vnf_name
- name of the VNF being tested.components
- all the components in a clustered VNF. If it is a single VNF, only one component is included.flavors
- for components, specify what instance sizes can be used during test.deploy
- include information needed to deploy the VNF.script
- indicates the full path to VNF deployment script.server
- specification and requirements on physical server to run tests.number
- number of servers to use for VNF deployment.knobs
- knobs to test. Support pinning
for vCPU core pinning, and sr-iov
. More knob-support will be provided.
mode
- information about testing mode. Only one of the three testing modes can be activated during the entire testing process.sizes
- use custom sizing test mode. If is option is set, all other options in mode
will be ignored. Specify all deployment sizes to test, each of which should include component name, flavor name and number of instance for that flavor. If the flavor name given here is inconsistent with the ones given in components
, this specific deployment size will be ignored.
# Exmaple of three Clearwater deployment sizes to test: bn1s-sp1s-hs1s, bn1m-sp2s-h2s, bn1l-sp1m2s-hs2m,
sizes:
- bono: [cw1.small: 1]
sprout: [cw1.small: 1]
homestead: [cw1.small: 1]
homer: [cw1.small: 1]
- bono: [cw1.medium: 1]
sprout: [cw1.small: 2]
homestead: [cw1.small: 2]
homer: [cw1.small: 1]
- bono: [cw1.large: 1]
sprout: [cw1.meidum: 1, cw1.small: 2]
homestead: [cw1.medium: 2]
homer: [cw1.small: 1]
search
- define search algorithm: directed
and exhaustive
.resource
- define resource constrains for directed
and exhaustive
, including vcpu
, mem
and net
. Currently, only one of three can be defined at a time.target-load
- define the target workload for directed search.scale
- define scaling method for directed search.generator
- indicates the full path of workload generator script.rates
- information about at what rate to generate workload. Only one of custom rates and auto increase can be defined.custom
- defines specific rates to test. range
- defines the minimum and maximum value of workload rates.increase
- defines how to increase the rate linearly. May support other rate increasing mode in future.repeat
- defines how many times to repeat the same test for averaging results.stop
- defines when to stop testing the current deployment size. This is useful when the VNF performance reaches system capacity before maximum rate is applied. It is currently set to a proportion of workload rates. For instance, if the performance indicator (e.g., successful call rate) drops below 0.8 of target workload rate, testing will stop.