Cloud desktop virtual network equipment and network card straight-through on-demand hybrid forwarding method

Abstract

The invention provides a cloud desktop virtual network device and network card straight-through on-demand hybrid forwarding method, belongs to the technical field of cloud desktop and virtualization, and realizes a flexible flow scheduling strategy on a bonding aggregation port through statistics and network quality analysis of user network messages based on aggregation binding of an openvswitch virtual network card and a sriov straight-through network card. On one hand, based on bandwidth requirements, large-bandwidth flow is guided to a high-performance straight-through network card, and low-bandwidth flow is guided to an ovs virtual switch network card; and on the other hand, based on network quality parameters such as packet loss rate and retransmission rate, the traffic which is high in packet loss rate and retransmission rate and urgently needs to be guaranteed preferentially is switched to the straight-through network card to be processed. Therefore, the network forwarding performance is dynamically optimized, and on-demand allocation of resources and improvement of user experience are realized.

Description

Technical Field

[0001] The present invention relates to the fields of cloud desktop and virtualization technology, and in particular to a cloud desktop virtual network device and a network card direct-through on-demand hybrid forwarding method. Background Art

[0002] Currently, common virtualized network technologies on cloud platforms fall into two main categories: network forwarding based on the OpenVSwitch virtual switch and network forwarding based on hard-pass through network adapters (VFs) or PFs. These two approaches differ in forwarding performance, with hard-pass through outperforming OpenVSwitch. OpenVSwitch virtual switch forwarding typically uses DPDK to optimize the forwarding process to improve performance. However, compared to hard-pass through network adapters, DPDK still cannot meet the requirements for high-bandwidth, high-performance, and low-latency network forwarding.

[0003] When complex network requirements exist in production networks, traditional virtualized network forwarding methods cannot flexibly address diverse traffic forwarding requirements. For example, when a production network contains both high-bandwidth traffic, such as video streaming, which requires high real-time performance, and low-bandwidth traffic, such as text and image access, which requires lower real-time performance, forwarding methods with better forwarding performance should be prioritized for high-real-time, high-bandwidth traffic, while lower-real-time, low-bandwidth traffic should be forwarded using methods with lower forwarding performance. However, in current scenarios, users use hard-pass solutions to uniformly handle both types of traffic to ensure network service quality, resulting in wasted network forwarding resources and increased platform operating costs. Alternatively, users use OpenVSWITCH to forward both types of traffic, causing network service delays and lags, impacting the customer experience. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides a cloud desktop virtual network device and a network card direct on-demand hybrid forwarding method. On the one hand, it can be applied to scenarios where different forwarding methods are used to cope with different forwarding requirements, and on the other hand, it can provide a capability to enhance network quality. By simultaneously binding the openvswitch virtualized network card and the sriov hard-pass network card in an aggregation manner, statistics, network quality analysis and diversion for user network messages are achieved. According to network statistics, large bandwidth traffic is distributed to the sriov hard-pass network card for high-performance forwarding, and low bandwidth traffic is distributed to the openvswitch virtualized network card for ordinary forwarding; ensuring that network forwarding capabilities are allocated on demand. According to network quality analysis, traffic with higher retransmission rate and packet loss rate is distributed to the sriov hard-pass network card for priority processing, and traffic with lower retransmission rate and packet loss rate is distributed to the openvswitch virtualized network card for ordinary forwarding; ensuring that messages that need to be forwarded are quickly processed and optimizing user experience. Based on the above two solutions, network forwarding performance optimization is achieved.

[0005] The technical solution of the present invention is:

[0006] A method for on-demand hybrid forwarding of cloud desktop virtual network devices and network cards. This method aggregates SRIOV and OVS network cards through bonding to form a single logical interface. A bonding diversion scheduling scheme based on traffic statistics prioritizes high-bandwidth traffic forwarding through the SRIOV network card. A bonding diversion scheduling scheme based on network quality analysis prioritizes high-loss and high-latency traffic forwarding through the SRIOV network card. Dynamic optimization and closed-loop control ensure that high-bandwidth traffic is automatically downgraded back to the OVS channel after forwarding, and the high-bandwidth threshold is dynamically adjusted.

[0007] Furthermore, the construction of a hybrid NIC aggregation interface involves: using the Linux kernel bonding driver to bind the physical Sriov NIC and the OVS-based virtual NIC into the same logical bonding group; configuring a unified MAC address and IP address for the bonding group to provide external services; and implementing automatic switching in the event of a NIC failure through a heartbeat detection mechanism.

[0008] Furthermore, intelligent traffic diversion based on traffic statistics includes: building a bandwidth statistics model in the aggregation port, and real-time statistics of session bandwidth according to source / destination IP and transport layer five-tuple; setting dynamic bandwidth thresholds based on historical traffic analysis to identify large-bandwidth sessions; implementing diversion strategies in the Bonding scheduling logic, and accurately matching the large-traffic session packets with bandwidth exceeding the threshold identified in real time according to their source / destination IP and port, and forcibly directing their traffic to the Sriov direct-through network card in the Bonding group for processing.

[0009] Build a traffic monitoring system, deploy traffic sampling probes on aggregation ports, and use the sFlow / IPFIX protocol for traffic sampling.

[0010] Furthermore, sensitive flow acceleration based on network quality assessment includes: combining active network detection with passive traffic analysis mechanisms; comprehensively evaluating the network forwarding quality of each session; implementing a quality-driven diversion strategy in the Bonding-driven scheduling logic, and for low-quality traffic sessions with high packet loss, high latency, and high jitter, directing their packets to the Sriov direct-pass network card in the Bonding group; and utilizing the hardware acceleration capabilities of the Sriov network card to process such sensitive traffic, significantly reducing its network transmission delay.

[0011] Furthermore, the two-layer closed-loop dynamic scheduling and optimization mechanism includes: implementing the state migration function. When the high-bandwidth or low-quality session diverted to the sriov pass-through network card ends, its subsequent traffic is automatically downgraded and migrated back to the OVS virtual network card channel for processing; real-time monitoring of the utilization rate of the sriov pass-through channel; dynamic adjustment of the diversion strategy parameters according to the current network card load status; and implementation of the sriov channel overload protection mechanism.

[0012] When the SR-IOV channel utilization is greater than 80%, only high-risk traffic is allowed to pass through to achieve overload protection.

[0013] The beneficial effects of the present invention are

[0014] Based on real-time traffic statistics, it intelligently directs high-bandwidth traffic to the high-performance Sriov network interface card (NIC) and offloads low-bandwidth traffic to Openvsiwtch. Furthermore, based on analysis of network quality indicators, it dispatches quality-sensitive, critical traffic that urgently needs forwarding to the more reliable Sriov forwarding path. This overcomes the limitations of traditional single forwarding modes and can simultaneously meet the differentiated needs of different business traffic flows. While improving the quality of critical business services, it also avoids the waste of high-cost direct-through resources for all traffic, optimizing overall computing and network resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the workflow of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0017] This invention proposes an on-demand hybrid forwarding method for virtual network cards (NICs) and direct-through NICs. Traditional virtualized network forwarding methods lack the flexibility to address diverse traffic forwarding requirements. This invention proposes a virtual network device and NIC direct-through on-demand hybrid forwarding technology. This technology is applicable to scenarios where different forwarding methods are used to address diverse forwarding requirements and also provides a means to enhance network quality. By simultaneously binding an OpenvSwitch virtualized NIC and a Sriov hard-pass NIC in an aggregated manner, user network message statistics, network quality analysis, and traffic diversion are implemented. Based on network statistics, high-bandwidth traffic is distributed to the Sriov hard-pass NIC for high-performance forwarding, while low-bandwidth traffic is distributed to the OpenvSwitch virtualized NIC for standard forwarding, ensuring on-demand allocation of network forwarding capacity. Based on network quality analysis, traffic with high retransmission and packet loss rates is prioritized by the Sriov hard-pass NIC, while traffic with low retransmission and packet loss rates is distributed to the OpenvSwitch virtualized NIC for standard forwarding. This ensures rapid processing of urgently needed messages and optimizes the user experience. Based on these two solutions, network forwarding performance is optimized.

[0018] This invention achieves intelligent traffic diversion by constructing a logical aggregation interface between the OpenvSwitch virtualized network card and the Sriov hard-wired network card, combining dynamic traffic statistics with a network quality analysis engine. A unified logical interface is created to bind the two types of network cards to form a hybrid network card aggregation. Traffic monitoring is achieved by collecting metrics such as bandwidth, retransmission rate, and packet loss rate in real time. A diversion strategy is generated based on traffic characteristics and network quality analysis. Millisecond-level traffic scheduling is achieved through aggregated network port content scheduling. This allows high-bandwidth traffic to be directly passed through SR-IOV to ensure performance, high-packet-loss traffic to be prioritized and passed through to reduce latency, and ordinary traffic to be distributed using the OVS solution.

[0019] Step 1: Create a hybrid NIC aggregate interface

[0020] Based on the Linux kernel bonding driver, the SR-IOV VF network card and the OVS virtual network card are added to the same logical bonding group, and a unified MAC address and IP address are set for external exposure to shield the differences in the underlying network cards. Automatic switching of network card failures is achieved through the heartbeat detection mechanism.

[0021] Step 2: Traffic diversion strategy based on traffic statistics

[0022] Build a traffic monitoring system and deploy traffic sampling probes on aggregation ports. Use sFlow / IPFIX protocols for traffic sampling. Establish a bandwidth statistics model to calculate BPS by source / destination IP address and quintuple. Set dynamic thresholds and define maximum bandwidth based on historical traffic analysis.

[0023] The traffic diversion strategy is implemented by using the monitoring system to identify high-bandwidth sessions in real time. This traffic is then diverted within the bonding scheduling logic. Packets with bandwidth exceeding the threshold are matched based on source, destination IP, and port numbers. These high-bandwidth packets are then directed to the bonding group's Sriov passthrough network card. The hardware passthrough network card directly processes the packets, bypassing the kernel protocol stack.

[0024] Step 3: Traffic diversion strategy based on network quality analysis

[0025] Combining active probing results with passive analysis implements a network quality-driven traffic diversion mechanism. Passive analysis captures TCP header information and calculates retransmission and packet loss rates; active probing sends probe packets to the destination IP address to calculate RTT and jitter. A traffic health scoring model is established to assess network forwarding quality.

[0026] The traffic diversion strategy is implemented by the quality analysis engine, which flags low-quality traffic sessions. This is then diverted within the bonding scheduling logic, directing packets with high-risk traffic characteristics to the Sriov passthrough network interface card (NIC) in the bonding group. The Sriov NIC processes sensitive traffic through hardware acceleration, reducing network latency.

[0027] Step 4: Dynamic Scheduling and Strategy Optimization

[0028] A two-layer control loop is established to implement positive feedback and protection mechanisms. State migration is implemented, automatically downgrading sessions to OVS channels when high-bandwidth flows end. Traffic diversion thresholds are dynamically adjusted based on NIC load. When SR-IOV channel utilization exceeds 80%, only high-risk traffic is allowed through, achieving overload protection.

[0029] The above description is only a preferred embodiment of the present invention and is only used to illustrate the technical solution of the present invention, and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention are included in the scope of protection of the present invention.

Claims

1. A cloud desktop virtual network device and network card direct on-demand hybrid forwarding method, characterized in that: Bonding aggregates the SRIOV and OVS network cards to form a single logical interface; the bonding diversion scheduling solution based on traffic statistics prioritizes forwarding of high-bandwidth traffic through the SRIOV network card; the bonding diversion scheduling solution based on network quality analysis prioritizes forwarding of high-packet loss and high-latency traffic through the SRIOV network card; dynamic optimization and closed-loop control automatically downgrade high-bandwidth traffic to the OVS channel after forwarding is completed, and dynamically adjusts the high-bandwidth threshold.

2. The method according to claim 1, characterized in that Use the Linux kernel Bonding driver to bind the physical sriov network card and the OVS-based virtual network card to the same logical Bonding group; configure a unified MAC address and IP address for the Bonding group to provide external services.

3. The method according to claim 2, characterized in that Automatic switching of network card failures is achieved through the heartbeat detection mechanism.

4. The method according to claim 1, wherein Build a bandwidth statistics model on the aggregation interface to calculate session bandwidth in real time based on source / destination IP and transport layer quintuples. Set dynamic bandwidth thresholds based on historical traffic analysis to identify high-bandwidth sessions. Implement a traffic diversion strategy in the Bonding scheduling logic. For high-traffic session packets with bandwidth exceeding the threshold identified in real time, perform precise matching based on their source / destination IP and port, and forcibly direct their traffic to the Sriov direct-through network card in the Bonding group for processing.

5. The method according to claim 4, characterized in that Build a traffic monitoring system, deploy traffic sampling probes on aggregation ports, and use the sFlow / IPFIX protocol for traffic sampling.

6. The method according to claim 1, wherein This approach combines active network detection with passive traffic analysis mechanisms; comprehensively evaluates the network forwarding quality of each session; and implements a quality-driven traffic diversion strategy in the Bonding-driven scheduling logic. For low-quality traffic sessions with high packet loss, high latency, and high jitter, their packets are directed to the Sriov direct-connect network card within the Bonding group. The Sriov network card's hardware acceleration capabilities are used to process such sensitive traffic, reducing its network transmission latency.

7. The method according to claim 1, characterized in that Implement the state migration function. When a high-bandwidth or low-quality session diverted to the sriov passthrough network card ends, its subsequent traffic will be automatically downgraded and migrated back to the OVS virtual network card channel for processing; monitor the utilization of the sriov passthrough channel in real time; dynamically adjust the diversion policy parameters according to the current network card load status; and implement the overload protection mechanism of the sriov channel.

8. The method according to claim 7, characterized in that When the SR-IOV channel utilization is greater than 80%, only high-risk traffic is allowed to pass through to achieve overload protection.

Images