diff options
Diffstat (limited to 'docs/testing/user/userguide/nsb/tc_prox_context_load_balancer_port.rst')
| -rw-r--r-- | docs/testing/user/userguide/nsb/tc_prox_context_load_balancer_port.rst | 109 |
1 files changed, 109 insertions, 0 deletions
diff --git a/docs/testing/user/userguide/nsb/tc_prox_context_load_balancer_port.rst b/docs/testing/user/userguide/nsb/tc_prox_context_load_balancer_port.rst new file mode 100644 index 000000000..ada02f829 --- /dev/null +++ b/docs/testing/user/userguide/nsb/tc_prox_context_load_balancer_port.rst @@ -0,0 +1,109 @@ +.. This work is licensed under a Creative Commons Attribution 4.0 International +.. License. +.. http://creativecommons.org/licenses/by/4.0 +.. (c) OPNFV, 2017 Intel Corporation. + +******************************************************* +Yardstick Test Case Description: NSB PROX Load Balancer +******************************************************* + ++-----------------------------------------------------------------------------+ +|NSB PROX test for NFVI characterization | +| | ++--------------+--------------------------------------------------------------+ +|test case id | tc_prox_{context}_lb-{port_num} | +| | | +| | * context = baremetal or heat_context | +| | * port_num = 4 | +| | | ++--------------+--------------------------------------------------------------+ +|metric | * Network Throughput; | +| | * TG Packets Out; | +| | * TG Packets In; | +| | * VNF Packets Out; | +| | * VNF Packets In; | +| | * Dropped packets; | +| | | ++--------------+--------------------------------------------------------------+ +|test purpose | The applciation transmits packets on one port and revieves | +| | them on 4 ports. | +| | The conventional 5-tuple is used in this test as it requires | +| | some extraction steps and allows defining enough distinct | +| | values to find the performance limits. | +| | | +| | The load is increased (adding more ports if needed) while | +| | packets are load balanced using a hash table of 8M entries | +| | | +| | The number of packets per second that can be forwarded | +| | determines the KPI. The default packet size is 64 bytes. | +| | | ++--------------+--------------------------------------------------------------+ +|configuration | The Load Balancer test cases are listed below: | +| | | +| | * tc_prox_baremetal_lb-4.yaml | +| | * tc_prox_heat_context_lb-4.yaml | +| | | +| | Test duration is set as 300sec for each test. | +| | Packet size set as 64 bytes in traffic profile. | +| | These can be configured | +| | | ++--------------+--------------------------------------------------------------+ +|test tool | PROX | +| | PROX is a DPDK application that can simulate VNF workloads | +| | and can generate traffic and used for NFVI characterization | +| | | ++--------------+--------------------------------------------------------------+ +|applicability | The PROX Load Balancer test cases can be configured with | +| | different: | +| | | +| | * packet sizes; | +| | * test durations; | +| | * tolerated loss; | +| | | +| | Default values exist. | +| | | ++--------------+--------------------------------------------------------------+ +|pre-test | For Openstack test case image (yardstick-samplevnfs) needs | +|conditions | to be installed into Glance with Prox and Dpdk included in | +| | it. The test need multi-queue enabled in Glance image. | +| | | +| | For Baremetal tests cases Prox and Dpdk must be installed in | +| | the hosts where the test is executed. The pod.yaml file must | +| | have the necessary system and NIC information | +| | | ++--------------+--------------------------------------------------------------+ +|test sequence | description and expected result | +| | | ++--------------+--------------------------------------------------------------+ +|step 1 | For Baremetal test: The TG and VNF are started on the hosts | +| | based on the pod file. | +| | | +| | For Heat test: Two host VMs are booted, as Traffic generator | +| | and VNF(Load Balancer workload) based on the test flavor. | +| | | ++--------------+--------------------------------------------------------------+ +|step 2 | Yardstick is connected with the TG and VNF by using ssh. | +| | The test will resolve the topology and instantiate the VNF | +| | and TG and collect the KPI's/metrics. | +| | | ++--------------+--------------------------------------------------------------+ +|step 3 | The TG will send packets to the VNF. If the number of | +| | dropped packets is more than the tolerated loss the line | +| | rate or throughput is halved. This is done until the dropped | +| | packets are within an acceptable tolerated loss. | +| | | +| | The KPI is the number of packets per second for 78 bytes | +| | packet size with an accepted minimal packet loss for the | +| | default configuration. | +| | | ++--------------+--------------------------------------------------------------+ +|step 4 | In Baremetal test: The test quits the application and unbind | +| | the dpdk ports. | +| | | +| | In Heat test: Two host VMs are deleted on test completion. | +| | | ++--------------+--------------------------------------------------------------+ +|test verdict | The test case will achieve a Throughput with an accepted | +| | minimal tolerated packet loss. | ++--------------+--------------------------------------------------------------+ + |