Method and system for multi-service multiplexing of RDMA high-performance network on virtualized host

Abstract

The invention discloses an RDMA high-performance network multi-service multiplexing method and system on a virtualized host. The method comprises the following operation steps that network service data programmed through a Verbs interface by using an RDMA protocol and network service data programmed through a socket interface by using a TCP / IP protocol are transmitted at the same time; and performing dynamic balance operation of each level on the network service data, transmitting the network service data subjected to the dynamic balance operation of each level to a physical network card, packaging the network service data to obtain an Ethernet message, and sending the Ethernet message to a physical network.

Description

Technical Field

[0001] The present invention relates to the field of network allocation, and in particular to a method and system for multiplexing RDMA high-performance network multiple services on a virtualized host. Background Art

[0002] High-performance network is a technology that eliminates the increase in delay and degradation of network quality caused by congestion or packet loss during data transmission in the network. High-performance network can ensure real-time and high bandwidth during data transmission.

[0003] It provides a guarantee of network reliability for the stable operation of services that have special requirements for high bandwidth and low latency.

[0004] Currently, when using high-performance networks, the mainstream mostly uses the RoCE (RDMA over ConvergedEthernt) protocol. The RoCE protocol is divided into RoCEv1 and RoCEv2. RoCEv1 is implemented based on the link layer (Layer 2 network) and cannot be deployed across network segments. RoCEv2 is based on the network layer and has good adaptability to existing services and cross-segment deployment capabilities. In the RoCE protocol, the transport layer uses the InfiniBand protocol, which is an RDMA technology that can be implemented based on the existing Ethernet. It needs to be supported by a physical network card that supports the protocol. The protocol encapsulation and decapsulation during network data transmission are implemented by the network card hardware.

[0005] However, when services with special requirements for network performance and ordinary services are running on the server at the same time, it is necessary to configure an independent physical network card that supports high-performance RDMA protocol for the high-performance network service, and deploy one or more non-RDMA physical network cards for ordinary services.

[0006] If ordinary services and services requiring high-performance networks are simply mixed and run on a high-bandwidth network card that supports the RDMA protocol, bandwidth competition will occur between services requiring high-performance networks and ordinary services, which may cause services requiring high-performance networks to have excessive latency and bandwidth cannot be guaranteed, resulting in unstable business operations. Summary of the invention

[0007] The purpose of the present invention is to provide a method and system for multiplexing RDMA high-performance network multiple services on a virtualized host, which solves the above technical problems pointed out in the prior art.

[0008] The present invention provides a method for multiplexing RDMA high-performance network multiple services on a virtualized host, comprising the following steps: Simultaneously transmit network service data programmed using the RDMA protocol through the Verbs interface and network service data programmed using the TCP / IP protocol through the socket interface; Dynamic balancing operations are performed on the network service data at various levels, and the network service data transmission physical network card after the dynamic balancing operations at various levels is encapsulated to obtain Ethernet messages, and the Ethernet messages are sent to the physical network.

[0009] Preferably, the network service data includes: high-performance network required service data and non-high-performance network required service data.

[0010] Preferably, the non-high-performance network required service data includes: ordinary non-high-performance network service data and network service data of a virtual machine.

[0011] Preferably, the dynamic balancing operations at each level include: initial bandwidth allocation, real-time traffic monitoring, bandwidth usage analysis, and dynamic feedback balancing.

[0012] Preferably, the network service data is transmitted to a physical network card, and the physical network card performs dynamic balancing operations at various levels on the network service data to obtain an Ethernet message. The specific operation steps are as follows: Performing initial bandwidth allocation according to the type of the network service data, allocating different bandwidth ratios according to different types of network service data, and using the remaining bandwidth ratio as reserved bandwidth; For the types of the allocated different network service data, the traffic usage of the physical network card is monitored, RDMA traffic and ordinary TCP traffic, to obtain real-time traffic monitoring data; Preset time interval e; calculate the bandwidth usage ratio of the physical network card for the RDMA traffic and the ordinary TCP traffic of the real-time traffic monitoring data according to the preset time interval e; The bandwidth ratios of different types of network service data are adjusted by adjusting the bandwidth usage ratio and the ratio of the initial bandwidth allocation, thereby obtaining a dynamic balancing operation of the network service data.

[0013] Preferably, the bandwidth ratios of different types of network service data are adjusted by adjusting the bandwidth usage ratio and the ratio of the initial bandwidth allocation, so as to obtain a dynamic balancing operation of the network service data. The specific operation steps are as follows: Determining whether to adjust the bandwidth ratio of different types of network service data according to the ratio of the bandwidth usage ratio to the initial bandwidth allocation; If the bandwidth usage ratio of the high-performance network required service data of the RDMA traffic is greater than the initial bandwidth allocation ratio, then the reserved bandwidth ratio is reallocated to the high-performance network required service data of the RDMA traffic; If the bandwidth usage ratio of the high-performance network-required service data of the RDMA traffic is less than the initial bandwidth allocation ratio, the reserved bandwidth ratio is used to reallocate the bandwidth ratio to the non-high-performance network-required service data of the common TCP traffic; Preset an initial bandwidth usage ratio threshold t and an initial bandwidth usage ratio threshold y; The initial bandwidth usage ratio threshold y is greater than the initial bandwidth usage ratio threshold t; If the bandwidth usage ratio of the high-performance network required business data of the RDMA traffic is less than the initial bandwidth usage ratio threshold y and greater than the initial bandwidth usage ratio threshold t, the reserved bandwidth ratio is reallocated to the high-performance network required business data of the RDMA traffic; If the bandwidth usage ratio of the high-performance network required service data of the RDMA traffic is greater than the initial bandwidth usage ratio threshold y, the reserved bandwidth ratio is reallocated to the high-performance network required service data of the RDMA traffic.

[0014] Accordingly, the present invention also proposes a system for multiplexing RDMA high-performance network services on a virtualized host, comprising: a programming module; a dynamic balancing module; The programming module is used to simultaneously transmit network service data programmed using the RDMA protocol through the Verbs interface and network service data programmed using the TCP / IP protocol through the socket interface; The dynamic balancing module is used to transmit the network service data to the physical network card, and the physical network card performs dynamic balancing operations at various levels on the network service data to obtain Ethernet messages, and sends the Ethernet messages to the physical network.

[0015] Correspondingly, the present application also relates to a storage medium, in which a computer program is stored. When the computer program is executed by a processor, the steps of a method for multiplexing RDMA high-performance network multiple services on a virtualized host are implemented.

[0016] Compared with the prior art, the embodiments of the present invention have at least the following technical advantages: By analyzing the above-mentioned method of multiplexing multiple services of RDMA high-performance network on a virtualized host provided by the present invention, it can be known that in the scenario where both high-performance applications are programmed using the RDMA protocol through the Verbs interface and ordinary services are programmed using the TCP / IP protocol through the Socket interface, high-performance network service data and non-high-performance network service data are directly transmitted to the network card. Through the Verbs interface, applications can control data transmission in a more fine-grained manner, including high-performance data streams (such as HPC, storage, database, etc.) and ordinary network traffic (such as ordinary TCP / IP traffic, virtual machine traffic, etc.); high-performance services can be directly transmitted through RDMA, reducing the intervention of CPU and memory, and improving data transmission speed and efficiency.

[0017] Furthermore, the physical network card is responsible for "dynamically balancing" the data traffic according to different types of business requirements. Specifically, the physical network card processes network business data based on bandwidth allocation strategies, real-time traffic monitoring, bandwidth usage analysis, and dynamic feedback balancing mechanisms, and finally encapsulates the data into Ethernet packets and sends them to the physical network.

[0018] Furthermore, appropriate bandwidth resources are initially allocated based on the priorities of different types of services (for example, high-performance services are usually given higher priority and bandwidth); the physical network card monitors the real-time transmission of each type of traffic, checks bandwidth usage, and whether there is network congestion or insufficient bandwidth; the system analyzes the bandwidth usage efficiency of each type of traffic, identifies which services occupy too much bandwidth and which services fail to fully utilize the bandwidth, thereby providing a basis for subsequent adjustments; and dynamically adjusts bandwidth allocation based on traffic monitoring and bandwidth analysis results.

[0019] For example, if a certain type of service (such as high-performance service) requires more bandwidth, the system can adjust the bandwidth allocation according to actual needs, avoiding bandwidth resource waste or bottlenecks. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 An overall flow chart of a method for multiplexing RDMA high-performance network multiple services on a virtualized host provided in Embodiment 1 of the present invention; Figure 2A flow chart of network service data allocation for a method for multiplexing RDMA high-performance network services on a virtualized host provided in Embodiment 1 of the present invention; Figure 3 A flowchart of dynamic feedback balancing of a method for multiplexing RDMA high-performance network multiple services on a virtualized host provided in Embodiment 1 of the present invention; Figure 4 A flowchart of a system for multiplexing RDMA high-performance network multiple services on a virtualized host provided in Embodiment 2 of the present invention; Figure 5 The present invention is a schematic diagram of the structure of a storage medium to which the above-mentioned method for multiplexing RDMA high-performance network multiple services on a virtualized host is applied.

[0022] Reference numerals: programming module 10 ; dynamic balancing module 20 ; processor 1110 ; communication interface 1120 ; memory 1130 ; storage medium 1140 . DETAILED DESCRIPTION

[0023] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments.

[0024] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.

[0025] The present invention is further described in detail below through specific embodiments in conjunction with the accompanying drawings.

[0026] Embodiment 1 like Figure 1 As shown, the present invention proposes a method for multiplexing RDMA high-performance network multiple services on a virtualized host, comprising the following steps: S1: Simultaneously transmit network service data programmed through the Verbs interface using the RDMA protocol (i.e., RoCEv2 protocol) and network service data programmed through the socket interface using the TCP / IP protocol (i.e., service data to be processed); The network service data includes: high-performance network required service data and non-high-performance network required service data; The non-high-performance network required service data includes: ordinary non-high-performance network service data and network service data of virtual machines; S2: Perform dynamic balancing operations on the network service data at each level, and encapsulate the network service data transmission physical network card after the dynamic balancing operations at each level (i.e., a physical network card that supports the RoCEv2 protocol) to obtain an Ethernet message, and send the Ethernet message to the physical network.

[0027] The dynamic balancing operations at each level include: initial bandwidth allocation, real-time traffic monitoring, bandwidth usage analysis, and dynamic feedback balancing; It should be noted that RoCEv2 is an Ethernet-based remote direct memory access (RDMA) protocol that allows efficient data exchange between different hosts over the network, avoiding frequent interference with the CPU and memory during data transmission.

[0028] The RoCEv2 protocol is mainly used in application scenarios such as high-performance computing (HPC), storage, and big data analysis.

[0029] Encapsulation and processing of physical network cards: When business data reaches a physical network card that supports RoCEv2, the physical network card is responsible for encapsulating the data according to the protocol requirements.

[0030] This means that the physical network card will encapsulate the RDMA data sent by the application layer into an Ethernet message with an IP header and an Ethernet header according to the RoCEv2 protocol, and send it to the destination host through the physical network.

[0031] This process is automatically completed by the network card hardware, greatly reducing the burden on the host CPU.

[0032] Verbs interface In the RoCEv2 protocol, applications communicate with the network card through the Verbs interface. The Verbs interface is a low-level programming interface that enables applications to directly access RDMA capabilities, send data, and control data flows.

[0033] Specifically, Figure 2 As shown, in step S2, dynamic balancing operations at various levels are performed on the network service data, and the specific operation steps are as follows: S21: performing initial bandwidth allocation according to the type of the network service data, allocating different bandwidth ratios according to different types of network service data, and using the remaining bandwidth ratio as reserved bandwidth; It should be noted that on a high-performance physical network card that supports RDMA, first, initial bandwidth allocation is performed for traffic of different types of network service data (such as service data requiring high-performance networks, service data requiring non-high-performance networks, and reserved bandwidth (that is, used for dynamic adjustment and increasing the proportion of network service data with too low bandwidth)).

[0034] For example, set 50% bandwidth for high-performance network services, 40% bandwidth for ordinary network services, and 10% bandwidth as reserved bandwidth (for dynamic adjustment); S22: monitoring the RDMA traffic and the common TCP traffic of the physical network card for the types of the allocated different network service data, and obtaining real-time traffic monitoring data; It should be noted that the data of RDMA traffic and ordinary TCP traffic are collected by real-time monitoring of the traffic usage of the physical network card (ie, monitoring the usage of business data requiring high-performance networks and business data requiring non-high-performance networks).

[0035] The traffic data will be used for subsequent bandwidth usage analysis and adjustment.

[0036] RDMA traffic is an efficient network communication technology (i.e., high-performance network required business data) that allows the memory of a computer system to interact directly with the memory of a remote computer without intervention from the operating system or CPU.

[0037] The advantage of RDMA is that it can significantly reduce latency and improve throughput, and is particularly suitable for applications that require fast data transmission (such as high-performance computing, storage access, distributed databases, etc.); Ordinary TCP traffic is a connection-oriented, reliable transport protocol (i.e., non-high-performance network-required business data) that is widely used in computer networks to ensure the correct order and integrity of data.

[0038] Unlike RDMA, TCP traffic is managed and processed by the operating system kernel; TCP implements flow control and congestion control to ensure that the network does not become overly congested under high load conditions.

[0039] S23: Preset time interval e; calculate the bandwidth usage ratio of the physical network card for the RDMA traffic (i.e., high-performance network required business data) and the ordinary TCP traffic (i.e., non-high-performance network required business data) of the real-time traffic monitoring data according to the preset time interval e; S24: adjusting the bandwidth ratio of different types of network service data by the bandwidth usage ratio and the initial bandwidth allocation ratio (i.e., whether the bandwidth ratio of service data required by the high-performance network is greater than the bandwidth ratio of service data required by the non-high-performance network, thereby determining whether to allocate bandwidth to the bandwidth ratio of service data required by the high-performance network from the reserved bandwidth ratio), thereby obtaining a dynamic balancing operation of network service data; It should be noted that the preset time interval e (eg, 60 seconds) calculates the bandwidth usage (ie, bandwidth usage ratio) of RDMA traffic and common traffic on the network card at regular intervals (eg, 60 seconds) based on real-time traffic monitoring data.

[0040] This bandwidth usage ratio, together with the initial bandwidth allocation ratio, is used to determine whether bandwidth adjustment is required.

[0041] Specifically, Figure 3 As shown, in step S24, the bandwidth ratios of different types of network service data are adjusted by the bandwidth usage ratio and the initial bandwidth allocation ratio, so as to obtain a dynamic balancing operation of the network service data. The specific operation steps are as follows: S241: judging whether to adjust the bandwidth ratios of different types of network service data according to the ratio of the bandwidth usage ratio and the ratio of the initial bandwidth allocation; S242: if the bandwidth usage ratio of the high-performance network required service data of the RDMA traffic is greater than the initial bandwidth allocation ratio, reallocate the bandwidth ratio of the reserved bandwidth to the high-performance network required service data of the RDMA traffic; It should be noted that if the actual bandwidth usage of RDMA traffic exceeds the initial bandwidth (that is, exceeds the bandwidth ratio initially allocated, exceeding 50% of the allocation), it indicates that the RDMA traffic demand has increased, and the system will allocate more bandwidth from the reserved bandwidth to meet the RDMA traffic demand.

[0042] S243: if the bandwidth usage ratio of the high-performance network-required service data of the RDMA traffic is less than the initial bandwidth allocation ratio, reallocate the bandwidth ratio of the reserved bandwidth to the non-high-performance network-required service data of the common TCP traffic; It should be noted that if the bandwidth usage of RDMA traffic is less than the initial bandwidth (that is, the high-performance network of RDMA traffic requires that the bandwidth allocation ratio of business data is less than 80% of the bandwidth allocated at the beginning), and the bandwidth usage of ordinary services has exceeded 90%, a part of the reserved bandwidth will be allocated to ordinary network services to avoid too low bandwidth for ordinary services.

[0043] At this time, the high-performance network bandwidth still maintains the initial allocation ratio.

[0044] S244: Preset an initial bandwidth usage ratio threshold t (i.e., initial bandwidth usage 80%) and an initial bandwidth usage ratio threshold y (i.e., initial bandwidth usage 90%); The initial bandwidth usage ratio threshold y (i.e., initial bandwidth usage 90%) is greater than the initial bandwidth usage ratio threshold t (i.e., initial bandwidth usage 80%); If the bandwidth usage ratio of the high-performance network required business data of the RDMA traffic is less than the initial bandwidth usage ratio threshold y and greater than the initial bandwidth usage ratio threshold t, the reserved bandwidth ratio is reallocated to the high-performance network required business data of the RDMA traffic; It should be noted that if the RDMA traffic bandwidth usage is greater than 80% but less than 90%, and the ordinary network service bandwidth usage is less than 90%, part of the reserved bandwidth will be allocated to high-performance services (RDMA), and the ordinary service bandwidth will be restored to its initial state to avoid excessive bandwidth occupation by ordinary services.

[0045] S245: if the bandwidth usage ratio of the high-performance network required service data of the RDMA traffic is greater than the initial bandwidth usage ratio threshold y, reallocate the bandwidth ratio of the reserved bandwidth to the high-performance network required service data of the RDMA traffic; It should be noted that if the RDMA traffic bandwidth usage exceeds 90%, it means that the bandwidth demand of high-performance services has reached the upper limit.

[0046] At this time, if the adjusted bandwidth is lower than the initial bandwidth, the system will allocate all the bandwidth from the reserved bandwidth to high-performance services and restore the bandwidth of ordinary services to the initial state to ensure the stability of the high-performance network.

[0047] Embodiment 2 like Figure 4 As shown, accordingly, the present invention also proposes a system for multiplexing RDMA high-performance network multiple services on a virtualized host, comprising: a programming module 10; a dynamic balancing module 20; The programming module 10 is used to simultaneously transmit network service data programmed using the RDMA protocol through the Verbs interface and network service data programmed using the TCP / IP protocol through the socket interface; The dynamic balancing module 20 is used to perform dynamic balancing operations on the network service data at each level, and to encapsulate the network service data transmission physical network card after the dynamic balancing operations at each level to obtain Ethernet messages, and to send the Ethernet messages to the physical network.

[0048] Embodiment 3 On the other hand, this third embodiment is based on the method for multiplexing RDMA high-performance network multiple services on a virtualized host provided in the first embodiment of the invention, and further provides a computer storage medium 1140 (storage medium for short).

[0049] like Figure 5 FIG. 1 is a schematic diagram of a computer storage medium structure framework provided in Embodiment 3 of the present invention, which includes: A memory 1130, used for storing computer programs; The communication interface 1120 is used to connect the memory 1130 to the processor 1110; The processor 1110 is used to execute a computer program to implement a method for multiplexing RDMA high-performance network multiple services on a virtualized host, which is involved in an embodiment disclosed in combination with any of the above implementation modes.

[0050] In summary, the present invention proposes a method and system for multiplexing RDMA high-performance network multiple services on a virtualized host, which directly transmits high-performance network service data and non-high-performance network service data to the network card in a scenario where high-performance applications use the RDMA protocol to program through the Verbs interface and ordinary services use the TCP / IP protocol to program through the Socket interface.

[0051] Through the Verbs interface, applications can control data transmission in a more fine-grained manner, including high-performance data streams (such as HPC, storage, databases, etc.) and ordinary network traffic (such as ordinary TCP / IP traffic, virtual machine traffic, etc.); high-performance services can be transmitted directly through RDMA, reducing the intervention of the CPU and memory, and improving data transmission speed and efficiency.

[0052] Furthermore, the physical network card is responsible for encapsulating the data traffic in a "dynamically balanced" manner according to different types of business requirements.

[0053] Specifically, the physical network card processes network service data according to mechanisms such as bandwidth allocation strategy, real-time traffic monitoring, bandwidth usage analysis, and dynamic feedback balancing, and finally encapsulates the data into Ethernet packets and sends them to the physical network.

[0054] Furthermore, appropriate bandwidth resources are initially allocated based on the priorities of different types of services (for example, high-performance services are usually given higher priority and bandwidth); the physical network card monitors the real-time transmission of each type of traffic, checks bandwidth usage, and whether there is network congestion or insufficient bandwidth; the system analyzes the bandwidth usage efficiency of each type of traffic, identifies which services occupy too much bandwidth and which services fail to fully utilize the bandwidth, thereby providing a basis for subsequent adjustments; and dynamically adjusts bandwidth allocation based on traffic monitoring and bandwidth analysis results.

[0055] For example, if a certain type of service (such as high-performance service) requires more bandwidth, the system can adjust the bandwidth allocation according to actual needs, avoiding bandwidth resource waste or bottlenecks.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. A person skilled in the art may modify the technical solutions described in the above embodiments, or replace part or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for multiplexing RDMA high-performance network services on a virtualized host, characterized in that: The steps are as follows: Simultaneously transmit network service data programmed using the RDMA protocol through the Verbs interface and network service data programmed using the TCP / IP protocol through the socket interface; Dynamic balancing operations are performed on the network service data at various levels, and the network service data transmission physical network card after the dynamic balancing operations at various levels is encapsulated to obtain Ethernet messages, and the Ethernet messages are sent to the physical network.

2. According to claim 1, a method for multiplexing RDMA high-performance network services on a virtualized host, characterized in that: The network service data includes: high-performance network required service data and non-high-performance network required service data.

3. According to claim 2, a method for multiplexing RDMA high-performance network services on a virtualized host, characterized in that: The non-high-performance network required service data includes: ordinary non-high-performance network service data and network service data of virtual machines.

4. According to claim 3, a method for multiplexing RDMA high-performance network services on a virtualized host, characterized in that: The dynamic balancing operations at each level include: initial bandwidth allocation, real-time traffic monitoring, bandwidth usage analysis, and dynamic feedback balancing.

5. According to claim 4, a method for multiplexing RDMA high-performance network services on a virtualized host, characterized in that: Perform dynamic balancing operations on the network service data at each level. The specific operation steps are as follows: Performing initial bandwidth allocation according to the type of the network service data, allocating different bandwidth ratios according to different types of network service data, and using the remaining bandwidth ratio as reserved bandwidth; For the types of the allocated different network service data, the traffic usage of the physical network card is monitored, RDMA traffic and ordinary TCP traffic, to obtain real-time traffic monitoring data; Preset time interval e; Calculate the bandwidth usage ratio of the physical network card for the RDMA traffic and the common TCP traffic of the real-time traffic monitoring data according to the preset time interval e; The bandwidth ratios of different types of network service data are adjusted by adjusting the bandwidth usage ratio and the ratio of the initial bandwidth allocation, thereby obtaining a dynamic balancing operation of the network service data.

6. According to claim 5, a method for multiplexing RDMA high-performance network services on a virtualized host, characterized in that: The bandwidth ratios of different types of network service data are adjusted by adjusting the bandwidth usage ratio and the initial bandwidth allocation ratio, thereby obtaining a dynamic balancing operation of network service data. The specific operation steps are as follows: Determining whether to adjust the bandwidth ratio of different types of network service data according to the ratio of the bandwidth usage ratio to the initial bandwidth allocation; If the bandwidth usage ratio of the high-performance network required service data of the RDMA traffic is greater than the initial bandwidth allocation ratio, then the reserved bandwidth ratio is reallocated to the high-performance network required service data of the RDMA traffic; If the bandwidth usage ratio of the high-performance network-required service data of the RDMA traffic is less than the initial bandwidth allocation ratio, the reserved bandwidth ratio is used to reallocate the bandwidth ratio to the non-high-performance network-required service data of the common TCP traffic; Preset an initial bandwidth usage ratio threshold t and an initial bandwidth usage ratio threshold y; The initial bandwidth usage ratio threshold y is greater than the initial bandwidth usage ratio threshold t; If the bandwidth usage ratio of the high-performance network required business data of the RDMA traffic is less than the initial bandwidth usage ratio threshold y and greater than the initial bandwidth usage ratio threshold t, the reserved bandwidth ratio is reallocated to the high-performance network required business data of the RDMA traffic; If the bandwidth usage ratio of the high-performance network required service data of the RDMA traffic is greater than the initial bandwidth usage ratio threshold y, the reserved bandwidth ratio is reallocated to the high-performance network required service data of the RDMA traffic.

7. A system for multiplexing RDMA high-performance network services on a virtualized host, comprising: Programming module; Dynamic balancing module; The programming module is used to simultaneously transmit network service data programmed through the Verbs interface using the RDMA protocol and network service data programmed through the socket interface using the TCP / IP protocol; The dynamic balancing module is used to perform dynamic balancing operations on the network service data at each level, and to encapsulate the network service data transmission physical network card after the dynamic balancing operations at each level to obtain Ethernet messages, and to send the Ethernet messages to the physical network.

8. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the steps of the method for multiplexing RDMA high-performance network multiple services on a virtualized host are implemented as described in any one of claims 1 to 6.

Images