to protect the host's data, further enhancing the efficiency and flexibility of the device.

[0031] In an implementation form of the first aspect, the fabric manager is further configured to generate an encryption key and send the encryption key to the memory controller, and change the encryption key after the host writes data in the at least one memory device and before the data path between the host and the at least one memory device of the memory pool is disabled. The memory pool further comprises an encryption engine configured to communicate with the memory controller and with the one or more memory devices. Additionally, the memory controller is further configured to protect the data of the host in the at least one memory device after the data path between the host and the at least one memory device of the memory pool is disabled, by using the encryption engine.

[0032] This provides the advange of ensuring data protection without requiring the hosts to enable, cofigure and manage the encryption of their own data, which enhances flexibility and efficiency. Morover, this is applicable to all CPUs.

[0033] In an implementation form of the first aspect, the fabric manager is further configured to enable the data path between the host and the at least one memory device of the memory pool through the encryption engine, and disable the data path between the host and the at least one memory device of the memory pool through the encryption engine.

[0034] In an implementation form of the first aspect, protecting the data of the host in the at least one memory device after the data path between the host and the at least one memory device of the memory pool is disabled, by using the encryption engine, includes controlling, by the memory controller, the encryption engine to encrypt the data of the host using the encryption key sent by the fabric manager, and providing the encrypted data of the host to the at least one memory device.

[0035] This provides a further manner to protect the host's data by encrypting it "on the fly" within the memory pool by means of the encyrption engine. Notably, the generated encryption key may be unique for each host, furhter enhancing the data protection.

[0036] In an implementation form of the first aspect, the at least one memory device is further configured to hold the encrypted data of the host.

[0037] According to a second aspect, a method for a disaggregated compute system is provided. The disaggregated compute system includes a fabric manager configured to communicate with a plurality of hosts and with a memory pool of the system, and the memory pool includes a memory controller and one or more memory devices, where the one or more memory devices are configured to hold data of the hosts. The method includes: enabling, by the fabric manager, a data path between any host and at least one of the memory devices of the memory pool; disabling, by the fabric manager, the data path between the host and the at least one memory device of the memory pool after the host writes data in the at least one memory device; and configuring, by the memory controller, the memory pool to protect the data of the host in the at least one memory device after the data path between the host and the at least one memory device of the memory pool is disabled.

[0038] This provides the advantage of ensuring that the rack or datacenter that provides the memory pooling service does not need to rely on the hosts for protecting the data in the one or more memory devices.

[0039] Furthermore, the disaggregated compute system can use any CPU and is not bound to products that provide the specific memory security features.

[0040] In an implementation form of the second aspect, the method further includes controlling, by the fabric manager, the memory pool to allocate a memory space for the host in the at least one memory device. The fabric manager may be a combined hardware-software system that may be configured to enable allocating the resources of the common memory pool and to enable connectivity of any host to the memory pool.

[0041] In an implementation form of the second aspect, after disabling the data path between the host and the at least one memory device of the memory pool after the host writes data in the at least one memory device, the method further includes enabling, by the fabric manager, a local data path between the memory controller and the at least one memory device of the memory pool.

[0042] This enables protection of the user's data without burdening the users and hosts with additional requirements and additional management.

[0043] In an implementation form of the second aspect, protecting the data of the host in the at least one memory device includes cleaning the memory space for the host allocated in the at least one memory device of the memory pool.

[0044] When switching a memory resource between different hosts, after detaching a host from the memory device of the memory pool and prior to connecting another host to the memory device, the host's data residing on the memory device may be cleaned, thus preventing the host's data to be subsequently accessed by any other host.

[0045] In an implementation form of the second aspect, cleaning the memory space for the host allocated in the at least one memory device of the memory pool includes erasing the memory space for the host allocated in the at least one memory device of the memory pool, alternatively overwriting the memory space for the host in the at least one memory device of the memory pool using random data.

[0046] This provides the advantage of enabling different manners to protect the host's data, enhancing the efficiency and flexibility.

[0047] In an implementation form of the second aspect, the method further includes: enabling the data path between the host and the at least one memory device of the memory pool through a fabric mesh, the fabric mesh being configured to communicate with the fabric manager, with the plurality of hosts and with the memory pool; and disabling the data path between the host and the at least one memory device of the memory pool through the fabric mesh.

[0048] That is, the connectivity of the hosts to the at least one memory device of the memory pool may be provided by the fabric mesh. The fabric mesh may comprise one or more physical implementations.

[0049] In an implementation form of the second aspect, the memory controller includes a memory host device and, after disabling the data path between the host and at least one memory device of the memory pool after the host writes data in the at least one memory device, the method further includes enabling, by the fabric manager, a data path between the memory host device and the at least one memory device through the fabric mesh.

[0050] The fabric mesh, thus, may further enable the connectivity of the memory host device to the at least one memory device within the memory pool.

[0051] In an implementation form of the second aspect, the memory controller includes a reset module host, and after disabling the data path between the host and the at least one memory device of the memory pool after the host writes data in the at least one memory device, the method further includes enabling, by the fabric manager, a data path between the reset module host and the at least one memory device through the fabric mesh.

[0052] The fabric mesh, thus, may further provide the connectivity of the reset module host to the at least one memory device within the memory pool.

[0053] In an implementation form of the second aspect, protecting the data of the host in the at least one memory device after the data path between the host and the at least one memory device of the memory pool is disabled includes performing a low level reset of the at least one memory device of the memory pool.

[0054] This provides a further manner to protect the host's data, further enhancing the efficiency and flexibility of the device. In an implementation form of the second aspect, the method further includes: generating, by the fabric manager, an encryption key; sending, by the fabric manager, the encryption key to the memory controller; changing, by the fabric manager, the encryption key after the host writes data in the at least one memory device and before the data path between the host and the at least one memory device of the memory pool is disabled; and protecting, by the memory controller, the data of the host in the at least one memory device after the data path between the host and the at least one memory device of the memory pool is disabled, by using an encryption engine, the encryption engine configured to communicate with the memory controller and with the one or more memory devices.

[0055] This provides the advange of ensuring data protection without requiring the hosts to enable, cofigure and manage the encryption of their own data, which enhances flexibility and efficiency. Morover, this is applicable to all CPUs.

[0056] In an implementation form of the second aspect, the method further includes: enabling, by the fabric manager, the data path between the host and the at least one memory device of the memory pool through the encryption engine; and disabling, by the fabric manager, the data path between the host and the at least one memory device of the memory pool through the encryption engine.

[0057] In an implementation form of the second aspect, protecting the data of the host in the at least one memory device after the data path between the host and the at least one memory device of the memory pool is disabled, by using the encryption engine, includes controlling, by the memory controller, the encryption engine to encrypt the data of the host using the encryption key sent by the fabric manager, and providing the encrypted data of the host to the at least one memory device.

[0058] This provides a further manner to protect the host's data by encrypting it "on the fly" within the memory pool by means of the encyrption engine. Notably, the generated encryption key may be unique for each host, furhter enhancing the data protection.

[0059] In an implementation form of the second aspect, the at least one memory device is further configured to hold the encrypted data of the host. The method according to the second aspect comprises the features of the corresponding implementation forms of the device of the first aspect.

[0060] According to a third aspect, a computer program product is provided, including a program code for carrying out, when implemented on a processor, the method according to the second aspect and its implementation forms.

[0061] The computer program product according to the third aspect comprises the features of the corresponding implementation forms of the method of the second aspect.

[0062] The method according to the second aspect and the computer program product according to the third aspect and their implementation forms provide the same advantages and effects as described above for the device of the first aspect and its respective implementation forms.

[0063] The solutions according to this disclosure address the security aspect from the memory pool perspective and may provide one or more of the following advantages:

[0064] • The solutions are applicable to all CPU.

[0065] • They have a negligible performance impact.

[0066] • The solutions do not require the host to enable, configure and manage advanced security features.

[0067] • The datacenter owner can guaranty memory isolation without any requirements from the hosts and his users.

[0068] It has to be noted that all devices, elements, units and means described in the present application could be implemented in the software or hardware elements or any kind of combination thereof. All steps which are performed by the various entities described in the present application as well as the functionalities described to be performed by the various entities are intended to mean that the respective entity is adapted to or configured to perform the respective steps and functionalities. Even if, in the following description of specific embodiments, a specific functionality or step to be performed by external entities is not reflected in the description of a specific detailed element of that entity which performs that specific step or functionality, it should be clear for a skilled person that these methods and functionalities can be implemented in respective software or hardware elements, or any kind of combination thereof.

[0069] BRIEF DESCRIPTION OF DRAWINGS

[0070] The above described aspects and implementation forms will be explained in the following description of specific embodiments in relation to the enclosed drawings, in which:

[0071] FIG. 1 schematically depicts a disaggregated compute system according to this disclosure;

[0072] FIG. 2 schematically depicts a disaggregated compute system according to this disclosure;

[0073] FIG. 3 schematically depicts a disaggregated compute system according to this disclosure;

[0074] FIG. 4 schematically depicts a disaggregated compute system according to this disclosure;

[0075] FIG. 5 schematically depicts a disaggregated compute system according to this disclosure;

[0076] FIG. 6 schematically depicts a disaggregated compute system according to this disclosure;

[0077] FIG. 7 schematically depicts a disaggregated compute system according to this disclosure;

[0078] FIG. 8 schematically depicts a disaggregated compute system according to this disclosure;

[0079] FIG. 9 schematically depicts a fabric manager according to this disclosure; FIG. 10 shows an example of the disaggregated compute system shown in FIG. 2;

[0080] FIG. 11 shows an example of the disaggregated compute system shown in FIG. 3;

[0081] FIG. 12 shows an example of the disaggregated compute system shown in FIG. 5;

[0082] FIGS. 13A-B show an example of the disaggregated compute system shown in FIG. 8;

[0083] FIG. 14 shows an example of an implementation of a disaggregated compute system according to this disclosure;

[0084] FIG. 15 shows a method for disaggregated compute system according to this disclosure;

[0085] FIG. 16 schematically depicts a conventional disaggregated compute system;

[0086] FIG. 17 schematically depicts a conventional disaggregated compute system;

[0087] DETAILED DESCRIPTION OF EMBODIMENTS

[0088] FIG. 1 shows an exemplary embodiment of a disaggregated compute system 100 according to this disclosure.

[0089] The disaggregated compute system 100 comprises a fabric manager 110, a plurality of hosts 120 and a memory pool 130.

[0090] The fabric manager 110 is configured to communicate with the plurality of hosts 120 and with the memory pool 130.

[0091] The memory pool 130 comprises a memory controller 131 and one or more memory devices 132. Each of the one or more memory devices 132 is configured to hold data of the hosts 120.

[0092] The one or more memory devices 132 may be implemented in, or may be part of, a memory module. The memory pool may be a separate module located outside of an enclosure of one or more of the plurality of hosts 120.

[0093] The fabric manager 110 is further configured to enable a data path 121 between any host 120 and at least one of the memory devices 132 of the memory pool 130. Further, after the host 120 writes data in the at least one memory device 132, the fabric manager 110 is configured to disable the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130.

[0094] The fabric manager 110 may be a combined hardware- software system that may enable allocating the resources of the memory pool 130, and may further enable connectivity to each host 120.

[0095] The memory controller 131 is configured to configure the memory pool 130 to protect the data of the host 120 in the at least one memory device 132 after the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 is disabled.

[0096] The fabric manager 110 is further configured to control the memory pool 130 to allocate a memory space for the host 120 in the at least one memory device 132. Thereby, the host 120 may be able to write the data in the at least one memory device 132.

[0097] After the host 120 writes data in the at least one memory device 132 and after the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 has been disabled, the fabric manager 110 is further configured to enable a local data path 133 between the memory controller 131 and the at least one memory device 132 of the memory pool 130.

[0098] In other words, the fabric manager 110 may enable the memory controller 131 to gain ownership on the at least one memory device 132 of the memory pool 130.

[0099] The memory controller 131 may use the local data path 133 to protect the data of the host 120 in the at least one memory device 132 of the memory pool 130. Protecting the data of the host 120, by the memory controller, in the at least one memory device 132 comprises cleaning the memory space for the host 120 allocated in the at least one memory device 132 of the memory pool 130.

[0100] The memory controller may clean the memory space for the host 120 allocated in the at least one memory device 132 of the memory pool 130 by using the local data path 133.

[0101] The cleaning of the memory space for the host 120 allocated in the at least one memory device 132 of the memory pool 130 comprises erasing the memory space for the host 120 allocated in the at least one memory device 132 of the memory pool 130.

[0102] Alternative, the cleaning of the memory space for the host 120 allocated in the at least one memory device 132 of the memory pool 130 comprises overwriting the memory space for the host 120 in the at least one memory device 132 of the memory pool 130 using random data. The random data may be generated by the memory controller 131.

[0103] The disaggregated compute system 100 according to this disclosure may comprise a processor or processing circuitry (not shown) configured to perform, conduct or initiate the various operations of the disaggregated compute system 100 described herein. The processing circuitry may comprise hardware and / or the processing circuitry may be controlled by software. The hardware may comprise analog circuitry or digital circuitry, or both analog and digital circuitry. The digital circuitry may comprise components such as application-specific integrated circuits (ASICs), field-programmable arrays (FPGAs), digital signal processors (DSPs), or multipurpose processors. The disaggregated compute system 100 may further comprise memory circuitry, which stores one or more instruction(s) that can be executed by the processor or by the processing circuitry, in particular under control of the software. For instance, the memory circuitry may comprise a non-transitory storage medium storing executable software code which, when executed by the processor or the processing circuitry, causes the various operations of the disaggregated compute system 100 to be performed. The processing circuitry may comprise one or more processors and a non-transitory memory connected to the one or more processors. The non-transitory memory may carry executable program code which, when executed by the one or more processors, causes the disaggregated compute system 100 to perform, conduct or initiate the operations or methods described herein. FIG. 2 shows an exemplary embodiment of the disaggregated compute system 100 according to this disclosure, which builds on the disaggregated compute system 100 shown in FIG. 1. Same elements are labelled with the same reference signs.

[0104] The afore-detailed description may be applied to the disaggregated compute system 100 of FIG. 2, except for the hereinafter-mentioned noticeable differences.

[0105] In the embodiment of FIG. 2, the disaggregated compute system 100 further comprises a fabric mesh 240. The fabric mesh 240 is configured to communicate with the fabric manager 110, with the plurality of hosts 120 and with the memory pool 130.

[0106] Further, the fabric mesh 240 is configured to enable the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 through the fabric mesh 240. Additionally, the fabric mesh 240 is configured to disable the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 through the fabric mesh 240.

[0107] That is, in the disaggregated compute system 100 of the embodiment of FIG. 2, the fabric manager 100 may be configured to enable the data path 121 between the host 120 and the at least one memory devices 132 through the fabric mesh 240.

[0108] Thereby, in this exemplary embodiment, the connectivity of the host 120 to the at least one memory device 132 may be via the fabric mesh 240, wherein the fabric mesh 240 may comprise one or more physical implementations.

[0109] Then, after the host 120 writes data in the at least one memory device 132, the fabric manager 100 may be configured to disable the data path 121 between the host 120 and the at least one memory device 132 through the fabric mesh 240.

[0110] The fabric mesh 240 according to this disclosure may comprise a processor or processing circuitry (not shown) configured to perform, conduct or initiate the various operations of the fabric mesh 240 described herein. The processing circuitry may comprise hardware and / or the processing circuitry may be controlled by software. The hardware may comprise analog circuitry or digital circuitry, or both analog and digital circuitry. The digital circuitry may comprise components such as application-specific integrated circuits (ASICs), field- programmable arrays (FPGAs), digital signal processors (DSPs), or multi-purpose processors. The fabric mesh 240 may further comprise memory circuitry, which stores one or more instruction(s) that can be executed by the processor or by the processing circuitry, in particular under control of the software. For instance, the memory circuitry may comprise a non-transitory storage medium storing executable software code which, when executed by the processor or the processing circuitry, causes the various operations of the fabric mesh 240 to be performed. The processing circuitry may comprise one or more processors and a non-transitory memory connected to the one or more processors. The non-transitory memory may carry executable program code which, when executed by the one or more processors, causes the fabric mesh 240 to perform, conduct or initiate the operations or methods described herein.

[0111] FIG. 3 shows an exemplary embodiment of the disaggregated compute system 100 according to this disclosure, which builds on the disaggregated compute system 100 shown in FIG. 2. Same elements are labelled with the same reference signs.

[0112] The afore-detailed description may be applied to the disaggregated compute system 100 of FIG. 3, except for the hereinafter-mentioned noticeable differences. In this exemplary embodiment, the memory controller 131 comprises a memory host device 331.

[0113] The enabling, by the fabric manager 110, the local data path 133 between the memory controller

[0114] 131 and the at least one memory device 132 of the memory pool 130 comprises enabling, by the fabric manager 110, a data path 333 between the memory host device 331 and the at least one memory device 132 through the fabric mesh 240.

[0115] The memory host device 331 may use the data path 333 to protect the data of the host 120 in the at least one memory device 132 of the memory pool 130.

[0116] Protecting the data of the host 120, by the memory controller, in the at least one memory device

[0117] 132 comprises cleaning the memory space for the host 120 allocated in the at least one memory device 132 of the memory pool 130.

[0118] The memory host device 331 may be configured to configure the memory pool 130 to protect the data of the host 120 in the at least one memory device 132 after the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 is disabled, by cleaning the memory space for the host 120 allocated in the at least one memory device 132.

[0119] The memory host device 331 may be configured to clean the memory space for the host 120 allocated in the at least one memory device 132 of the memory pool 130 by using the data path 333.

[0120] The cleaning of the memory space for the host 120 allocated in the at least one memory device 132 of the memory pool 130 comprises erasing the memory space for the host 120 allocated in the at least one memory device 132 of the memory pool 130.

[0121] Alternative, the cleaning of the memory space for the host 120 allocated in the at least one memory device 132 of the memory pool 130 comprises overwriting the memory space for the host 120 in the at least one memory device 132 of the memory pool 130 using random data. The random data may be generated by the memory host device 331.

[0122] FIG. 4 shows an exemplary embodiment of the disaggregated compute system 100 according to this disclosure, which builds on the disaggregated compute system 100 shown in FIG. 1. Same elements are labelled with the same reference signs.

[0123] The afore-detailed description of FIG. 1 may be applied to the disaggregated compute system 100 of FIG. 4, except for the hereinafter- mentioned noticeable differences. In this exemplary embodiment, the memory controller 131 comprises a reset module host 431.

[0124] The enabling, by the fabric manager 110, the local data path 133 between the memory controller 131 and the at least one memory device 132 of the memory pool 130 comprises enabling, by the fabric manager 110, a data path 433 between the reset module host 431 and the at least one memory device 132 of the memory pool 130.

[0125] Further, protecting the data of the host 120 in the at least one memory device 132 after the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 is disabled comprises performing, by the reset module host 431, a low level reset of the at least one memory device 132 of the memory pool 130. The reset module host 431 may be configured to perform the low level reset of the at least one memory device 132 of the memory pool 130 by using the data path 433.

[0126] In this exemplary embodiment, the one or more memory devices 132 may be further configured to support the low level reset performed by the reset module host 431.

[0127] FIG. 5 shows an exemplary embodiment of the disaggregated compute system 100 according to this disclosure, which builds on the disaggregated compute system 100 shown in FIG. 4. Same elements are labelled with the same reference signs.

[0128] The afore-detailed description may be applied to the disaggregated compute system 100 of FIG. 5, except for the hereinafter-mentioned noticeable differences. In this embodiment, the disaggregated compute system 100 further comprises the fabric mesh 240.

[0129] The fabric manager 100 may be configured to enable the data path 121 between the host 120 and the at least one memory device 132 through the fabric mesh 240, and to further disable the data path 121 between the host 120 and the at least one memory device 132 through the fabric mesh 240.

[0130] That is, the fabric mesh 240 is configured to enable the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 through the fabric mesh 240. Additionally, the fabric mesh 240 is configured to disable the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 through the fabric mesh 240.

[0131] Thereby, in this exemplary embodiment, the connectivity of the host 120 to the at least one memory device 132 may be via the fabric mesh 240, wherein the fabric mesh 240 may comprise one or more physical implementations.

[0132] Further, protecting the data of the host 120 in the at least one memory device 132 after the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 is disabled comprises performing, by the reset module host 431, the low level reset of the at least one memory device 132 of the memory pool 130. The reset module host 431 may be configured to perform the low level reset of the at least one memory device 132 of the memory pool 130 by using the data path 433, as disclosed above in the embodiment of FIG. 4. The details are not repeated again.

[0133] In this exemplary embodiment, the one or more memory devices 132 may be further configured to support the low level reset performed by the reset module host 431.

[0134] FIG. 6 shows an exemplary embodiment of the disaggregated compute system 100 according to this disclosure, which builds on the disaggregated compute system 100 shown in FIG. 5. Same elements are labelled with the same reference signs.

[0135] The afore-detailed description may be applied to the disaggregated compute system 100 of FIG. 6, except for the hereinafter-mentioned noticeable differences. In this exemplary embodiment, the enabling, by the fabric manager 110, the local data path 133 between the memory controller 131 and the at least one memory device 132 of the memory pool 130 comprises enabling a data path 633 between the reset module host 431 and the at least one memory device 132 of the memory pool 130 through the fabric mesh 240.

[0136] Thereby, in this embodiment, the fabric mesh 240 may enable the connectivity of the host 120 to the at least one memory device 132 as well as the connectivity of the reset module host 431 to the at least one memory device 132 of the memory pool 130.

[0137] Further, protecting the data of the host 120 in the at least one memory device 132 after the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 is disabled comprises performing, by the reset module host 431, the low level reset of the at least one memory device 132 of the memory pool 130.

[0138] The reset module host 431 may be configured to perform the low level reset of the at least one memory device 132 by using the data path 633.

[0139] In the exemplary embodiment of FIG. 6 the one or more memory devices 132 may be further configured to support the low level reset performed by the reset module host 431. FIG. 7 shows an exemplary embodiment of the disaggregated compute system 100 according to this disclosure, which builds on the disaggregated compute system 100 shown in FIG. 1. Same elements are labelled with the same reference signs.

[0140] The afore-detailed description may be applied to the disaggregated compute system 100 of FIG. 7, except for the hereinafter-mentioned noticeable differences. In this exemplary embodiment, the memory pool 130 further comprises an encryption engine 710 configured to communicate with the memory controller 131 and with the one or more memory devices 132.

[0141] The fabric manager 110 is further configured to generate an encryption key 711. Further, the fabric manager 110 sends the encryption key 711 to the memory controller 131. The generated encryption key 711 may be unique for the host.

[0142] Further, the fabric manager 110 is configured to enable the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 through the encryption engine 710.

[0143] The fabric manager 110 may be further configured to enable a data path 733 between the memory controller 131 and the encryption engine 710.

[0144] After the host 120 writes data in the at least one memory device 132, the fabric manager 110 is configured to disable the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 through the encryption engine 710.

[0145] Then, after the host 120 writes data in the at least one memory device 132 and before the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 is disabled, the fabric manager 110 is configured to change the encryption key 711.

[0146] The memory controller 131 is then configured to protect the data of the host 120 in the at least one memory device 132 after the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 is disabled, by using the encryption engine 710 and by using the encryption key 711 that is received from the fabric manager 110 before the fabric manager 110 changes the key. Further, the memory controller may be configured to protect the data of the host 120 in the at least one memory device 132 by using the encryption engine 710 through the data path 733.

[0147] In this exemplary embodiment, protecting the data of the host 120 in the at least one memory device 132 after the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 is disabled, by using the encryption engine 710, comprises controlling, by the memory controller 131, the encryption engine 710 to encrypt the data of the host 120 using the encryption key 711 sent by the fabric manager 110, and further providing the encrypted data of the host 120 to the at least one memory device 132.

[0148] Alternatively, the memory controller 131 may be configured to protect the data of the host 120 in the at least one memory device 132 by using the encrypting engine 710 while the host 120 writes the data.

[0149] That is, the host 120 may write data in the at least one memory device 132 by using the encryption engine 710, the encryption engine 710 being configured to protect the data of the host 120 by encrypting the data using the encryption key 711 sent by the fabric manager, and further providing the encrypted data of the host 120 to the at least one memory device 132.

[0150] After the host 120 writes the data, the fabric manager 110 may be configured to disable the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130. Next, the fabric manager 110 may be configured to change the encryption key 711.

[0151] Thereby, the host's data may be encrypted on the fly via the encryption engine 710.

[0152] Further, in the exemplary embodiment of FIG. 7, the at least one memory device 132 is configured to hold the encrypted data of the host 120.

[0153] FIG. 8 shows an exemplary embodiment of the disaggregated compute system 100 according to this disclosure, which builds on the disaggregated compute system 100 shown in FIG. 7. Same elements are labelled with the same reference signs. The afore-detailed description may be applied to the disaggregated compute system 100 of FIG. 8, except for the hereinafter-mentioned noticeable differences. The disaggregated compute system 100 shown in FIG. 8 further comprises the fabric mesh 240.

[0154] The fabric manager 110 is configured to enable the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 through the encryption engine 710 by using the fabric mesh 240.

[0155] Additionally, after the host 120 writes data in the at least one memory device 132, the fabric manager 110 is configured to disable the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 through the encryption engine 710 by using the fabric mesh 240.

[0156] FIG. 9 shows an exemplary embodiment of the fabric manager 110 comprised in the disaggregated compute system 100 according to this disclosure and shown in FIG. 1 to FIG. 8.

[0157] The fabric manager 110 according to this disclosure may comprise a processor or processing circuitry (not shown) configured to perform, conduct or initiate the various operations of the fabric manager 110 described herein. The processing circuitry may comprise hardware and / or the processing circuitry may be controlled by software. The hardware may comprise analog circuitry or digital circuitry, or both analog and digital circuitry. The digital circuitry may comprise components such as application-specific integrated circuits (ASICs), field- programmable arrays (FPGAs), digital signal processors (DSPs), or multi-purpose processors. The fabric manager 110 may further comprise memory circuitry, which stores one or more instruction(s) that can be executed by the processor or by the processing circuitry, in particular under control of the software. For instance, the memory circuitry may comprise a non-transitory storage medium storing executable software code which, when executed by the processor or the processing circuitry, causes the various operations of the fabric manager 110 to be performed. The processing circuitry may comprise one or more processors and a non-transitory memory connected to the one or more processors. The non-transitory memory may carry executable program code which, when executed by the one or more processors, causes the fabric manager 110 to perform, conduct or initiate the operations or methods described herein. The fabric manager 110 comprises a compute node control unit 9111 in communication with each of the plurality of hosts 120 of the disaggregated compute system 100. The compute node control unit 9111 may be configured to enable the data path 121 between any host 120 and at least one of the memory devices 132 of the memory pool 130.

[0158] Further, the compute node control unit 9111 may be configured to disable the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 after the host 120 writes data in the at least one memory device 132.

[0159] The fabric manager 110 further comprises a memory control unit 9113, configured to communicate with the memory pool 130. The memory control unit 9113 may be configured to control the memory pool 130 to allocate a memory space for the host 120 in the at least one memory device 132.

[0160] Further, the memory control unit 9113 may be configured to enable the local data path 133 between the memory controller 131 and the at least one memory device 132 of the memory pool 130.

[0161] The memory control unit 9113 may be further configured to generate the encryption key 711 and send the encryption key 711 to the memory controller 131. Further, the memory control unit 9113 may be configured to change the encryption key 711.

[0162] The memory control unit 9113 may be further configured to enable a data path 733 between the memory controller 131 and the encryption engine 710.

[0163] The fabric manager 110 may further comprise a mesh control unit 9112 configured to control the fabric mesh 240 to enable the data path 121 between any host 120 and the at least one memory device 132 of the memory pool 130 through the fabric mesh 240.

[0164] The mesh control unit 9112 may be further configured to control the fabric mesh 240 to disable the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 after the host 120 writes data in the at least one memory device 132 through the fabric mesh 240. Further, the mesh control unit 9112 may be configured to control the fabric mesh 240 to enable the data path data path 333 between the memory host device 331 and the at least one memory device 132.

[0165] The mesh control unit 9112 may be further configured to control the fabric mesh 240 to enable the data path data path 633 between the reset module host 431 and the at least one memory device 132.

[0166] FIG. 10 depicts an example of the disaggregated compute system 100 shown in FIG. 2. In this example, the disaggregated compute system 100 may comprise two hosts 120, exemplary Host A and Host B, the fabric manager 110, the memory pool 130 and the fabric mesh 240. As depicted in FIG. 10(a), the fabric manager 110 may configure Host A 120, the fabric mesh 240 and the memory pool 130 to enable the data path 121 between Host A 120 and a memory device 132 of the memory pool 130. Then, Host A 120 may write data to the memory device 132 on the memory pool 130.

[0167] Then, FIG. 10(b) depicts that the memory controller 131 may gain ownership on the memory device 132 as follows: The fabric manager 110 may configure Host A 120, the fabric mesh 240 and the memory pool 130 to disable (or disconnect) the data path 121 between Host A 120 and a memory device 132 on the memory pool 130. Next, the fabric manager 110 may configure the memory controller 131 to enable the local data path 133 within a memory pool enclosure, i.e., between the memory controller 131 and the memory device 132.

[0168] Next, an example depicting the cleansing of the memory device 132 is shown in FIG. 10(c). The memory controller 131 may use the data path (or data link) 133 to the memory device 132 to cleanse the data written by Host A in the memory device 132 by, for example, overwriting the entire memory space of the memory device 132. Thereby, the memory device 132 may be clean and without any residual data.

[0169] Then, the disaggregated compute system 100 may perform a new memory allocation, as depicted in FIG. 10(d). That is, the fabric manager 110 may configure another host 120, i.e., Host B, the fabric mesh 240 and the memory pool 130 to enable a data path 121 between Host B 120 and a memory device 132 on the memory pool 130. The memory device 132 may be, for example and not as a limitation, the same memory device 132 in which Host A has written data. Notably, as the memory device 132 is already clean, Host B 120 may not have access to data previously written to the memory by a different host, e.g., by Host A 120.

[0170] FIG. 11 depicts an example of the disaggregated compute system 100 shown in FIG. 3. In this example, the disaggregated compute system 100 may comprise two hosts 120, exemplary Host A and Host B, the fabric manager 110, the memory pool 130 and the fabric mesh 240. The memory controller 131 may comprise the memory host device 331. As depicted in FIG. 11(a), the fabric manager 110 may configure Host A 120, the fabric mesh 240 and the memory pool 130 to enable the data path 121 between Host A 120 and a memory device 132 of the memory pool 130. Then, Host A 120 may write data to the memory device 132 on the memory pool 130.

[0171] FIG. 11(b) depicts that the memory host device 331 may gain ownership of the memory device 132. That is, the fabric manager 110 may configure Host A 120, the fabric mesh 240 and the memory pool 130 to disable the data path 121 between Host A 120 and the memory device 132 on the memory pool 130. Then, the fabric manager 110 may configure the fabric mesh 240 and the memory pool 130 to enable the data path 333 between the memory host device 331 and the memory device 132.

[0172] Next, in the example of FIG. 11(c), the memory cleansing is depicted. The memory host device 331 may use the data path (or link) 133 to the memory device 132 to cleanse the memory device 132 by overwriting the entire memory space of the memory device 132. Thereby, the memory device 132 may be clean and without any residual data of Host A 120.

[0173] Then, the disaggregated compute system 100 may perform a new memory allocation, as depicted in FIG.11(d). That is, the fabric manager 1110 may configure another host 120, i.e., Host B, the fabric mesh 240 and the memory pool 130 to enable a data path 121 between Host B 120 and a memory device 132 on the memory pool 130. The memory device 132 may be, for example and not as a limitation, the same memory device 132 in which Host A has written data. Notably, as the memory device 132 may be already clean, Host B 120 may not have access to data previously written to the memory device 132 by a different host, e.g., by Host A 120.

[0174] FIG. 12 depicts an example of the disaggregated compute system 100 shown in FIG. 5. In this example, the disaggregated compute system 100 may comprise two hosts 120, exemplary Host A and Host B, the fabric manager 110, the memory pool 130 and the fabric mesh 240. The memory controller 131 may comprise the reset module host 431. As depicted in FIG. 12(a), the fabric manager 110 may configure Host A 120, the fabric mesh 240 and the memory pool 130 to enable the data path 121 between Host A 120 and a memory device 132 of the memory pool 130. Then, Host A 120 may write data to the memory device 132 on the memory pool 130.

[0175] FIG. 12(b) depicts that the reset module host 431 may gain ownership of the memory device 132. That is, the fabric manager 110 may configure Host A 120, the fabric mesh 240 and the memory pool 130 to disable the data path 121 between Host A 120 and the memory device 132 on the memory pool 130. Then, the fabric manager 110 may configure the fabric mesh 240 and the memory pool 130 to enable the data path 433 between the reset module host 431 and the memory device 132 through the fabric mesh 240.

[0176] Next, in the example of FIG. 12(c), the memory cleansing is depicted. The reset module host 431 may use the data path (or link) 633 to the memory device 132 to cleanse the memory device 132 by providing a dedicated low level reset of the memory device 132. Such a reset mechanism may be supported by the hardware of the memory device 132. Thereby, the memory device 132 may be clean and without any residual data of Host A 120.

[0177] Then, the disaggregated compute system 100 may perform a new memory allocation, as depicted in FIG. 12(d). The fabric manager 110 may configure another host 120, i.e., Host B, the fabric mesh 240 and the memory pool 130 to enable a data path 121 between Host B 120 and a memory device 132 on the memory pool 130. The memory device 132 may be, for example and not as a limitation, the same memory device 132 in which Host A has written data. Notably, as the memory device 132 may be already clean, Host B 120 may not have access to data previously written to the memory device 132 by a different host, e.g., by Host A 120.

[0178] FIGS. 13A and 13B depict an example of the disaggregated compute system 100 shown in FIG. 8. In this example, the disaggregated compute system 100 may comprise two hosts 120, exemplary Host A and Host B, the fabric manager 110, the memory pool 130 and the fabric mesh 240. The memory pool 130 may further comprise the encryption engine 710.

[0179] The top part of FIG. 13 A depicts that the fabric manager 110 may configure the memory pool 130 to enable a data path compression within the memory pool 130. That is, the fabric manager 110 may generate the encryption key 711 and may subsequently provide the encryption key 711 to the memory controller 131. This can be achieved, for example, as the disaggregated compute system 100 is powered on. Alternatively, the fabric manager 110 may generate a new encryption key for any new host 120 using the disaggregated compute system 100.

[0180] Next, the bottom part of FIG. 13 A depicts a manner in which Host A may write data to the memory module 132 on the memory pool 130. The fabric manager 110 may configure Host A 120, the fabric mesh 240 and the memory pool 130 to enable the data path 121 between Host A 120 and the memory device 132 on the memory pool 130 through the encryption engine 710. Then, Host A may write data to the memory device 132 on the memory pool 130 thorugh the encryption engine 710. That is, the data may be encrypted on the fly within the memory pool by means of the encyrption engine 710. Notably, the encryption key 711 is unique to this current host (or owner) 120 of the memory pool 130.

[0181] The top part of Fig.13B depicts that after Host A 120 writes the encrypted data in the memory device 132 of the memory pool 130, the fabric manager 110 may change the encryption key 711 before the data path 121 is disabled, thereby protecting the data from Host A.

[0182] Next, the disaggregated compute system 100 may perform a new memory allocation, as depicted in the bottom part of FIG. 13B. The fabric manager 110 may configure another host 120, i.e., Host B, the fabric mesh 240 and the memory pool 130 to enable a data path 121 between Host B 120 and the memory device 132 on the memory pool 130. Notably, since Host B lacks the proper encryption key 711 that was generated for Host A 120, Host B cannot decipher the data of Host A 120. Furhter, the fabric manager 110 may generate a new encryption key 711 and may assign it to Host B 120. Then, Host B 120 may use the new encryption key 711 to write data to the memory device 132 on the memory pool 130.

[0183] FIG. 14 shows an exemplary implementation of the disaggregated compute system 100 according to this disclosure. The disaggregated compute system 100 shown in FIG. 1 to FIG. 8 may be implemented with the memory pool 130 and with the fabric manager 110, which may orchestrate the host 120 connectivity and the factuality of the disaggregated compute system 100 as disclosed above.

[0184] FIG. 15 shows an exemplary embodiment of a method 1500 for a disaggregated compute system 100 according to this disclosure. The disaggregated compute system 100 comprising a fabric manager 110 configured to communicate with a plurality of hosts 120 and with a memory pool 130 of the system. The memory pool 130 comprises a memory controller 131 and one or more memory devices 132. The one or more memory devices 132 are configured to hold data of the hosts 120. The method 1500 may be performed by the exemplary embodiments of the disaggregated compute system 100 according to this disclosure.

[0185] The method 1500 comprises a step 1501 of enabling, by the fabric manager 110, a data path 121 between any host 120 and at least one of the memory devices 132 of the memory pool 130.

[0186] Then, the method 1500 comprises a step 1502 of disabling, by the fabric manager 110, the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 after the host 120 writes data in the at least one memory device 132.

[0187] Further, in step 1503, the method 1500 comprises configuring, by the memory controller 131, the memory pool 130 to protect the data of the host 120 in the at least one memory device 132 after the data path 121 between the host 120 and the at least one memory device 132 of the memory pool 130 is disabled.

[0188] The method 1500 may further comprise actions according to the exemplary embodiments of the disaggregated compute system 100 shown in FIG. 1, FIG. 2, FIG. 3, FIG. 4, FIG. 5, FIG. 6, FIG. 7 and FIG. 8. Hence, the method 1500 achieves the same advantages as the disaggregated compute system 100 disclosed above.

[0189] This disclosure further provides a computer program product comprising a program code for carrying out, when implemented on a processor, the method 1500. The computer program may be included in a computer readable medium of the computer program product. The computer readable medium may comprise essentially any memory, such as a ROM (Read-Only Memory), a PROM (Programmable Read-Only Memory), a 15 EPROM (Erasable PROM), a Flash memory, an EEPROM (Electrically Erasable PROM), or a hard disk drive.

[0190] The present disclosure has been described in conjunction with various embodiments as examples as well as implementations. However, other variations can be understood and effected by those persons skilled in the art and practicing the claimed matter, from the studies of the drawings, this disclosure and the independent claims. In the claims as well as in the description the word "comprising" does not exclude other elements or steps and the indefinite article "a" or "an" does not exclude a plurality. A single element or other unit may fulfill the functions of several entities or items recited in the claims. The mere fact that certain measures are recited in the mutual different dependent claims does not indicate that a combination of these measures cannot be used in an advantageous implementation.

Claims

CLAIMS1. A disaggregated compute system (100) comprising: a fabric manager (110) configured to communicate with a plurality of hosts (120) and with a memory pool (130) of the system, the memory pool (130) comprising a memory controller (131) and one or more memory devices (132), wherein the one or more memory devices (132) are configured to hold data of the hosts (120); wherein the fabric manager (110) is further configured to: enable a data path (121) between any host (120) and at least one of the memory devices(132) of the memory pool (130); and disable the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) after the host (120) writes data in the at least one memory device (132); and wherein the memory controller (131) is configured to: configure the memory pool (130) to protect the data of the host (120) in the at least one memory device (132) after the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) is disabled.

2. The disaggregated compute system (100) according to claim 1, wherein the fabric manager (110) is further configured to control the memory pool (130) to allocate a memory space for the host (120) in the at least one memory device (132).

3. The disaggregated compute system (100) according to one of the claims 1 or 2, wherein after disabling the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) after the host (120) writes data in the at least one memory device (132), the fabric manager (110) is further configured to: enable a local data path (133) between the memory controller (131) and the at least one memory device (132) of the memory pool (130).

4. The disaggregated compute system (100) according to one of the claims 1 to 3, wherein protecting the data of the host (120) in the at least one memory device (132) comprises: cleaning the memory space for the host (120) allocated in the at least one memory device5. The disaggregated compute system (100) according to claim 4, wherein cleaning the memory space for the host (120) allocated in the at least one memory device (132) of the memory pool (130) comprises: erasing the memory space for the host (120) allocated in the at least one memory device (132) of the memory pool (130); or overwriting the memory space for the host (120) in the at least one memory device (132) of the memory pool (130) using random data.

6. The disaggregated compute system (100) according to one of the preceding claims 1 to 5, further comprising a fabric mesh (240), the fabric mesh (240) being configured to communicate with the fabric manager (110), with the plurality of hosts (120), and with the memory pool (130), and the fabric mesh (240) is further configured to: enable the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) through the fabric mesh (240); and disable the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) through the fabric mesh (240).

7. The disaggregated compute system (100) according to one of the claims 1 to 6, wherein the memory controller (131) comprises a memory host device (331); and wherein, after disabling the data path (121) between the host (120) and at least one memory device (132) of the memory pool (130) after the host (120) writes data in the at least one memory device (132), the fabric manager (110) is further configured to: enable a data path (333) between the memory host device (331) and the at least one memory device (132) through the fabric mesh (240).

8. The disaggregated compute system (100) according to one of the claims 1 to 6, wherein the memory controller (131) comprises a reset module host (431); and wherein, after disabling the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) after the host (120) writes data in the at least one memory device (132), the fabric manager (110) is further configured to: enable a data path (633) between the reset module host (431) and the at least one memory device (132) through the fabric mesh (240).

9. The disaggregated compute system (100) according to claim 8, wherein protecting the data of the host (120) in the at least one memory device (132) after the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) is disabled comprises: performing a low level reset of the at least one memory device (132) of the memory pool (130).

10. The disaggregated compute system (100) according to claim 1, wherein the fabric manager (110) is further configured to: generate an encryption key (711) and send the encryption key (711) to the memory controller (131); and change the encryption key (711) after the host (120) writes data in the at least one memory device (132) and before the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) is disabled; wherein the memory pool (130) further comprises an encryption engine (710) configured to communicate with the memory controller (131) and with the one or more memory devices (132); and wherein the memory controller (131) is further configured to: protect the data of the host (120) in the at least one memory device (132) after the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) is disabled, by using the encryption engine (710).

11. The disaggregated compute system (100) according to claim 10, wherein the fabric manager (110) is further configured to: enable the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) through the encryption engine (710); and disable the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) through the encryption engine (710).

12. The disaggregated compute system (100) according to one of the claims 10 or 11, wherein protecting the data of the host (120) in the at least one memory device (132) after the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) is disabled, by using the encryption engine (710), comprises:controlling, by the memory controller (131), the encryption engine (710) to encrypt the data of the host (120) using the encryption key (711) sent by the fabric manager (110); and providing the encrypted data of the host (120) to the at least one memory device (132).

13. The disaggregated compute system (100) according to one of the claims 10 to 12, wherein the at least one memory device (132) is further configured to hold the encrypted data of the host (120).

14. A method (1500) for a disaggregated compute system (100), the disaggregated compute system (100) comprising a fabric manager (110) configured to communicate with a plurality of hosts (120) and with a memory pool (130) of the system, the memory pool (130) comprising a memory controller (131) and one or more memory devices (132), wherein the one or more memory devices (132) are configured to hold data of the hosts (120); wherein the method (1500) comprises: enabling (1501), by the fabric manager (110), a data path (121) between any host (120) and at least one of the memory devices (132) of the memory pool (130); disabling (1502), by the fabric manager (110), the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) after the host (120) writes data in the at least one memory device (132); and configuring (1503), by the memory controller (131), the memory pool (130) to protect the data of the host (120) in the at least one memory device (132) after the data path (121) between the host (120) and the at least one memory device (132) of the memory pool (130) is disabled.

15. A computer program product comprising a program code for carrying out, when implemented on a processor, the method (1500) according to claim 14.

Citation Information

Patent Citations

  • End-to-end secure data storage in a dispersed storage network

    US20170123907A1

  • Service mesh offload to network devices

    US20210243247A1

  • Confidential computing environment for service mesh on a network interface device

    US20220329573A1

  • Management method and management system for computer system

    WO2015107676A1

Cited By

  • Memory controller, operating method thereof, and storage device for ensuring initialization of security parameter

    US12530286B2

  • Memory controller, operating method thereof, and storage device for ensuring initialization of security parameter

    US20250225069A1