Help us improve your experience.

Let us know what you think.

Do you have time for a two-minute survey?

 
 

Layer 2 and Layer 3 RFC 2544-Based Benchmarking Test Overview

In ACX Series routers, RFC 2544-based benchmark tests can be run to measure the performance characteristics of the E-Line, E-LAN, and EVPL services.

Note:

  • Prior to Junos OS Evolved Release 22.4R1, ACX7100 routers can be configured only as a Layer 3 reflector (family inet). Starting in Junos OS Evolved Release 22.4R1, ACX7100 routers can also be configured as a Layer 2 reflector (family ccc or ethernet-switching.) Starting in Junos OS Evolved Release 24.2R1, these routers can also be configured as Layer 2 ingress reflectors.

  • Starting in Junos OS Evolved Release 22.4R1, ACX7509 and ACX7024 routers can be configured as Layer 2 or Layer 3 reflectors. Starting in Junos OS Evolved Release 24.2R1, these routers can also be configured as Layer 2 ingress reflectors.

  • Starting in Junos OS Evolved Release 23.4R1, ACX7332 and ACX7348 routers can be configured as Layer 2 or Layer 3 reflectors.

  • Starting in Junos OS Evolved Release 23.4R1, ACX7024, ACX7024X, ACX7100, and ACX7509 routers can be configured as Layer 3 initiators.

  • Starting in Junos OS Evolved 24.2R1, ACX7024, ACX7024X, ACX7100, ACX7332, and ACX7509 routers can be configured as Layer 2 initiators.

  • ACX5448, ACX5048, and ACX5096 routers can be configured only as a Layer 2 reflector (family bridge or ccc). ACX5048 and ACX5096 routers support only E-Line services.

  • You can configure the test on the following underlying services:

    • Between two IPv4 endpoints—In this mode, the generator sends test packets to a user-configured IP destination or UDP port (which is on the reflector).

    • Between two user-to-network interfaces (UNIs) of Ethernet Virtual Connection (EVC), Ethernet Private Line (EPL, also called E-Line), Ethernet Virtual Private Line (EVPL), EVC (EPL, EVPL)—One end is configured as the generator or initiator and the other end acts as the reflector. The generator receives the test packets that are returned from the reflector and computes the test results.

    Note:

    Benchmarking tests are not supported for IPv6-based services.

  • You cannot perform multiple simultaneous RFC 2544-based benchmarking tests on the same pseudowire.

  • Interoperation of the RFC 2544 benchmarking tests with other third-party customer premises equipment (CPE) that provides embedded or dedicated benchmarking test capability is not supported.

  • Fragmented test-frames and one-way measurements of frames are not supported. You must configure one end or the source device to initiate and terminate test frames and the other end or the destination device to reflect the received frames back to the initiator.

  • RFC 2544 generator and reflector are supported with testing bandwidth up to 1 Gbps. ACX5048 and ACX5096 routers supports test bandwidth of up to 40 Gbps.

  • RFC2544 Layer 2 reflection supports these Layer 2 services : L2 (Bridge), L2CKT, L2VPN, EVPN-VPWS, EVPN-FXC, EVPN-MPLS, and VPLS. You can configure Layer 2 reflection only in the egress direction. Starting in Junos OS Evolved Release 24.2R1, the ACX7024, ACX7100, and ACX7509 routers can also be configured as Layer 2 ingress reflectors. Layer 2 reflection occurs at the UNI interface for unicast traffic only. The MAC addresses are always swapped after reflection. You can configure swapping for IP addresses or UDP ports using the ip-swap or udp-tcp-port-swap statements.

  • RFC2544 Layer 3 reflection supports IPv4 or L3VPN traffic. You can configure Layer 3 reflection only in the ingress direction. Only traffic destined for the host is reflected; transit traffic is not affected. The IP addresses and UDP addresses are swapped after reflection.

  • Supported transport mechanisms for ELAN include:

    • Multipoint Q-in-Q over provider bridged networks

    • Provider Backbone Bridge (Mac-in-Mac)

    • VPLS over IP/MPLS

    • Ethernet VPN, EVPN-MPLS, and EVPN-VXLAN

  • Supported transport mechanisms for E-Line include:

    • Ethernet pseudo-wires

    • Q-in-Q

    • Provider Backbone Bridge (Mac-in-Mac)

    • Bridge domains with two logical interfaces; one for service provider and one for customer VLANs.

  • The test session is supported in out-of-service mode for the underlying service. You must not transmit any traffic to the UNI port, configured as a generator or a reflector, that is being tested during the duration of the test. However, other services that are not configured for the testing session are not impacted.

  • Devices embedded with benchmarking test capabilities (generators and reflectors) interoperate with other Juniper Networks devices that support the RFC 2544-based generator or reflector functionality.

  • RFC 2544 generator traffic undergoes the same traffic classifier and policer or shaper processing as the ingress customer traffic from the UNI port.

  • RFC 2544 generator produces a report with clear details of pass or fail for each critical testing metric, based on the configured thresholds.

  • The testing packets can be configured and the format of the packet depends on the underlying service on which the test is configured. For IP-based service, the IP or port values can be configured. For Ethernet-based service, unicast untagged or VLAN ID-tagged dot1p formats (IEEE 802.1p or packet classification Layer 2 headers) are supported. The Ethernet destination address and source address that you configured are used.

  • You can run RFC 2544 benchmarking inet tests on Layer 3 VPNs or virtual routers.

  • For an inet service, each test session needs to use a unique UDP port. On the initiator device, the source UDP port that you specify by using the source-udp-port statement must be unique and not used by other UDP services that terminate at the initiator. On the reflector device, the UDP port of the destination to be used in the UDP header for the generated frames by using the destination-udp-port statement must be unique and not used by other UDP services that terminate at the reflector.

  • You must start the test on the router that operates as the reflector before you start the test on router that functions as the initiator.

  • You must configure the size of the test packet based on the configured MTU of the packets.

  • For computation of the test results for a user-to-network interface (UNI) or ingress direction of an Ethernet pseudowire service, the customer edge (CE) device that is configured as a reflector for inet must have the reflected destination address resolved using ARP or a statically configured route must be present on the CE device to connect to the initiator.

  • For benchmarking tests on the UNI direction of an Ethernet pseudowire service, if reflection mode is configured, you must configure a static ARP entry. Otherwise, the tests fail when test frames on the UNI interface are reflected. ARP resolution does not enable a successful reflection of test frames for UNI interfaces.

  • For a CCC family and with the test performed in the egress or network-to-network interface (NNI) direction, the tests stop on the initiator and reflector when the pseudowire goes down.

  • For an RFC 2544 test that is run in the egress or network-to-network interface (NNI) direction of an Ethernet service for a CCC family, the ingress features are not applied.

  • In ACX5048 and ACX5096 routers, for a CCC family, the pseudowire has to be opened prior to the start of the RFC 2544 test and during the course of the test.

  • The configured packet size denotes the untagged packet size. Any additional VLAN in the payload causes the packet length to be increased correspondingly.

  • For an inet service, if you configure an interface on an initiator for the RFC 2544-based benchmarking test to be run without specifying the source IPv4 address for the test frames, the primary IP address of the interface is used for the test frames. If the primary IP address is not configured, the first IPv4 address of the interface is used. Similarly, for an unnumbered interface on an initiator on which the RFC 2544 test is run, the primary or the first IP address of the donor loopback interface is retrieved and used in the test frames. You must explicitly configure the source IPv4 address for the test frames by using the source-ipv4-address statement if you want a particular address to be used.

  • RFC 2544 test generates packets for performance benchmarking testing. The packets can be destined for known or unknown unicast MAC addresses, and they can be either tagged or untagged frames. UDP/IP packet is used as the frame payload. Refer to Configuring RFC 2544-Based Benchmarking Tests for the frame fields that can be configured.

  • Supported outer TPIDs for tagged frames are 0x8100, 0x88a8, 0x9100, and 0x9200.

  • RFC 2544 benchmark tests can be run in out-of-service and in in-service modes.

    Note:

    In out-of-service mode, while the test is running, all the data traffic sent to and from the UNI port under test on the service is interrupted. Control protocol packets are not interrupted.

    In in-service mode, while the test is running, only the data traffic corresponding to the test session is interrupted, rest of the data traffic flow sent to and from the UNI port under test on the service are not affected. Control protocol packets are not interrupted.

  • The source MAC address, destination MAC address, and the UNI port under test configured uniquely identifies the RFC 2544 benchmark test session (or test stream).

  • You can run only one test at a time. Multiple simultaneous tests cannot be run at a time.

  • The maximum theoretical test bandwidth supported by ACX Series routers for RFC 2544 test initiator or reflector is 1 Gbps. On ACX5048 and ACX5096 routers, the maximum theoretical test bandwidth supported for RFC 2544 reflector is 40 Gbps.

  • RFC 2544 tests can be run with different frame sizes. In ACX Series routers, the supported frame sizes are 64, 68, 72, 128, 256, 512, 768, 1024, 1280, 1518, 1522, 1600, 1728, 2496, 3584, 4016, 9104, and 9136 bytes.

  • The test uses round-trip traffic for performance measurement.

  • A history of the test results is stored in memory.

  • The test results can be copied to the local file system or a remote file system, optionally.

Note:

RFC 2544 tests cannot compute the performance attributes of multicast or broadcast traffic streams.

For ACX routers running Junos OS Evolved, the number of RFC 2544 reflector test sessions supported varies according to the interface speed as shown in Table 1.

Table 1: Number of RFC2544 Reflector Sessions Supported
Interface Speed ACX7509 ACX7100 ACX7024

10G

16 sessions

16 sessions

4 sessions

25G

12 sessions

16 sessions

4 sessions

50G

6 sessions

16 sessions

Not supported

100G

3 sessions

16 sessions

1 session

400G

None

3 sessions

None

For ACX routers running Junos OS Evolved that are initiating Layer 3 tests, we support 64 sessions per Packet Forwarding Engine.

Related Documentation