Environment and Resources

Kubernetes cluster with 4 worker nodes, sharing hardware configuration shown in a table below, is deployed in OpenStack cloud operating system. The test components in docker containers are further deployed on the Kubernetes cluster.

...

CPU

...

Test architecture

In order to conduct client tests this will be conducted in following architecture:

Image Added

• HV-VES Client - produces high amount of events for processing.
• Processing Consumer - consumes events from Kafka topics and creates performance metrics.
• Offset Consumer - reads Kafka offsets.
• Prometheus - sends requests for performance metrics to HV-VES, Processing Consumer and Offset Consumer, provides data to Grafana.
• Grafana - delivers analytics and its visualization.

Link between HV-VES Client and HV-VES is TLS secured (provided scripts generate and place certificates on proper containers).

Environment and Resources

Kubernetes cluster with 4 worker nodes, sharing hardware configuration shown in a table below, is deployed in OpenStack cloud operating system. The test components in docker containers are further deployed on the Kubernetes cluster.

Network Performance

Pod measurement method

In order to check cluster network performance tests with usage of Iperf3 have been applied. Iperf is a tool for measurement of the maximum bandwidth on IP networks, it runs on two mode: server and client. We used a docker image: networkstatic/iperf3.

Following deployment creates a pod with iperf (server mode) on one worker, and one pod with iperf client for each worker.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: iperf3-server
  namespace: onap
  labels:

Network Performance

Pod measurement method

In order to check cluster network performance tests with usage of Iperf3 have been applied. Iperf is a tool for measurement of the maximum bandwidth on IP networks, it runs on two mode: server and client. We used a docker image: networkstatic/iperf3.

Following deployment creates a pod with iperf (server mode) on one worker, and one pod with iperf client for each worker.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: iperf3-server
  namespace: onap
  labels:
    app: iperf3-server
spec:
  replicas: 1
  selector:
    matchLabels:
      app: iperf3-server
spec:
  replicas: 1
  selector:
    matchLabels:
      app: iperf3-server
  template:
    metadata:
      labels:
        app: iperf3-server
    spec:
      containers:
        - name: iperf3-server
          image: networkstatic/iperf3
          args: ['-s']
          ports:
            - containerPort: 5201
              name: server


---

apiVersion: v1
kind: Service
metadata:
  name: iperf3-server
  namespace: onap
spec:
  selector:
    app: iperf3-server
  ports:
    - protocol: TCP
      port: 5201
      targetPort: server



---

apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: iperf3-clients
  namespace: onap
  labels:
    app: iperf3-client
spec:
  selector:
    matchLabels:
      app: iperf3-client
  template:
    metadata:
      labels:
        app: iperf3-client
    spec:
      containers:
        - name: iperf3-client
          image: networkstatic/iperf3
          command: ['/bin/sh', '-c', 'sleep infinity']

...

• worker1 (136 MBytes/sec)

  Code Block
  title results collapse true
  Connecting to host iperf3-server, port 5201 [ 4] local 10.42.5.127 port 39752 connected to 10.43.25.161 port 5201 [ ID] Interval Transfer Bandwidth Retr Cwnd [ 4] 0.00-1.00 sec 141 MBytes 141 MBytes/sec 32 673 KBytes [ 4] 1.00-2.00 sec 139 MBytes 139 MBytes/sec 0 817 KBytes [ 4] 2.00-3.00 sec 139 MBytes 139 MBytes/sec 0 936 KBytes [ 4] 3.00-4.00 sec 138 MBytes 137 MBytes/sec 0 1.02 MBytes [ 4] 4.00-5.00 sec 138 MBytes 137 MBytes/sec 0 1.12 MBytes [ 4] 5.00-6.00 sec 129 MBytes 129 MBytes/sec 0 1.20 MBytes [ 4] 6.00-7.00 sec 129 MBytes 129 MBytes/sec 0 1.27 MBytes [ 4] 7.00-8.00 sec 134 MBytes 134 MBytes/sec 0 1.35 MBytes [ 4] 8.00-9.00 sec 135 MBytes 135 MBytes/sec 0 1.42 MBytes [ 4] 9.00-10.00 sec 135 MBytes 135 MBytes/sec 45 1.06 MBytes - - - - - - - - - - - - - - - - - - - - - - - - - [ ID] Interval Transfer Bandwidth Retr [ 4] 0.00-10.00 sec 1.32 GBytes 136 MBytes/sec 77 sender [ 4] 0.00-10.00 sec 1.32 GBytes 135 MBytes/sec receiver

  
  

• worker2 (87 MBytes/sec)

  Code Block
  title results collapse true
  Connecting to host iperf3-server, port 5201 [ 4] local 10.42.3.188 port 35472 connected to 10.43.25.161 port 5201 [ ID] Interval Transfer Bandwidth Retr Cwnd [ 4] 0.00-1.00 sec 88.3 MBytes 88.3 MBytes/sec 121 697 KBytes [ 4] 1.00-2.00 sec 96.2 MBytes 96.3 MBytes/sec 0 796 KBytes [ 4] 2.00-3.00 sec 92.5 MBytes 92.5 MBytes/sec 0 881 KBytes [ 4] 3.00-4.00 sec 90.0 MBytes 90.0 MBytes/sec 0 957 KBytes [ 4] 4.00-5.00 sec 87.5 MBytes 87.5 MBytes/sec 0 1.00 MBytes [ 4] 5.00-6.00 sec 88.8 MBytes 88.7 MBytes/sec 0 1.06 MBytes [ 4] 6.00-7.00 sec 80.0 MBytes 80.0 MBytes/sec 0 1.12 MBytes [ 4] 7.00-8.00 sec 81.2 MBytes 81.3 MBytes/sec 25 895 KBytes [ 4] 8.00-9.00 sec 85.0 MBytes 85.0 MBytes/sec 0 983 KBytes [ 4] 9.00-10.00 sec 83.8 MBytes 83.7 MBytes/sec 0 1.03 MBytes - - - - - - - - - - - - - - - - - - - - - - - - - [ ID] Interval Transfer Bandwidth Retr [ 4] 0.00-10.00 sec 873 MBytes 87.3 MBytes/sec 146 sender [ 4] 0.00-10.00 sec 870 MBytes 87.0 MBytes/sec receiver

  
  

• worker3 (135 MBytes/sec)

  Code Block
  title results collapse true
  Connecting to host iperf3-server, port 5201 [ 4] local 10.42.4.182 port 35288 connected to 10.43.25.161 port 5201 [ ID] Interval Transfer Bandwidth Retr Cwnd [ 4] 0.00-1.00 sec 129 MBytes 129 MBytes/sec 45 1.17 MBytes [ 4] 1.00-2.00 sec 134 MBytes 134 MBytes/sec 32 1.25 MBytes [ 4] 2.00-3.00 sec 135 MBytes 135 MBytes/sec 0 1.32 MBytes [ 4] 3.00-4.00 sec 139 MBytes 139 MBytes/sec 0 1.40 MBytes [ 4] 4.00-5.00 sec 144 MBytes 144 MBytes/sec 0 1.47 MBytes [ 4] 5.00-6.00 sec 131 MBytes 131 MBytes/sec 45 1.14 MBytes [ 4] 6.00-7.00 sec 129 MBytes 129 MBytes/sec 0 1.25 MBytes [ 4] 7.00-8.00 sec 134 MBytes 134 MBytes/sec 0 1.33 MBytes [ 4] 8.00-9.00 sec 138 MBytes 138 MBytes/sec 0 1.39 MBytes [ 4] 9.00-10.00 sec 135 MBytes 135 MBytes/sec 0 1.44 MBytes - - - - - - - - - - - - - - - - - - - - - - - - - [ ID] Interval Transfer Bandwidth Retr [ 4] 0.00-10.00 sec 1.31 GBytes 135 MBytes/sec 122 sender [ 4] 0.00-10.00 sec 1.31 GBytes 134 MBytes/sec receiver

  
  

• worker0 (2282 MBytes/sec) (iperf client and server exist on same worker )

  Code Block
  title results collapse true
  Connecting to host iperf3-server, port 5201 [ 4] local 10.42.6.132 port 51156 connected to 10.43.25.161 port 5201 [ ID] Interval Transfer Bandwidth Retr Cwnd [ 4] 0.00-1.00 sec 2.13 GBytes 2185 MBytes/sec 0 536 KBytes [ 4] 1.00-2.00 sec 1.66 GBytes 1702 MBytes/sec 0 621 KBytes [ 4] 2.00-3.00 sec 2.10 GBytes 2154 MBytes/sec 0 766 KBytes [ 4] 3.00-4.00 sec 1.89 GBytes 1937 MBytes/sec 0 1.01 MBytes [ 4] 4.00-5.00 sec 1.87 GBytes 1914 MBytes/sec 0 1.39 MBytes [ 4] 5.00-6.00 sec 2.76 GBytes 2826 MBytes/sec 0 1.39 MBytes [ 4] 6.00-7.00 sec 1.81 GBytes 1853 MBytes/sec 792 1.09 MBytes [ 4] 7.00-8.00 sec 2.54 GBytes 2600 MBytes/sec 0 1.21 MBytes [ 4] 8.00-9.00 sec 2.70 GBytes 2763 MBytes/sec 0 1.34 MBytes [ 4] 9.00-10.00 sec 2.82 GBytes 2889 MBytes/sec 0 1.34 MBytes - - - - - - - - - - - - - - - - - - - - - - - - - [ ID] Interval Transfer Bandwidth Retr [ 4] 0.00-10.00 sec 22.3 GBytes 2282 MBytes/sec 792 sender [ 4] 0.00-10.00 sec 22.3 GBytes 2282 MBytes/sec receiver

Average speed (without worker 0 )  : 119 MBytes/sec

HV-VES Performance

Preconditions

Before start tests, download docker image of producer which is available here. To extract image locally use command: 

docker load < hv-collector-go-client.tar.gz  

• [ 4] 0.00-10.00 sec 22.3 GBytes 2282 MBytes/sec 792 sender [ 4] 0.00-10.00 sec 22.3 GBytes 2282 MBytes/sec receiver

  
  

Average speed (without worker 0 )  : 119 MBytes/sec

Test Setup

Preconditions

• Installed ONAP (Frankfurt)
• Plain TCP connection between HV-VES and clients (default configuration)
• Metric port exposed on HV-VES service

In order to reach metrics endpoint in HV-VES there is a need to add the following lines in the ports section of HV-VES service configuration file:

  - name: port-t-6060
	port: 6060
	protocol: TCP
	targetPort: 6060

Before start tests, download docker image of producer which is available here:

To extract image locally use command: 

docker load < hv-collector-go-client.tar.gz  

Modify tools/performance/cloud/producer-pod.yaml file to use the above image and set imagePullPolicy to IfNotPresent:

...
spec:
  containers:
    - name: hv-collector-producer
      image: onap/org.onap.dcaegen2.collectors.hv-ves.hv-collector-go-client:latest
      imagePullPolicy: IfNotPresent
      volumeMounts:
...

To execute performance tests we have to run functions from a shell script cloud-based-performance-test.sh in HV-VES project directory: ~/tools/performance/cloud/

1. First we have to generate certificates in ~/tools/ssl folder by using gen_certs. This step only needs to be performed during the first test setup (or if the generated files have been deleted).

   Code Block
   title Generating certificates
   ./cloud-based-performance-test.sh gen_certs

   
   

2. Then we call setup in order to send certificates to HV-VES, and deploy Consumers, Prometheus, Grafana and create their ConfigMaps.

   Code Block
   title Setting up the test environment
   ./cloud-based-performance-test.sh setup

   
   

3. After completing previous steps we can call the start function, which provides Producers and starts the test.

   Code Block
   title Performing the test
   ./cloud-based-performance-test.sh start

   For the start function we can use optional arguments:

   --load	should the test keep defined number of running producers until script interruption (false)
   --containers	number of producer containers to create (1)
   --properties-file	path to file with benchmark properties (./test.properties)
   --retention-time-minutes	retention time of messages in kafka in minutes (60)

   Example invocations of test start:

   Code Block
   title Starting performance test with single producers creation
   ./cloud-based-performance-test.sh start --containers 10

   The command above starts the test that creates 10 producers which send the amount of messages defined in test.properties once.

   Code Block
   title Starting performance test with constant messages load
   ./cloud-based-performance-test.sh start --load true --containers 10 --retention-time-minutes 30

   This invocation starts load test, meaning the script will try to keep the amount of running containers at 10 with kafka message retention of 30 minutes.

   
   

   The test.properties file contains Producers and Consumers configurations and it allows setting following properties:

   Producer
   hvVesAddress	HV-VES address (dcae-hv-ves-collector.onap:6061)
   client.count	Number of clients per pod (1)
   message.size	Size of a single message in bytes (16384)
   message.count	Amount of messages to be send by each client (1000)
   message.interval	Interval between messages in miliseconds (1)
   Certificates paths
   client.cert.path	Path to cert file (/ssl/client.p12)
   client.cert.pass.path	Path to cert's pass file (/ssl/client.pass)
   Consumer
   kafka.bootstrapServers	Adress of Kafka service to consume from (message-router-kafka:9092)
   kafka.topics	Kafka topics to subscribe to (HV_VES_PERF3GPP)

   
   Results can be accessed under following links:

   • Prometheus: http://slave1:30000/graph?g0.range_input=1h&g0.expr=hv_kafka_consumer_travel_time_seconds_count&g0.tab=1
   • Grafana: http://slave1:30001/d/V94Kjlwmz/hv-ves-processing?orgId=1&refresh=5s

HV-VES Performance test results

With dmaap Kafka

Conditions

...

Raw results data with screenshots can be found in following files:

• Series 1 - results_series_1_with_dmaap.zip
• Series 2 - results_series_2_with_dmaap.zip

Test Results - series 1

Expand
title Click here to see results...
Below tables show the test results across a wide range of containers' number.  NUMBER OF PRODUCERS TOTAL MESSAGES PROCESSED DIFFERENCE BETWEEN ALL MESSAGES AND SENT TO HV-VES AVERAGE PROCESSING TIME IN HV-VES WITHOUT ROUTING [ms] AVERAGE LATENCY TO HV-VES OUTPUT WITH ROUTING [ms] PEAK INCOMING DATA RATE [MB/s] PEAK PROCESSING MESSAGE QUEUE SIZE PEAK CPU LOAD [%] PEAK MEMORY USAGE [GB] RESULTS PRESENTED IN GRAFANA 2 180000 0 0.03 54 1.6 175 2.9 0.38 Image RemovedImage RemovedImage RemovedImage AddedImage AddedImage Added 2 180000 0 0.021 8.5 1.6 6 3.2 0.37 Image RemovedImage RemovedImage RemovedImage AddedImage AddedImage Added 2 180000 0 0.022 14.8 1.6 66 2.4 0.38 Image RemovedImage RemovedImage RemovedImage AddedImage AddedImage Added 2 180000 0 0.023 19.4 1.6 59 2.2 0.38 Image RemovedImage RemovedImage RemovedImage AddedImage AddedImage Added 2 180000 0 0.021 7.6 1.6 58 2.6 0.36 Image AddedImage Added Image Added Image RemovedImage Removed Image Removed NUMBER OF PRODUCERS TOTAL MESSAGES PROCESSED DIFFERENCE BETWEEN ALL MESSAGES AND SENT TO HV-VES AVERAGE PROCESSING TIME IN HV-VES WITHOUT ROUTING [ms] AVERAGE LATENCY TO HV-VES OUTPUT WITH ROUTING [ms] PEAK INCOMING DATA RATE [MB/s] PEAK PROCESSING MESSAGE QUEUE SIZE PEAK CPU LOAD [%] PEAK MEMORY USAGE [GB] RESULTS PRESENTED IN GRAFANA 4 315868 44824 82 260 3,2 2400 3 0,4 Image RemovedImage Added Image RemovedImage RemovedImage AddedImage Added 4 359193 807 3660 10408 3,2 3100 4,9 0,78 Image RemovedImage RemovedImage RemovedImage AddedImage AddedImage Added 4 359401 591 1350 3486 3.7 3000 4,9 0,56 Image RemovedImage RemovedImage RemovedImage AddedImage AddedImage Added 4 360000 0 1.5 130 3.3 1200 4.1 0.39 Image RemovedImage RemovedImage AddedImage Added Image RemovedImage Added 4 360000 0 0.02 57 3.3 180 3.5 0.38 Image RemovedImage RemovedImage RemovedImage AddedImage AddedImage Added NUMBER OF PRODUCERS TOTAL MESSAGES PROCESSED DIFFERENCE BETWEEN ALL MESSAGES AND SENT TO HV-VES AVERAGE PROCESSING TIME IN HV-VES WITHOUT ROUTING [ms] AVERAGE LATENCY TO HV-VES OUTPUT WITH ROUTING [ms] PEAK INCOMING DATA RATE [MB/s] PEAK PROCESSING MESSAGE QUEUE SIZE PEAK CPU LOAD [%] PEAK MEMORY USAGE [GB] RESULTS PRESENTED IN GRAFANA 6 358475 1525 6600 17875 4.9 4900 5.8 0.97 Image RemovedImage RemovedImage RemovedImage AddedImage AddedImage Added 6 358494 1506 7080 20182 4.7 4500 4.5 0.97 Image RemovedImage RemovedImage RemovedImage AddedImage AddedImage Added 6 359244 756 4900 17231 4.9 1900 3.9 0.93 Image RemovedImage RemovedImage RemovedImage AddedImage AddedImage Added 6 358944 1056 6150 17200 4.9 4800 5 0.97 Image RemovedImage RemovedImage RemovedImage AddedImage AddedImage Added 6 358689 1311 5410 17202 4.9 4100 5 0.96 Image RemovedImage RemovedImage RemovedImage AddedImage AddedImage Added

Test Results - series 2

Expand
title Click here to see results...
NUMBER OF PRODUCERS TOTAL MESSAGES PROCESSED DIFFERENCE BETWEEN ALL MESSAGES AND SENT TO HV-VES AVERAGE PROCESSING TIME IN HV-VES WITHOUT ROUTING [ms] AVERAGE LATENCY TO HV-VES OUTPUT WITH ROUTING [ms] PEAK INCOMING DATA RATE [MB/s] PEAK PROCESSING MESSAGE QUEUE SIZE PEAK CPU LOAD [%] PEAK MEMORY USAGE [GB] RESULTS PRESENTED IN GRAFANA 2 180000 0 0.026 8.9 1.6 11 3 0.35 Image RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage AddedImage AddedImage AddedImage AddedImage AddedImage Added 2 180000 0 0.026 11.3 1.6 23 3.2 0.35 Image Added Image AddedImage AddedImage AddedImage AddedImage Added Image Removed Image RemovedImage RemovedImage RemovedImage RemovedImage Removed NUMBER OF PRODUCERS TOTAL MESSAGES PROCESSED DIFFERENCE BETWEEN ALL MESSAGES AND SENT TO HV-VES AVERAGE PROCESSING TIME IN HV-VES WITHOUT ROUTING [s] AVERAGE LATENCY TO HV-VES OUTPUT WITH ROUTING [ms] PEAK INCOMING DATA RATE [MB/s] PEAK PROCESSING MESSAGE QUEUE SIZE PEAK CPU LOAD [%] PEAK MEMORY USAGE [GB] RESULTS PRESENTED IN GRAFANA 4 360000 0 0.022 41.3 3.2 130 4 0.35 Image AddedImage AddedImage Added Image AddedImage AddedImage Added Image RemovedImage RemovedImage Removed Image RemovedImage RemovedImage Removed 4 360000 0 0.023 40.7 3.2 370 4.2 0.35 Image RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage AddedImage AddedImage AddedImage AddedImage AddedImage Added NUMBER OF PRODUCERS TOTAL MESSAGES PROCESSED DIFFERENCE BETWEEN ALL MESSAGES AND SENT TO HV-VES AVERAGE PROCESSING TIME IN HV-VES WITHOUT ROUTING [ms] AVERAGE LATENCY TO HV-VES OUTPUT WITH ROUTING [ms] PEAK INCOMING DATA RATE [MB/s] PEAK PROCESSING MESSAGE QUEUE SIZE PEAK CPU LOAD [%] PEAK MEMORY USAGE [GB] RESULTS PRESENTED IN GRAFANA 6 359317 683 6240 15965 4.8 4500 4.3 1 Image RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage AddedImage AddedImage AddedImage AddedImage AddedImage Added 6 359217 783 6490 16834 4.8 4200 5.8 1 Image RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage AddedImage AddedImage AddedImage AddedImage AddedImage Added

No DMaaP Kafka SetUp

...

Modify tools/performance/cloud scripts to match the names in your deployments, described in the previous step. Here is a diff file (you may need to adapt it to the current code situation):

tools.diff 

Go to tools/performance/cloud and reboot the environment:

./reboot-test-environment.sh -v

Now you are ready to run the test.

Without DMaaP Kafka

Conditions

To gather tcpdump data another container was added to hv-ves deployment in kubernetes and producer-pod.yaml (command to get tcpdump data file: kubectl cp -n onap <pod-name>:/tcpdump.pcap -c tcpdump ./<pod-name>.pcap)the environment:

./reboot-test-environment.sh -v

Now you are ready to run the test.

Without DMaaP Kafka

Conditions

Tests were performed with following configuration:

...

Raw results data with screenshots can be found in following files:

• Series 1 - results_series_1.zip
• Series 2 - results_series_2.zip

To see custom Kafka metrics you may want to change kafka-and-producers.json (located in HV-VES project directory: tools/performance/cloud/grafana/dashboards) to

...

Expand
title Click here to see results...
Below tables show the test results across a wide range of containers' number.  NUMBER OF PRODUCERS TOTAL MESSAGES PROCESSED DIFFERENCE BETWEEN ALL MESSAGES AND SENT TO HV-VES AVERAGE PROCESSING TIME IN HV-VES WITHOUT ROUTING [ms] AVERAGE LATENCY TO HV-VES OUTPUT WITH ROUTING [ms] PEAK INCOMING DATA RATE [MB/s] PEAK PROCESSING MESSAGE QUEUE SIZE PEAK CPU LOAD [%] PEAK MEMORY USAGE [GB] RESULTS PRESENTED IN GRAFANA 2 180000 0 0.024 1.5 1.6 3 3.5 0.37 Image RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage AddedImage AddedImage AddedImage AddedImage AddedImage Added 4 360000 0 0.020 1.7 3.2 3 5.7 0.37 Image RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage AddedImage AddedImage AddedImage AddedImage AddedImage Added 6 540000 0 0.02 2.6 4.8 24 6.0 0.37 Image AddedImage Added Image AddedImage AddedImage AddedImage Added Image RemovedImage Removed Image RemovedImage RemovedImage RemovedImage Removed 8 720000 0 0.02 2.8 6.4 14 8.5 0.37 Image RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage AddedImage AddedImage AddedImage AddedImage AddedImage Added 10 900000 0 0.02 7.3 8.1 552 8.5 0.38 Image RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage AddedImage AddedImage AddedImage AddedImage AddedImage Added 12 10800000 0 0.02 78 9.7 1077 7.5 0.41 Image AddedImage AddedImage Added Image AddedImage Added Image Added Image RemovedImage RemovedImage Removed Image RemovedImage Removed Image Removed 14 12600000 0 6140 19000 13.0 10630 9.8 0.99 Image RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage RemovedImage AddedImage AddedImage AddedImage AddedImage AddedImage Added

Test results - series 2