Data center high-speed switch testing method for cross-domain heterogeneous distributed learning

Through the data center high-speed switch testing method for cross-domain heterogeneous distributed learning, covering basic functions, operation and maintenance functions, performance and reliability testing, the problem that existing testing methods cannot fully evaluate 400G switches, and realize the 400G switches comprehensive performance and reliability evaluation.

CN118972297APending Publication Date: 2024-11-15CHINA ACADEMY OF INFORMATION & COMM
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410721049.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-05
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing high-speed switch testing methods for data centers cannot be comprehensively tested from multiple dimensions, resulting in the test results that cannot truly reflect actual needs. Especially in a cross-domain heterogeneous distributed learning environment, there is a lack of testing methods for 400G switches.

Method used

A high-speed switch testing method for cross-domain heterogeneous distributed learning is proposed, covering basic functional testing, operation and maintenance functional testing, performance testing and reliability testing, including optical module information reading, Fec testing, port splitting, and 1588 clock Test steps and expected results in multiple aspects such as synchronization function, BFD fast switching function, BGP EVPN Vxlan forwarding, IPv6 over IPv4 Vxlan forwarding, ARP broadcast suppression function, cross-chassis link bundling function, and fault scenario testing.

Benefits of technology

Through all-round and multi-angle testing, the comprehensive capability evaluation indicators of 400G switches are provided to ensure that the performance evaluation meets actual needs and can truly reflect the comprehensive performance and reliability of the switch.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118972297A_ABST
    Figure CN118972297A_ABST
Patent Text Reader

Abstract

The invention discloses a data center high-speed switch testing method oriented to cross-domain heterogeneous distributed learning, and relates to the technical field of data center switch testing, and the data center high-speed switch testing method oriented to cross-domain heterogeneous distributed learning comprises the following steps: S1, testing items of all dimensions; s2, basic function testing; s3, testing operation and maintenance functions; s4, performance testing; and S5, reliability testing. According to the cross-domain heterogeneous distributed learning-oriented data center high-speed switch testing method, project testing is carried out from multiple dimensions, and in the process of carrying out normative testing on a cross-domain heterogeneous distributed learning data center 400G switch, testing and result provision are carried out aiming at various performances; therefore, the performance evaluation of the 400G switch serving the cross-domain heterogeneous distributed learning data center can really approach the actual demand.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of data center switch testing, and in particular to a data center high-speed switch testing method oriented to cross-domain heterogeneous distributed learning. Background Art

[0002] Data centers are the core support platform for cloud computing. The development of cloud computing has posed severe challenges to the network architecture of data centers. With the progress of digitalization, data processing, storage and transmission have developed rapidly. The demand for high bandwidth has made short-distance interconnection a bottleneck for system development. 400G optical modules mainly complete the PMD function of 400G Ethernet (or include some PMA layer functions) to ultimately realize the transmission of electrical signals on optical media. With the expansion of network scale, especially with the development of cloud services, east-west traffic has increased sharply, accounting for about 70% of the total traffic. This has led to an increasing number of network nodes within data centers and an increasing variety of interconnection port types. The number of optical modules used in data centers has increased exponentially, and the impact of optical interconnection on the stability of network systems has increased day by day. Since the performance of 400G optical modules has a huge impact on the performance of network systems, the testing of switches has great significance and value for actual use. Data center switch standards are becoming more and more mature, but there is currently a lack of test methods for testing high-speed switches in data centers that serve cross-domain heterogeneous distributed learning, which makes the evaluation of high-speed switches difficult.

[0003] Currently, existing patents have proposed performance testing methods for switches and provided testing devices, such as patent CN202311026312.5, a method for testing terminal devices based on testing switches and testing switch networking. This patent provides a method for testing terminal devices based on testing switches and testing switch networking, provides test steps and expected results, obtains key indicators, and conducts a comprehensive performance test on the switch. However, the patent does not test from multiple dimensions, resulting in the test results not being truly close to actual needs.

[0004] Therefore, in view of this, the existing structure and deficiencies are studied and improved, and a data center high-speed switch testing method for cross-domain heterogeneous distributed learning is proposed. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a data center high-speed switch testing method for cross-domain heterogeneous distributed learning, which solves the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention is implemented by the following technical scheme: a data center high-speed switch testing method for cross-domain heterogeneous distributed learning, the data center high-speed switch testing method for cross-domain heterogeneous distributed learning comprising the following steps:

[0007] S1. Test items in each dimension:

[0008] Dimensions include basic function testing, operation and maintenance function testing, performance testing, and reliability testing;

[0009] Among them, the basic function test includes optical module information reading; Fec test; port splitting; 1588 clock synchronization function; BFD fast switching function; BGP EVPN Vxlan forwarding; IPv6 over IPv4 Vxlan forwarding; ARP broadcast suppression function; cross-chassis link bundling function test; cross-chassis link bundling fault scenario test;

[0010] S2. Basic function test:

[0011] Test 1: Optical module information reading test. Its purpose is to verify that the device supports the optical port optical power reading function. The test steps are as follows: For the 400GE interface, the DUTs are connected through 400G optical modules. When the interface is UP, read the optical module information and get the expected result 1.

[0012] Shutdown the interface and read the optical module information again, with the expected result 2.

[0013] Use snmp to read the optical module information, and the expected result is 3;

[0014] Test 2: Fec test, the purpose of which is to test DUT's support for RS-FEC function. The test steps are as follows: TC1 and DUT-1 are interconnected through 5m DAC, and RS-FEC is turned on on TC1 side; there is expected result 1;

[0015] RS-fec is enabled on all links, and TC1 and TC2 exchange 128-byte 400G line-speed traffic for one hour, with the expected result 2.

[0016] Test 3: Port split test, the purpose of which is to verify the port splitting capability. The test steps are as follows: DUT is for a 400GE interface, and the test port can be split into 4*100GE ports. The split does not require a board or device reset to take effect. After the split, the tester is connected to 100GE, and the tester verifies the traffic with the expected result 1;

[0017] Test 4: 1588 clock synchronization function test, the purpose of the test is to test whether the device can support the 1588 clock synchronization function. The test steps are as follows: globally enable PTP on DUT1, configure the device type to BC, and configure the local priority 1 to 10; globally enable PTP on DUT2, configure the device type to BC, enable 1588 synchronization, and configure the clock source priority to 2;

[0018] Enable PTP on the interconnection interface between DUT-1 and DUT-2, and the expected result 1 can be obtained;

[0019] Test 5: BFD fast switching function test, the purpose of the test is to test whether the device can support the BFD fast switching function. The test steps are as follows: configure the interconnection interface of DUT1 and DUT2 as a Layer 3 interface, and configure IP addresses respectively;

[0020] Configure static routes and priorities on DUT_1 and DUT_2. The interconnection link of Port_2 with a high priority is the primary link, and the interconnection link of Port_3 with a low priority is the backup link.

[0021] Enable BFD for DUT1 and DUT2, associate with static routing, configure the BFD check time to 3ms, and the number of detections to 3. Check the BFD session, and you will see the expected result 1.

[0022] Send test traffic from port Tport_1 to port Tport_2, with expected result 2;

[0023] Disconnect the main link of Port_2 between DUT_1 and DUT_1, and the expected result is 3;

[0024] Restore the main link of Port_2 between DUT_1 and DUT_2, and get the expected result 4;

[0025] Repeat steps 3 and 4 for Expected Result 3 and Expected Result 4 for 3 times, and record the packet loss time.

[0026] Test 6: BGP EVPN Vxlan forwarding test, the purpose of the test is to test whether the device can support BGP EVPNVxlan function. The test steps are as follows: configure the interface IP address on DUT_1 and DUT_2, configure the IGP routing protocol, publish the route of the interface IP address and the route of the loopback address, and ensure the three-layer interconnection of the network; configure EVPNBGP peers and vxlan tunnels on DUT_1 and DUT_2 devices, and get the expected result 1;

[0027] Disconnect the interconnection link between DUT1 and DUT2, and the expected result 2 is obtained;

[0028] Restore the disconnected link, configure tenant 1 gateway and other configurations on DUT_1 and DUT_2 devices, and send Layer 2 and Layer 3 IPv4 traffic on the tester Tport_1 and Tport_2 ports, with the expected result 3;

[0029] Disconnect the main link of Port_2 between DUT_1 and DUT_1, and the expected result is 3;

[0030] Configure the tenant 2 gateway and other configurations on the DUT_1 and DUT_2 devices. The tester Tport_1 and Tport_2 ports send cross-tenant IPv4 traffic, and the expected result 4 is obtained;

[0031] Test 7: IPv6 over IPv4 Vxlan forwarding test. The purpose of the test is to test whether the device can support the IPv6 over IPv4 Vxlan function. The test steps are as follows: configure the interface IP address on DUT_1 and DUT_2, configure the IGP routing protocol, publish the route of the interface IP address and the route of the loopback address, and ensure the three-layer interconnection of the network; configure the EVPN BGP peer and vxlan tunnel on the DUT_1 and DUT_2 devices, and get the expected result 1;

[0032] Disconnect the interconnection link between DUT1 and DUT2, and get the expected result 1; configure IPv6 over IPv4 configurations such as tenant 1 gateway on DUT_1 and DUT_2 devices, and send Layer 3 IPv6 traffic on the tester Tport_1 and Tport_2 ports, and get the expected result 2;

[0033] Configure IPv6over IPv4 configurations such as the tenant 2 gateway on DUT_1 and DUT_2 devices, and the tester Tport_1 and Tport_2 ports send cross-tenant IPv4 traffic, with expected result 3;

[0034] Test 8: ARP broadcast suppression function test, the purpose of the test is to test whether the device can support the ARP broadcast suppression function. The test steps are as follows: configure the interface IP address on DUT_1 and DUT_2, configure the IGP routing protocol, publish the route of the interface IP address and the route of the loopback address, and ensure the three-layer interconnection of the network; configure EVPN BGP peers and vxlan tunnels on DUT_1 and DUT_2 devices, and get the expected result 1; the tester Tport_1 and Tport_2 ports create virtual machines VM1 and VM2 in the same network segment, ping the gateway, and capture packets on VM2, and get the expected result 2; enable the ARP broadcast suppression function on DUT_3 and DUT_4, clear the arp table entries of VM1 and VM2, VM1 pings VM2, and captures packets on VM2, and get the expected result 3; the tester Tport_1 sends a line-speed arp request message to the Port_1 interface, and gets the expected result 4;

[0035] Test 9: Cross-chassis link bundling function test, the purpose of the test is to test whether the device can support the cross-chassis link bundling function. The test steps are as follows: configure the cross-chassis link bundling function on DUT1 and DUT2, and configure the L2 aggregation in lacp mode on DUT3 and DUT4, and the expected result 1 is obtained; configure EVPN BGP peers and vxlan tunnels on DUT_1 and DUT_2 devices, and the expected result 1 is obtained;

[0036] DUT1 and DUT2 devices are configured with two gateway IP addresses, and the interconnected ports and bundled ports are added to the same VLAN. The tester TC1 and TC2 ports send layer 3 traffic with a changed source IP, and the expected result 2 is obtained.

[0037] Disconnect the interconnection link between DUT1 and DUT3, and the expected result 3 is obtained;

[0038] Restore the disconnected link, with expected result 4;

[0039] Test 10: Cross-chassis link bundling failure scenario test. The purpose of the test is to test whether the device can support the switching and recovery functions of the cross-chassis link bundling failure scenario. The test steps are as follows: configure the cross-chassis link bundling function on DUT1 and DUT2, and configure the L2 aggregation in the lacp mode on DUT3 and DUT4, and the expected result 1 is obtained; configure EVPN BGP peers and vxlan tunnels on DUT_1 and DUT_2 devices, and the expected result 1 is obtained;

[0040] DUT1 and DUT2 devices are configured with two gateway IP addresses, and the interconnected ports and bundled ports are added to the same VLAN respectively; the tester TC1 and TC2 ports send layer 3 traffic with a changed source IP, and the expected result 2 is obtained; the interconnection link between DUT1 and DUT3 is disconnected, and the expected result 3 is obtained;

[0041] Restore the disconnected link, with expected result 4;

[0042] S3, operation and maintenance function test:

[0043] Test 11: Packet loss visualization function test. The purpose of the test is to test whether the device can support the packet loss visualization function. The test steps are as follows: configure IPv4 addresses on the interfaces connecting the DUT to the test instrument, and send traffic between TC1 and TC2, with the expected result 1;

[0044] The DUT enables the Buffer packet loss visualization function and configures the Server as a collector. TC1 and TC2 send line-speed traffic to TC3 to create a port congestion and packet loss scenario, which produces the expected result 2.

[0045] Test 12: Ultra-long delay visualization function test. The purpose of the test is to test whether the device can support the ultra-long delay visualization function. The test steps are as follows: configure IPv4 addresses on the interfaces connecting the DUT to the test instrument, and send traffic between TC1 and TC2, with the expected result 1;

[0046] Enable the latency visualization function on the DUT, configure the latency threshold, and configure the server as a collector. TC1 and TC2 send line-speed traffic to TC3 to create a scenario where port congestion increases latency, and the expected result is 2.

[0047] Test 13: Full-flow analysis function test. The purpose of the test is to test whether the device can support the full-flow analysis function. The test steps are as follows: configure IPv4 addresses on the interfaces of the DUT connected to the test instrument, and send TCP / UDP traffic between TC1 and TC2, with the expected result 1;

[0048] The DUT enables the full-flow analysis service and configures the Server as a collector, with the expected result 2.

[0049] Configure ACL to match the sent TCP / UDP traffic, and get the expected result 3;

[0050] Test 14: ioam flow detection function test, the purpose of the test is to test whether the device can support ioam flow detection function, the test steps are as follows: 1) DUT1, DUT2, DUT3 are configured as three-layer network interconnection, TC1 and TC2 gateways are configured on DUT1 and DUT2, TC1 sends TCP traffic to TC2, and the expected result 1 is obtained;

[0051] Enable the ioam function on DUT1, DUT2, and DUT3. Configure ACL matching traffic on DUT1 for sampling. Configure an IPv4 address on the interface connected to the server on DUT2, which is reachable to the server network. Send ioam packets to the analyzer. TC1 sends TCP traffic to TC2, with expected result 2.

[0052] S4. Performance test:

[0053] Test 15: MAC table entry rule test, the purpose of which is to test the maximum number of MAC address table entries that the device can support. The test steps are as follows: Test instrument TC1 inputs smac changing traffic, the number of changes is the number of mac address table entries declared by the DUT, and the expected result is 1;

[0054] TC2 enters the traffic whose dmac is the learned mac address table entry, and the expected result is 2;

[0055] Test 16: ARP table specification test, the purpose of the test is to test the maximum number of ARP table entries that the device can support. The test steps are as follows: 1) Configure IPv4 and IPv6 addresses on the interface connecting the DUT to the test instrument respectively; the test instrument TC1 enters the ARP reply message of the declared ARP specification at line speed, and the expected result is 1;

[0056] TC2 injects verification traffic with the destination address being the learned ARP table entry, and the expected result is 2.

[0057] The test instrument TC1 inputs the RA message of the claimed nd specification at line speed, and the expected result is 3;

[0058] TC2 injects verification traffic with the destination address being the learned ND entry, and the expected result is 4.

[0059] Test 17: FIB table rule test, the purpose of the test is to test the maximum number of FIB table items that the device can support. The test steps are as follows: DUT and TC1 establish an EBGP neighbor relationship; TC1 enters the BGP route of the FIB specification announced by the device, and the expected result is 1;

[0060] TC2 sends traffic with the destination address being the published route, with expected result 2.

[0061] Establish an EBGPv6 neighbor relationship between DUT and TC1, and repeat the steps of expected result 1 and expected result 2 using IPv6 routing;

[0062] Test 18: Full-frame performance forwarding test, the purpose of the test is to test the full-frame performance that the device can support. The test steps are as follows: The DUT is fully plugged with single boards, and all interfaces of the same type are connected as shown in the figure above. The first interface and the last interface are connected to the test instrument. The two interfaces are configured in the same VRF, and each VRF is configured with static routing to achieve forwarding; the RFC2544 test suite of the test instrument is used to test the IPv4 forwarding performance, with the expected result 1;

[0063] The dual-stack forwarding performance was tested using the RFC2544 test suite of the test instrument, with IPv4 and IPv6 traffic accounting for 50% each, and the expected result was 1; the "double snake" traversal test was performed on both the uplink and downlink ports at the same time;

[0064] Test 19: Equipment resource utilization test, the purpose of which is to test the full frame performance that the equipment can support. The test steps are as follows: On the basis that the equipment can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result is 1;

[0065] After the flow rate stabilizes, observe the equipment operation status and get the expected result 2;

[0066] S5. Reliability test:

[0067] Test 20: Main control board redundancy test, the purpose of the test is to test whether the device can support main control board redundancy. The test steps are as follows: on the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result 1 is obtained; unplug the main control board of the DUT, and the expected result 2 is obtained;

[0068] Test 21: Power redundancy test, the purpose of the test is to test whether the device can support power redundancy. The test steps are as follows: On the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result 1 is obtained; unplug a redundant power supply, and the expected result 2 is obtained;

[0069] Test 22: Fan redundancy test, the purpose of which is to test whether the device can support fan redundancy. The test steps are as follows: On the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result 1 is obtained; unplug a redundant fan of the DUT, and the expected result 2 is obtained;

[0070] Test 23: Switching network board redundancy test, the purpose of the test is to test whether the device can support network board redundancy. The test steps are as follows: on the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result 1 is obtained; unplug a redundant fan of the DUT, and the expected result 2 is obtained.

[0071] Furthermore, in step S1, the operation and maintenance function test includes packet loss visualization function; ultra-long delay visualization function; full flow analysis function; ioam flow detection function, and the performance test includes MAC table item rule test; ARP / ND table item rule test; FIB table item rule test; full frame performance forwarding; equipment resource utilization test, and the reliability test includes main control board redundancy test; power supply redundancy test; fan redundancy test; switching network board redundancy test.

[0072] Further, in the test 1 of step S2: expected result 1: the optical module information and optical power information can be read correctly; expected result 2: the optical module information and optical power information can be read correctly; expected result 3: snmp is supported to read the optical module and optical power information;

[0073] In test 2: Expected result 1: TC1 and DUT ports can be UP, and rs-fec negotiation is successful; Expected result 2: Statistics on traffic packet loss, bit error rate is less than 1x10e-12;

[0074] Test 3: Expected result 1: The split port forwards traffic at full rate without packet loss;

[0075] Test 4: Expected result 1: Check that the 1588v2 status of the device is normal;

[0076] In test five: Expected result 1: The session detection time meets the minimum detection cycle; Expected result 2: Messages are forwarded between DUT1 and DUT2 through the main link; Expected result 3: The link switching between DUT1 and DUT2 is successful, and the messages are forwarded from the backup link. The traffic switching time is calculated by the number of packet losses on the tester; Expected result 4: The link recovery between DUT1 and DUT2 is successful, and the messages are forwarded from the main link. The traffic recovery time is calculated by the number of packet losses on the tester;

[0077] In test six: Expected result 1: BGP EVPN neighbor status is normal, and vxlan tunnel is up; Expected result 2: BGP EVPN neighbor and vxlan tunnel status are abnormal; Expected result 3: Traffic is forwarded normally without packet loss; Expected result 4: Traffic is blocked;

[0078] Test 7: Expected result 1: BGP EVPN neighbor status is normal, vxlan tunnel is up; Expected result 2: Traffic is forwarded normally without packet loss; Expected result 3: Traffic is blocked;

[0079] Test 8: Expected result 1: BGP EVPN neighbor status is normal, vxlan tunnel is up; Expected result 2: ARP broadcast messages can be captured; Expected result 3: By querying the device ARP broadcast suppression table, the corresponding ip / mac information of VM1 and VM2 can be viewed, and only ARP unicast messages can be captured; Expected result 4: It can be seen on the device that the device has limited the message speed according to the default or configured speed limit value;

[0080] Test 9: Expected result 1: The cross-chassis link bundling status is normal, and the port aggregation status is normal; Expected result 2: Traffic is forwarded normally without packet loss, and load sharing is performed on the link; Expected result 3: Traffic is switched to the remaining links for forwarding, and the packet loss time is recorded; Expected result 4: Traffic is switched back, and the packet loss time is recorded;

[0081] Test 10: Expected result 1: The cross-chassis link bundling status is normal, and the port aggregation status is normal; Expected result 2: Traffic is forwarded normally without packet loss, and load is shared on the link; Expected result 3: Traffic is switched to the remaining links for forwarding, and the packet loss time is recorded; Expected result 4: Traffic is switched back, and the packet loss time is recorded.

[0082] Further, in the test eleven of step S3: Expected result 1: traffic is forwarded normally without packet loss; Expected result 2: packet loss visual flow table information can be viewed on the DUT, and a message containing packet loss information is received on the Server;

[0083] Test 12: Expected result 1: Traffic is forwarded normally without packet loss; Expected result 2: The latency visual flow table information can be viewed on the DUT, and the server receives the message containing latency information;

[0084] Test 13: Expected result 1: Traffic is forwarded normally without packet loss; Expected result 2: Statistical information can be viewed on the DUT, and the server receives messages containing statistical information. The flow information viewed on the device and the server is consistent with the flow information sent by the tester, and the number of messages received is consistent with TC2, which does not affect the forwarding traffic; Expected result 3: Statistical information can be viewed on the DUT, and the server receives messages containing statistical information. The inner messages are all messages that match the ACL. The flow information viewed on the device and the server is consistent with the flow information sent by the tester, which does not affect the forwarding traffic;

[0085] Test 14: Expected result 1: Traffic is forwarded normally without packet loss; Expected result 2: The server receives the ioam packet containing queue, latency and other information.

[0086] Further, in the test fifteen of step S4: Expected result 1: DUT can correctly learn the mac address table entry of the declared specification and record the specification value; Expected result 2: only TC1 can receive traffic, and TC3 cannot receive traffic;

[0087] Test 16: Expected result 1: DUT can correctly learn the declared number of ARP entries; Expected result 2: All learned ARP entries can be forwarded correctly; Expected result 3: DUT can correctly learn the declared number of ND entries; Expected result 4: All learned ND entries can be forwarded correctly;

[0088] Test 17: Expected result 1: DUT can correctly learn the published routes; Expected result 2: All traffic can be forwarded normally without packet loss;

[0089] Test 18: Expected result 1: Record the full frame performance data of the device: typical byte throughput.

[0090] Furthermore, in the test nineteen of step S4: expected result 1: the traffic can be forwarded normally; expected result 2: the CPU, memory and temperature of the device are recorded.

[0091] Furthermore, in the test twenty of step S5: expected result 1: the traffic can be forwarded normally; expected result 2: the device has an abnormal log alarm, and after the main master is unplugged, the device switches to the backup master and runs normally, there is no packet loss, and it runs for a long time for 1 hour, and the device operation status is observed.

[0092] Furthermore, in the test twenty-one of step S5: expected result 1: the traffic can be forwarded normally; expected result 2: the device has abnormal log alarm, the traffic does not lose packets, and it runs for a long time for 1 hour to observe the operation status of the device.

[0093] Furthermore, in the test twenty-two of step S5: expected result 1: the traffic can be forwarded normally; expected result 2: the device has an abnormal log alarm, the traffic does not lose packets, and it runs for a long time for 1 hour to observe the operation status of the device.

[0094] Furthermore, in the test twenty-three of step S5: expected result 1: the traffic can be forwarded normally; expected result 2: the device has abnormal log alarm, the traffic does not lose packets, and it runs for a long time for 1 hour to observe the operation status of the device.

[0095] The present invention provides a data center high-speed switch testing method for cross-domain heterogeneous distributed learning, which has the following beneficial effects:

[0096] 1. The test method for high-speed switches in data centers for cross-domain heterogeneous distributed learning tests the 400G switches serving cross-domain heterogeneous distributed learning data centers from four dimensions: basic functions, operation and maintenance functions, performance, and reliability. The test objectives, test topologies, test steps, and expected results of each aspect are given to measure the comprehensive capabilities of 400G switches. The 400G switch capabilities can be standardized from all aspects and angles, and each dimension is further refined into small items and tested. The specific test process and expected results are given, thereby providing comprehensive and effective indicators for the graded evaluation of 400G switches, thereby providing an important basis for the quality evaluation of 400G switches in data centers for cross-domain heterogeneous distributed learning.

[0097] 2. This data center high-speed switch testing method for cross-domain heterogeneous distributed learning, during the normative testing process of the 400G switch serving the cross-domain heterogeneous distributed learning data center, tests various performances and stipulates the results, so that the performance evaluation of the 400G switch serving the cross-domain heterogeneous distributed learning data center can be truly close to actual needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 A schematic diagram of a test topology of a test of the present invention;

[0099] Figure 2 This is a schematic diagram of the test topology of Test 2 of the present invention;

[0100] Figure 3 This is a schematic diagram of the test topology of test three of the present invention;

[0101] Figure 4 It is a schematic diagram of the test topology of test 4 of the present invention;

[0102] Figure 5 This is a schematic diagram of the test topology of Test 5 of the present invention;

[0103] Figure 6 It is a schematic diagram of the test topology of test 6 of the present invention;

[0104] Figure 7 It is a schematic diagram of the test topology of test seven of the present invention;

[0105] Figure 8 It is a schematic diagram of the test topology of test eight of the present invention;

[0106] Fig. 9 It is a schematic diagram of the test topology of test nine of the present invention;

[0107] Fig.10 A schematic diagram of the test topology of the present invention;

[0108] Fig.11 This is a schematic diagram of the test topology of Test 11 of the present invention;

[0109] Fig.12 A schematic diagram of a test topology for test twelve of the present invention;

[0110] Fig.13 This is a schematic diagram of the test topology of test 13 of the present invention;

[0111] Fig.14 A schematic diagram of a test topology for test 14 of the present invention;

[0112] Fig.15 A schematic diagram of a test topology for test 15 of the present invention;

[0113] Fig.16 A schematic diagram of the test topology of test 16 of the present invention;

[0114] Fig.17 A schematic diagram of the test topology of test 17 of the present invention;

[0115] Fig.18 It is a schematic diagram of the test topology of test 18 of the present invention;

[0116] Fig.19 A schematic diagram of a test topology for test 19 of the present invention;

[0117] Fig. 20 A schematic diagram of a test topology for test twenty of the present invention;

[0118] Fig.21 It is a schematic diagram of the test topology of test twenty-one of the present invention;

[0119] Fig. 22 It is a schematic diagram of the test topology of test twenty-two of the present invention;

[0120] Fig.23 This is a schematic diagram of the test topology of test twenty-three of the present invention;

[0121] Fig.24 It is a schematic diagram of the overall steps of the present invention. DETAILED DESCRIPTION

[0122] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0123] like Figure 1-Figure 24 As shown, the present invention provides a technical solution: a data center high-speed switch testing method for cross-domain heterogeneous distributed learning, and the data center high-speed switch testing method for cross-domain heterogeneous distributed learning includes the following steps:

[0124] S1. Test items in each dimension:

[0125] Dimensions include basic function testing, operation and maintenance function testing, performance testing, and reliability testing;

[0126] Among them, the basic function test includes optical module information reading; Fec test; port splitting; 1588 clock synchronization function; BFD fast switching function; BGP EVPN Vxlan forwarding; IPv6 over IPv4 Vxlan forwarding; ARP broadcast suppression function; cross-chassis link bundling function test; cross-chassis link bundling fault scenario test;

[0127] The operation and maintenance function test includes packet loss visualization function; ultra-long delay visualization function; full flow analysis function; ioam flow detection function; performance test includes MAC table rule test; ARP / ND table rule test; FIB table rule test; full frame performance forwarding; equipment resource utilization test; reliability test includes main control board redundancy test; power supply redundancy test; fan redundancy test; switching network board redundancy test;

[0128] S2. Basic function test:

[0129] Test 1: Optical module information reading test, test topology as follows Figure 1 As shown in the figure, the purpose of the test is to verify that the device supports the optical port optical power reading function. The test steps are as follows: For the 400GE interface, the DUTs are connected through 400G optical modules. When the interface is UP, the optical module information is read, and the expected result is 1.

[0130] Shutdown the interface and read the optical module information again, with the expected result 2.

[0131] Use snmp to read the optical module information, and the expected result is 3;

[0132] Expected result 1: The optical module information and optical power information can be read correctly; Expected result 2: The optical module information and optical power information can be read correctly; Expected result 3: SNMP is supported to read the optical module and optical power information;

[0133] Test 2: Fec test, test topology as follows Figure 2 As shown, the test purpose is to test the DUT's support for the RS-FEC function. The test steps are as follows: TC1 and DUT-1 are interconnected through a 5m DAC, and RS-FEC is turned on on the TC1 side; there is an expected result 1;

[0134] RS-fec is enabled on all links, and TC1 and TC2 exchange 128-byte 400G line-speed traffic for one hour, with the expected result 2.

[0135] Expected result 1: TC1 and DUT ports can be UP, and rs-fec negotiation is successful; Expected result 2: Statistics on traffic packet loss, the bit error rate is less than 1x10e-12;

[0136] Test 3: Port split test, test topology as follows Figure 3 As shown in the figure, the purpose of the test is to verify the port splitting capability. The test steps are as follows: DUT is for a 400GE interface. The test port can be split into 4*100GE ports. The splitting does not require a board or device reset to take effect. After the splitting, the 100GE interface is connected to the tester. The traffic verification through the tester has the expected result 1.

[0137] Expected result 1: The split port forwards traffic at full rate without packet loss;

[0138] Test 4: 1588 clock synchronization function test, test topology as follows Figure 4 As shown in the figure, the purpose of the test is to test whether the device can support the 1588 clock synchronization function. The test steps are as follows: globally enable PTP on DUT1, configure the device type to BC, and configure the local priority 1 to 10; globally enable PTP on DUT2, configure the device type to BC, enable 1588 synchronization, and configure the clock source priority to 2;

[0139] Enable PTP on the interconnection interface between DUT-1 and DUT-2, and the expected result 1 can be obtained;

[0140] Expected result 1: Check that the 1588v2 status of the device is normal;

[0141] Test 5: BFD fast switching function test, test topology as follows Figure 5 As shown in the figure, the purpose of the test is to test whether the device can support the BFD fast switching function. The test steps are as follows: configure the interconnection interfaces of DUT1 and DUT2 as Layer 3 interfaces respectively, and configure IP addresses respectively;

[0142] Configure static routes and priorities on DUT_1 and DUT_2. The interconnection link of Port_2 has a high priority and is the primary link. The interconnection link of Port_3 has a low priority and is the backup link.

[0143] Enable BFD for DUT1 and DUT2, associate with static routing, configure the BFD check time to 3ms, and the number of detections to 3. Check the BFD session, and you will see the expected result 1.

[0144] Send test traffic from port Tport_1 to port Tport_2, with expected result 2;

[0145] Disconnect the main link of Port_2 between DUT_1 and DUT_1, and the expected result is 3;

[0146] Restore the main link of Port_2 between DUT_1 and DUT_2, and get the expected result 4;

[0147] Repeat steps 3 and 4 for Expected Result 3 and Expected Result 4 for 3 times, and record the packet loss time.

[0148] Expected result 1: The session detection time meets the minimum detection cycle; Expected result 2: Messages are forwarded between DUT1 and DUT2 through the main link; Expected result 3: The link switching between DUT1 and DUT2 is successful, and the messages are forwarded from the backup link. The traffic switching time is calculated by the number of packet losses on the tester; Expected result 4: The link recovery between DUT1 and DUT2 is successful, and the messages are forwarded from the main link. The traffic recovery time is calculated by the number of packet losses on the tester;

[0149] Test 6: BGP EVPN Vxlan forwarding test, test topology as follows Figure 6 As shown in the figure, the purpose of the test is to test whether the device can support the BGP EVPN Vxlan function. The test steps are as follows: configure the interface IP address on DUT_1 and DUT_2, configure the IGP routing protocol, publish the route of the interface IP address and the route of the loopback address, and ensure the three-layer interconnection of the network; configure the EVPN BGP peer and vxlan tunnel on the DUT_1 and DUT_2 devices, and get the expected result 1;

[0150] Disconnect the interconnection link between DUT1 and DUT2, and the expected result 2 is obtained;

[0151] Restore the disconnected link, configure tenant 1 gateway and other configurations on DUT_1 and DUT_2 devices, and send Layer 2 and Layer 3 IPv4 traffic on the tester Tport_1 and Tport_2 ports, with the expected result 3;

[0152] Disconnect the main link of Port_2 between DUT_1 and DUT_1, and the expected result is 3;

[0153] Configure the tenant 2 gateway and other configurations on the DUT_1 and DUT_2 devices. The tester Tport_1 and Tport_2 ports send cross-tenant IPv4 traffic, and the expected result 4 is obtained;

[0154] Expected result 1: BGP EVPN neighbor status is normal, and the vxlan tunnel is up; Expected result 2: BGP EVPN neighbor and vxlan tunnel status are abnormal; Expected result 3: Traffic is forwarded normally without packet loss; Expected result 4: Traffic is blocked;

[0155] Test 7: IPv6 over IPv4 Vxlan forwarding test, test topology as follows Figure 7 As shown in the figure, the purpose of the test is to test whether the device can support IPv6 over IPv4 Vxlan function. The test steps are as follows: configure the interface IP address on DUT_1 and DUT_2, configure the IGP routing protocol, publish the route of the interface IP address and the route of the loopback address, and ensure the three-layer interconnection of the network; configure the EVPN BGP peer and vxlan tunnel on DUT_1 and DUT_2 devices, and get the expected result 1;

[0156] Disconnect the interconnection link between DUT1 and DUT2, and get the expected result 1; configure IPv6 over IPv4 configurations such as tenant 1 gateway on DUT_1 and DUT_2 devices, and send Layer 3 IPv6 traffic on the tester Tport_1 and Tport_2 ports, and get the expected result 2;

[0157] Configure IPv6over IPv4 configurations such as the tenant 2 gateway on DUT_1 and DUT_2 devices, and the tester Tport_1 and Tport_2 ports send cross-tenant IPv4 traffic, with expected result 3;

[0158] Expected result 1: BGP EVPN neighbor status is normal, and the vxlan tunnel is up; Expected result 2: Traffic is forwarded normally without packet loss; Expected result 3: Traffic is blocked;

[0159] Test 8: ARP broadcast suppression function test, test topology such as Figure 8As shown, the purpose of the test is to test whether the device can support the ARP broadcast suppression function. The test steps are as follows: configure the interface IP address on DUT_1 and DUT_2, configure the IGP routing protocol, publish the route of the interface IP address and the route of the loopback address, and ensure the three-layer interconnection of the network; configure EVPN BGP peers and vxlan tunnels on DUT_1 and DUT_2 devices, and get the expected result 1; the tester Tport_1 and Tport_2 ports create virtual machines VM1 and VM2 in the same network segment, ping the gateway, and capture packets on VM2, and get the expected result 2; enable the ARP broadcast suppression function on DUT_3 and DUT_4, clear the arp table entries of VM1 and VM2, VM1 pings VM2, and captures packets on VM2, and get the expected result 3; the tester Tport_1 sends a line-speed arp request message to the Port_1 interface, and gets the expected result 4;

[0160] Expected result 1: BGP EVPN neighbor status is normal, and vxlan tunnel is up; Expected result 2: ARP broadcast messages can be captured; Expected result 3: By querying the device ARP broadcast suppression table, the corresponding ip / mac information of VM1 and VM2 can be viewed, and only ARP unicast messages can be captured; Expected result 4: It can be seen on the device that the device has limited the message speed according to the default or configured speed limit value;

[0161] Test 9: Cross-chassis link bundling function test, test topology as follows Fig. 9 As shown, the purpose of the test is to test whether the device can support the cross-chassis link bundling function. The test steps are as follows: configure the cross-chassis link bundling function on DUT1 and DUT2, and configure the L2 aggregation in lacp mode on DUT3 and DUT4, and the expected result 1 is obtained; configure EVPNBGP peers and vxlan tunnels on DUT_1 and DUT_2 devices, and the expected result 1 is obtained;

[0162] DUT1 and DUT2 devices are configured with two gateway IP addresses, and the interconnected ports and bundled ports are added to the same VLAN. The tester TC1 and TC2 ports send layer 3 traffic with a changed source IP, and the expected result 2 is obtained.

[0163] Disconnect the interconnection link between DUT1 and DUT3, and the expected result 3 is obtained;

[0164] Restore the disconnected link, with expected result 4;

[0165] Expected result 1: The cross-chassis link bundling status is normal, and the port aggregation status is normal; Expected result 2: Traffic is forwarded normally without packet loss, and load sharing is performed on the link; Expected result 3: Traffic is switched to the remaining links for forwarding, and the packet loss time is recorded; Expected result 4: Traffic is switched back, and the packet loss time is recorded;

[0166] Test 10: Cross-chassis link bundling failure scenario test, test topology as follows Fig.10 As shown, the purpose of the test is to test whether the device can support the switching and recovery functions of the cross-chassis link bundling fault scenario. The test steps are as follows: configure the cross-chassis link bundling function on DUT1 and DUT2, and configure the L2 aggregation in the lacp mode on DUT3 and DUT4, and the expected result 1 is obtained; configure EVPN BGP peers and vxlan tunnels on DUT_1 and DUT_2 devices, and the expected result 1 is obtained;

[0167] DUT1 and DUT2 devices are configured with two gateway IP addresses, and the interconnected ports and bundled ports are added to the same VLAN respectively; the tester TC1 and TC2 ports send layer 3 traffic with a changed source IP, and the expected result 2 is obtained; the interconnection link between DUT1 and DUT3 is disconnected, and the expected result 3 is obtained;

[0168] Restore the disconnected link, with expected result 4;

[0169] Expected result 1: The cross-chassis link bundling status is normal, and the port aggregation status is normal;

[0170] Expected result 2: Traffic is forwarded normally without packet loss, and load is shared on the link; Expected result 3: Traffic is switched to the remaining links for forwarding, and the packet loss time is recorded; Expected result 4: Traffic is switched back, and the packet loss time is recorded;

[0171] S3, operation and maintenance function test:

[0172] Test 11: Packet loss visualization function test, test topology as follows Fig.11 As shown, the purpose of the test is to test whether the device can support the packet loss visualization function. The test steps are as follows: configure IPv4 addresses on the interfaces connecting the DUT to the test instrument, and send traffic between TC1 and TC2, with the expected result 1;

[0173] The DUT enables the Buffer packet loss visualization function and configures the Server as a collector. TC1 and TC2 send line-speed traffic to TC3 to create a port congestion and packet loss scenario, which produces the expected result 2.

[0174] Expected result 1: Traffic is forwarded normally without packet loss; Expected result 2: The packet loss visual flow table information can be viewed on the DUT, and the server receives the message containing the packet loss information;

[0175] Test 12: Ultra-long latency visualization function test, test topology such as Fig.12 As shown in the figure, the purpose of the test is to test whether the device can support the ultra-long delay visualization function. The test steps are as follows: configure IPv4 addresses on the interfaces connecting the DUT to the test instrument, and send traffic between TC1 and TC2, with the expected result 1;

[0176] Enable the latency visualization function on the DUT, configure the latency threshold, and configure the server as a collector. TC1 and TC2 send line-speed traffic to TC3 to create a scenario where port congestion increases latency, and the expected result is 2.

[0177] Expected result 1: Traffic is forwarded normally without packet loss; Expected result 2: The latency visual flow table information can be viewed on the DUT, and the server receives the message containing latency information;

[0178] Test 13: Full flow analysis function test, test topology such as Fig.13 As shown, the purpose of the test is to test whether the device can support the full-flow analysis function. The test steps are as follows: configure IPv4 addresses on the interfaces connecting the DUT to the test instrument, and send TCP / UDP traffic between TC1 and TC2, with the expected result 1;

[0179] The DUT enables the full-flow analysis service and configures the Server as a collector, with the expected result 2.

[0180] Configure ACL to match the sent TCP / UDP traffic, and get the expected result 3;

[0181] Expected result 1: Traffic is forwarded normally without packet loss; Expected result 2: Statistics can be viewed on the DUT, and the server receives messages containing statistics. The flow information viewed on the device and the server is consistent with the flow information sent by the tester, and the number of messages received is consistent with TC2, which does not affect the forwarding traffic; Expected result 3: Statistics can be viewed on the DUT, and the server receives messages containing statistics. The inner messages are all messages that match the ACL. The flow information viewed on the device and the server is consistent with the flow information sent by the tester, which does not affect the forwarding traffic;

[0182] Test 14: ioam flow detection function test, test topology as follows Fig.14 As shown, the purpose of the test is to test whether the device can support the ioam flow detection function. The test steps are as follows: 1) DUT1, DUT2, and DUT3 are configured as three-layer network interconnection, and the gateways of TC1 and TC2 are configured on DUT1 and DUT2. TC1 sends TCP traffic to TC2, and the expected result 1 is obtained;

[0183] Enable the ioam function on DUT1, DUT2, and DUT3. Configure ACL matching traffic on DUT1 for sampling. Configure an IPv4 address on the interface connected to the server on DUT2, which is reachable to the server network. Send ioam packets to the analyzer. TC1 sends TCP traffic to TC2, with expected result 2.

[0184] Expected result 1: Traffic is forwarded normally without packet loss; Expected result 2: The server receives the ioam packet containing queue, delay and other information;

[0185] S4. Performance test:

[0186] Test 15: MAC table rule test, test topology as follows Fig.15 As shown, the purpose of the test is to test the maximum number of MAC address table entries that the device can support. The test steps are as follows: Test instrument TC1 inputs smac changing traffic, the number of changes is the number of mac address table entries declared by the DUT, and the expected result is 1;

[0187] TC2 enters the traffic whose dmac is the learned mac address table entry, and the expected result is 2;

[0188] Expected result 1: DUT can correctly learn the mac address table entries of the declared specifications and record the specification values; Expected result 2: Only TC1 can receive traffic, and TC3 cannot receive traffic;

[0189] Test 16: ARP table specification test, test topology such as Fig.16 As shown, the purpose of the test is to test the maximum number of ARP entries that the device can support. The test steps are as follows: 1) Configure IPv4 and IPv6 addresses on the interface connecting the DUT to the test instrument respectively; the test instrument TC1 inputs the ARP reply message of the declared ARP specification at line speed, and the expected result is 1;

[0190] TC2 injects verification traffic with the destination address being the learned ARP table entry, and the expected result is 2.

[0191] The test instrument TC1 inputs the RA message of the claimed nd specification at line speed, and the expected result is 3;

[0192] TC2 injects verification traffic with the destination address being the learned ND entry, and the expected result is 4.

[0193] Expected result 1: DUT can correctly learn the declared number of ARP entries; Expected result 2: All learned ARP entries can be forwarded correctly; Expected result 3: DUT can correctly learn the declared number of ND entries; Expected result 4: All learned ND entries can be forwarded correctly;

[0194] Test 17: FIB table rule test, test topology such as Fig.17 As shown in the figure, the purpose of the test is to test the maximum number of FIB entries that the device can support. The test steps are as follows: DUT and TC1 establish an EBGP neighbor relationship; TC1 enters the BGP route of the FIB specification announced by the device, and the expected result is 1;

[0195] TC2 sends traffic with the destination address being the published route, with expected result 2.

[0196] Establish an EBGPv6 neighbor relationship between DUT and TC1, and repeat the steps of expected result 1 and expected result 2 using IPv6 routing;

[0197] Expected result 1: DUT can correctly learn the published routes; Expected result 2: All traffic can be forwarded normally without packet loss;

[0198] Test 18: Full frame performance forwarding test, test topology as follows Fig.18 As shown in the figure, the purpose of the test is to test the full frame performance that the device can support. The test steps are as follows: the DUT is fully plugged with single boards, and all interfaces of the same type are connected as shown in the figure above. The first interface and the last interface are connected to the test instrument. The two interfaces are configured in the same VRF, and each VRF is configured with static routing to achieve forwarding. The RFC2544 test suite of the test instrument is used to test the IPv4 forwarding performance, with the expected result 1.

[0199] The dual-stack forwarding performance was tested using the RFC2544 test suite of the test instrument, with IPv4 and IPv6 traffic accounting for 50% each, and the expected result was 1; the "double snake" traversal test was performed on both the uplink and downlink ports at the same time;

[0200] Expected result 1: Record the full frame performance data of the device: typical byte throughput;

[0201] Test 19: Device resource utilization test, test topology as follows Fig.19 As shown in the figure, the purpose of the test is to test the full frame performance that the device can support. The test steps are as follows: On the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result is 1;

[0202] After the flow rate stabilizes, observe the equipment operation status and get the expected result 2;

[0203] Expected result 1: Traffic can be forwarded normally; Expected result 2: The CPU, memory, and temperature of the device are recorded;

[0204] S5. Reliability test:

[0205] Test 20: Main control board redundancy test, test topology as follows Fig. 20 As shown, the purpose of the test is to test whether the device can support main control board redundancy. The test steps are as follows: On the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result 1 is obtained; unplug the main control board of the DUT, and the expected result 2 is obtained;

[0206] Expected result 1: Traffic can be forwarded normally; Expected result 2: The device has abnormal log alarms. After the main control is unplugged, the device switches to the backup control and runs normally. There is no packet loss. It runs for a long time for 1 hour. Observe the operation status of the device;

[0207] Test 21: Power redundancy test, test topology such as Fig.21 As shown in the figure, the purpose of the test is to test whether the device can support power redundancy. The test steps are as follows: On the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result 1 is obtained; unplug a redundant power supply, and the expected result 2 is obtained;

[0208] Expected result 1: Traffic can be forwarded normally; Expected result 2: The device has abnormal log alarms, traffic does not lose packets, and runs for 1 hour to observe the device operation status;

[0209] Test 22: Fan redundancy test, test topology as follows Fig. 22 As shown, the purpose of the test is to test whether the device can support fan redundancy. The test steps are as follows: On the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result 1 is obtained; unplug a redundant fan of the DUT, and the expected result 2 is obtained;

[0210] Expected result 1: Traffic can be forwarded normally; Expected result 2: The device has abnormal log alarms, traffic does not lose packets, and runs for 1 hour to observe the device operation status;

[0211] Test 23: Switching network board redundancy test, test topology as follows Fig.23 As shown, the purpose of the test is to test whether the device can support network board redundancy. The test steps are as follows: On the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result 1 is obtained; unplug a redundant fan of the DUT, and the expected result 2 is obtained;

[0212] Expected result 1: Traffic can be forwarded normally; Expected result 2: The device has abnormal log alarms, traffic does not lose packets, and runs for 1 hour to observe the device operation status;

[0213] The above tests 1 to 23 are all carried out under the premise that the topology devices are working normally.

[0214] Based on the above description, the data center high-speed switch testing method for cross-domain heterogeneous distributed learning tests the 400G switch serving the cross-domain heterogeneous distributed learning data center from four dimensions: basic functions, operation and maintenance functions, performance, and reliability. The test objectives, test topologies, test steps, and expected results of each aspect are given to measure the comprehensive capabilities of the 400G switch. The 400G switch capabilities can be standardized from all aspects and angles, and each dimension is further refined into small items and tested. The specific test process and expected results are given, thereby providing comprehensive and effective indicators for the grade grading evaluation of the 400G switch, thereby providing an important basis for the quality evaluation of the 400G switch in the data center for cross-domain heterogeneous distributed learning. In the process of normative testing of the 400G switch serving the cross-domain heterogeneous distributed learning data center, each performance is tested and the results are specified, so that the performance evaluation of the 400G switch serving the cross-domain heterogeneous distributed learning data center can be truly close to actual needs.

[0215] The embodiments of the present invention are given for the purpose of illustration and description, and are not intended to be exhaustive or to limit the invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

Claims

1. A data center high-speed switch testing method for cross-domain heterogeneous distributed learning, characterized by: The data center high-speed switch testing method for cross-domain heterogeneous distributed learning comprises the following steps: S1. Test items in each dimension: Dimensions include basic function testing, operation and maintenance function testing, performance testing, and reliability testing; Among them, the basic function test includes optical module information reading; Fec test; port splitting; 1588 clock synchronization function; BFD fast switching function; BGP EVPN Vxlan forwarding; IPv6 over IPv4 Vxlan forwarding; ARP broadcast suppression function; cross-chassis link bundling function test; cross-chassis link bundling fault scenario test; S2. Basic function test: Test 1: Optical module information reading test. Its purpose is to verify that the device supports the optical port optical power reading function. The test steps are as follows: For the 400GE interface, the DUTs are connected through 400G optical modules. When the interface is UP, read the optical module information and get the expected result 1. Shutdown the interface and read the optical module information again, with the expected result 2. Use snmp to read the optical module information, and the expected result is 3; Test 2: Fec test, the purpose of which is to test DUT's support for RS-FEC function. The test steps are as follows: TC1 and DUT-1 are interconnected through 5m DAC, and RS-FEC is turned on on TC1 side; there is expected result 1; RS-fec is enabled on all links, and TC1 and TC2 exchange 128-byte 400G line-speed traffic for 1 hour, with expected result 2. Test 3: Port split test, the purpose of which is to verify the port splitting capability. The test steps are as follows: DUT is for a 400GE interface, and the test port can be split into 4*100GE ports. The split does not require a board or device reset to take effect. After the split, the tester is connected to 100GE, and the tester verifies the traffic with the expected result 1; Test 4: 1588 clock synchronization function test, the purpose of the test is to test whether the device can support the 1588 clock synchronization function. The test steps are as follows: globally enable PTP on DUT1, configure the device type to BC, and configure the local priority 1 to 10; globally enable PTP on DUT2, configure the device type to BC, enable 1588 synchronization, and configure the clock source priority to 2; Enable PTP on the interconnection interface between DUT-1 and DUT-2, and the expected result 1 can be obtained; Test 5: BFD fast switching function test, the purpose of the test is to test whether the device can support the BFD fast switching function. The test steps are as follows: configure the interconnection interface of DUT1 and DUT2 as a Layer 3 interface, and configure IP addresses respectively; Configure static routes and priorities on DUT_1 and DUT_2. The interconnection link of Port_2 with a high priority is the primary link, and the interconnection link of Port_3 with a low priority is the backup link. Enable BFD for DUT1 and DUT2, associate with static routing, configure the BFD check time to 3ms, and the number of detections to 3. Check the BFD session, and you will see the expected result 1. Send test traffic from port Tport_1 to port Tport_2, with expected result 2; Disconnect the main link of Port_2 between DUT_1 and DUT_1, and the expected result is 3; Restore the main link of Port_2 between DUT_1 and DUT_2, and get the expected result 4; Repeat steps 3 and 4 for Expected Result 3 and Expected Result 4 for 3 times, and record the packet loss time. Test 6: BGP EVPN Vxlan forwarding test, the purpose of the test is to test whether the device can support BGP EVPNVxlan function. The test steps are as follows: configure the interface IP address on DUT_1 and DUT_2, configure the IGP routing protocol, publish the route of the interface IP address and the route of the loopback address, and ensure the three-layer interconnection of the network; configure EVPNBGP peers and vxlan tunnels on DUT_1 and DUT_2 devices, and get the expected result 1; Disconnect the interconnection link between DUT1 and DUT2, and the expected result 2 is obtained; Restore the disconnected link, configure tenant 1 gateway and other configurations on DUT_1 and DUT_2 devices, and send Layer 2 and Layer 3 IPv4 traffic on the tester Tport_1 and Tport_2 ports, with the expected result 3; Disconnect the main link of Port_2 between DUT_1 and DUT_1, and the expected result is 3; Configure the tenant 2 gateway and other configurations on the DUT_1 and DUT_2 devices. The tester Tport_1 and Tport_2 ports send cross-tenant IPv4 traffic, and the expected result 4 is obtained; Test 7: IPv6 over IPv4 Vxlan forwarding test. The purpose of the test is to test whether the device can support the IPv6 over IPv4 Vxlan function. The test steps are as follows: configure the interface IP address on DUT_1 and DUT_2, configure the IGP routing protocol, publish the route of the interface IP address and the route of the loopback address, and ensure the three-layer interconnection of the network; configure the EVPN BGP peer and vxlan tunnel on the DUT_1 and DUT_2 devices, and get the expected result 1; Disconnect the interconnection link between DUT1 and DUT2, and get the expected result 1; configure IPv6 over IPv4 configurations such as tenant 1 gateway on DUT_1 and DUT_2 devices, and send Layer 3 IPv6 traffic on the tester Tport_1 and Tport_2 ports, and get the expected result 2; Configure IPv6over IPv4 configurations such as the tenant 2 gateway on DUT_1 and DUT_2 devices, and the tester Tport_1 and Tport_2 ports send cross-tenant IPv4 traffic, with expected result 3; Test 8: ARP broadcast suppression function test, the purpose of the test is to test whether the device can support the ARP broadcast suppression function. The test steps are as follows: configure the interface IP address on DUT_1 and DUT_2, configure the IGP routing protocol, publish the route of the interface IP address and the route of the loopback address, and ensure the three-layer interconnection of the network; configure EVPN BGP peers and vxlan tunnels on DUT_1 and DUT_2 devices, and get the expected result 1; the tester Tport_1 and Tport_2 ports create virtual machines VM1 and VM2 in the same network segment, ping the gateway, and capture packets on VM2, and get the expected result 2; enable the ARP broadcast suppression function on DUT_3 and DUT_4, clear the arp table entries of VM1 and VM2, VM1 pings VM2, and captures packets on VM2, and get the expected result 3; the tester Tport_1 sends a line-speed arp request message to the Port_1 interface, and gets the expected result 4; Test 9: Cross-chassis link bundling function test, the purpose of the test is to test whether the device can support the cross-chassis link bundling function. The test steps are as follows: configure the cross-chassis link bundling function on DUT1 and DUT2, and configure the L2 aggregation in lacp mode on DUT3 and DUT4, and the expected result 1 is obtained; configure EVPN BGP peers and vxlan tunnels on DUT_1 and DUT_2 devices, and the expected result 1 is obtained; DUT1 and DUT2 devices are configured with two gateway IP addresses, and the interconnected ports and bundled ports are added to the same VLAN. The tester TC1 and TC2 ports send layer 3 traffic with a changed source IP, and the expected result 2 is obtained. Disconnect the interconnection link between DUT1 and DUT3, and the expected result is 3; Restore the disconnected link with expected result 4; Test 10: Cross-chassis link bundling failure scenario test. The purpose of the test is to test whether the device can support the switching and recovery functions of the cross-chassis link bundling failure scenario. The test steps are as follows: configure the cross-chassis link bundling function on DUT1 and DUT2, and configure the L2 aggregation in the lacp mode on DUT3 and DUT4, and the expected result 1 is obtained; configure EVPNBGP peers and vxlan tunnels on DUT_1 and DUT_2 devices, and the expected result 1 is obtained; DUT1 and DUT2 devices are configured with two gateway IP addresses, and the interconnected ports and bundled ports are added to the same VLAN respectively; the tester TC1 and TC2 ports send layer 3 traffic with a changed source IP, and the expected result 2 is obtained; the interconnection link between DUT1 and DUT3 is disconnected, and the expected result 3 is obtained; Restore the disconnected link with expected result 4; S3, operation and maintenance function test: Test 11: Packet loss visualization function test. The purpose of the test is to test whether the device can support the packet loss visualization function. The test steps are as follows: configure IPv4 addresses on the interfaces connecting the DUT to the test instrument, and send traffic between TC1 and TC2, with the expected result 1; The DUT enables the Buffer packet loss visualization function and configures the Server as a collector. TC1 and TC2 send line-speed traffic to TC3 to create a port congestion and packet loss scenario, which produces the expected result 2. Test 12: Ultra-long delay visualization function test. The purpose of the test is to test whether the device can support the ultra-long delay visualization function. The test steps are as follows: configure IPv4 addresses on the interfaces connecting the DUT to the test instrument, and send traffic between TC1 and TC2, with the expected result 1; Enable the latency visualization function on the DUT, configure the latency threshold, and configure the server as a collector. TC1 and TC2 send line-speed traffic to TC3 to create a scenario where port congestion increases latency, and the expected result is 2. Test 13: Full-flow analysis function test. The purpose of the test is to test whether the device can support the full-flow analysis function. The test steps are as follows: configure IPv4 addresses on the interfaces of the DUT connected to the test instrument, and send TCP / UDP traffic between TC1 and TC2, with the expected result 1; The DUT enables the full-flow analysis service and configures the Server as a collector, with the expected result 2. Configure ACL to match the sent TCP / UDP traffic, and get the expected result 3; Test 14: ioam flow detection function test, the purpose of the test is to test whether the device can support ioam flow detection function, the test steps are as follows: 1) DUT1, DUT2, DUT3 are configured as three-layer network interconnection, TC1 and TC2 gateways are configured on DUT1 and DUT2, TC1 sends TCP traffic to TC2, and the expected result 1 is obtained; Enable the ioam function on DUT1, DUT2, and DUT3. Configure ACL matching traffic on DUT1 for sampling. Configure an IPv4 address on the interface connected to the server on DUT2, which is reachable to the server network. Send ioam packets to the analyzer. TC1 sends TCP traffic to TC2, with expected result 2. S4. Performance test: Test 15: MAC table entry rule test, the purpose of which is to test the maximum number of MAC address table entries that the device can support. The test steps are as follows: Test instrument TC1 inputs smac changing traffic, the number of changes is the number of mac address table entries declared by the DUT, and the expected result is 1; TC2 enters the traffic whose dmac is the learned mac address table entry, and the expected result is 2; Test 16: ARP table specification test, the purpose of the test is to test the maximum number of ARP table entries that the device can support. The test steps are as follows: 1) Configure IPv4 and IPv6 addresses on the interface connecting the DUT to the test instrument respectively; the test instrument TC1 enters the ARP reply message of the declared ARP specification at line speed, and the expected result is 1; TC2 injects verification traffic with the destination address being the learned ARP table entry, and obtains the expected result 2. The test instrument TC1 inputs the RA message of the claimed nd specification at line speed, and the expected result is 3; TC2 injects verification traffic with the destination address being the learned ND entry, and the expected result is 4. Test 17: FIB table rule test, the purpose of the test is to test the maximum number of FIB table items that the device can support. The test steps are as follows: DUT and TC1 establish an EBGP neighbor relationship; TC1 enters the BGP route of the FIB specification announced by the device, and the expected result is 1; TC2 sends traffic with the destination address being the published route, with expected result 2. Establish an EBGPv6 neighbor relationship between DUT and TC1, and repeat the steps of expected result 1 and expected result 2 using IPv6 routing; Test 18: Full-frame performance forwarding test, the purpose of the test is to test the full-frame performance that the device can support. The test steps are as follows: The DUT is fully plugged with single boards, and all interfaces of the same type are connected as shown in the figure above. The first interface and the last interface are connected to the test instrument. The two interfaces are configured in the same VRF, and each VRF is configured with static routing to achieve forwarding; the RFC2544 test suite of the test instrument is used to test the IPv4 forwarding performance, with the expected result 1; The dual-stack forwarding performance was tested using the RFC2544 test suite of the test instrument, with IPv4 and IPv6 traffic accounting for 50% each, and the expected result was 1; the "double snake" traversal test was performed on both the uplink and downlink ports at the same time; Test 19: Equipment resource utilization test, the purpose of which is to test the full frame performance that the equipment can support. The test steps are as follows: On the basis that the equipment can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result is 1; After the flow rate stabilizes, observe the equipment operation status and get the expected result 2; S5. Reliability test: Test 20: Main control board redundancy test, the purpose of the test is to test whether the device can support main control board redundancy. The test steps are as follows: on the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result 1 is obtained; unplug the main control board of the DUT, and the expected result 2 is obtained; Test 21: Power redundancy test, the purpose of the test is to test whether the device can support power redundancy. The test steps are as follows: On the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result 1 is obtained; unplug a redundant power supply, and the expected result 2 is obtained; Test 22: Fan redundancy test, the purpose of which is to test whether the device can support fan redundancy. The test steps are as follows: On the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result 1 is obtained; unplug a redundant fan of the DUT, and the expected result 2 is obtained; Test 23: Switching network board redundancy test, the purpose of the test is to test whether the device can support network board redundancy. The test steps are as follows: on the basis that the device can forward traffic normally, the test instrument sends bidirectional traffic, and the expected result 1 is obtained; unplug a redundant fan of the DUT, and the expected result 2 is obtained.

2. The data center high-speed switch testing method for cross-domain heterogeneous distributed learning according to claim 1 is characterized in that: In step S1, the operation and maintenance function test includes packet loss visualization function; ultra-long delay visualization function; full flow analysis function; ioam flow detection function, the performance test includes MAC table item rule test; ARP / ND table item rule test; FIB table item rule test; full frame performance forwarding; equipment resource utilization test, and the reliability test includes main control board redundancy test; power supply redundancy test; fan redundancy test; switching network board redundancy test.

3. The data center high-speed switch testing method for cross-domain heterogeneous distributed learning according to claim 1 is characterized in that: In the test 1 of step S2: expected result 1: the optical module information and optical power information can be read correctly; expected result 2: the optical module information and optical power information can be read correctly; expected result 3: snmp is supported to read the optical module and optical power information; In test 2: Expected result 1: TC1 and DUT ports can be UP, and rs-fec negotiation is successful; Expected result 2: Statistics on traffic packet loss, the bit error rate is less than 1x10e-12; Test 3: Expected result 1: The split port forwards traffic at full rate without packet loss; Test 4: Expected result 1: Check that the 1588v2 status of the device is normal; In test five: Expected result 1: The session detection time meets the minimum detection cycle; Expected result 2: Messages are forwarded between DUT1 and DUT2 through the main link; Expected result 3: The link switching between DUT1 and DUT2 is successful, and the messages are forwarded from the backup link. The traffic switching time is calculated by the number of packet losses on the tester; Expected result 4: The link recovery between DUT1 and DUT2 is successful, and the messages are forwarded from the main link. The traffic recovery time is calculated by the number of packet losses on the tester; In test six: Expected result 1: BGP EVPN neighbor status is normal, vxlan tunnel is up; Expected result 2: BGP EVPN neighbor and vxlan tunnel status are abnormal; Expected result 3: Traffic is forwarded normally without packet loss; Expected result 4: Traffic is blocked; Test 7: Expected result 1: BGP EVPN neighbor status is normal, vxlan tunnel is up; Expected result 2: Traffic is forwarded normally without packet loss; Expected result 3: Traffic is blocked; Test 8: Expected result 1: BGP EVPN neighbor status is normal, vxlan tunnel is up; Expected result 2: ARP broadcast messages can be captured; Expected result 3: By querying the device ARP broadcast suppression table, the corresponding ip / mac information of VM1 and VM2 can be viewed, and only ARP unicast messages can be captured; Expected result 4: It can be seen on the device that the device has limited the message speed according to the default or configured speed limit value; Test 9: Expected result 1: The cross-chassis link bundling status is normal, and the port aggregation status is normal; Expected result 2: Traffic is forwarded normally without packet loss, and load sharing is performed on the link; Expected result 3: Traffic is switched to the remaining links for forwarding, and the packet loss time is recorded; Expected result 4: Traffic is switched back, and the packet loss time is recorded; Test 10: Expected result 1: The cross-chassis link bundling status is normal, and the port aggregation status is normal; Expected result 2: Traffic is forwarded normally without packet loss, and load is shared on the link; Expected result 3: Traffic is switched to the remaining links for forwarding, and the packet loss time is recorded; Expected result 4: Traffic is switched back, and the packet loss time is recorded.

4. The data center high-speed switch testing method for cross-domain heterogeneous distributed learning according to claim 1 is characterized in that: In the test 11 of step S3: Expected result 1: traffic is forwarded normally without packet loss; Expected result 2: packet loss visual flow table information can be viewed on the DUT, and the server receives a message containing packet loss information; Test 12: Expected result 1: Traffic is forwarded normally without packet loss; Expected result 2: The latency visual flow table information can be viewed on the DUT, and the server receives the message containing latency information; Test 13: Expected result 1: Traffic is forwarded normally without packet loss; Expected result 2: Statistical information can be viewed on the DUT, and the server receives messages containing statistical information. The flow information viewed on the device and the server is consistent with the flow information sent by the tester, and the number of messages received is consistent with TC2, which does not affect the forwarding traffic; Expected result 3: Statistical information can be viewed on the DUT, and the server receives messages containing statistical information. The inner messages are all messages that match the ACL. The flow information viewed on the device and the server is consistent with the flow information sent by the tester, which does not affect the forwarding traffic; Test 14: Expected result 1: Traffic is forwarded normally without packet loss; Expected result 2: The server receives the ioam packet containing queue, latency and other information.

5. The data center high-speed switch testing method for cross-domain heterogeneous distributed learning according to claim 1 is characterized in that: In the test fifteen of step S4: Expected result 1: DUT can correctly learn the mac address table entry of the declared specification and record the specification value; Expected result 2: Only TC1 can receive traffic, and TC3 cannot receive traffic; Test 16: Expected result 1: DUT can correctly learn the declared number of ARP entries; Expected result 2: All learned ARP entries can be forwarded correctly; Expected result 3: DUT can correctly learn the declared number of ND entries; Expected result 4: All learned ND entries can be forwarded correctly; Test 17: Expected result 1: DUT can correctly learn the published routes; Expected result 2: All traffic can be forwarded normally without packet loss; Test 18: Expected result 1: Record the full frame performance data of the device: typical byte throughput.

6. The data center high-speed switch testing method for cross-domain heterogeneous distributed learning according to claim 1 is characterized in that: In the test nineteen of step S4: expected result 1: traffic can be forwarded normally; expected result 2: the CPU, memory and temperature of the device are recorded.

7. The data center high-speed switch testing method for cross-domain heterogeneous distributed learning according to claim 1 is characterized in that: In the test twenty of step S5: Expected result 1: the traffic can be forwarded normally; Expected result 2: the device has an abnormal log alarm, and after the main master is unplugged, the device switches to the backup master and runs normally, there is no packet loss, and it runs for a long time for 1 hour, and observes the operation status of the device.

8. The data center high-speed switch testing method for cross-domain heterogeneous distributed learning according to claim 1 is characterized in that: In the test twenty-one of step S5: Expected result 1: the traffic can be forwarded normally; Expected result 2: the device has abnormal log alarm, the traffic does not lose packets, and it runs for a long time for 1 hour to observe the operation status of the device.

9. The data center high-speed switch testing method for cross-domain heterogeneous distributed learning according to claim 1 is characterized in that: In the test twenty-two of step S5: expected result 1: the traffic can be forwarded normally; expected result 2: the device has abnormal log alarm, the traffic does not lose packets, and it runs for a long time for 1 hour to observe the operation status of the device.

10. The data center high-speed switch testing method for cross-domain heterogeneous distributed learning according to claim 1, characterized in that: In the test twenty-three of step S5: expected result 1: the traffic can be forwarded normally; expected result 2: the device has abnormal log alarm, the traffic does not lose packets, and it runs for a long time for 1 hour to observe the operation status of the device.