This project collects the instructions to enable the Intel® QuickAssist Technology (QAT) on Origin community Distribution of Kubernetes. This technology improves performance by offloading the encryption/decryption and compression/decompression operations thereby reserving processor cycles for application and control processing.
This integration improves the response time of the OpenShift Router component which is an ingress point that routes external traffic to Kubernetes Service resources.
This project uses Vagrant tool for provisioning Virtual Machines automatically. The setup.sh script of the bootstrap-vagrant project contains the Linux instructions to install dependencies and plugins required for its usage. The following instruction installs Vagrant and Libvirt as Vagrant provider.
$ curl -fsSL http://bit.ly/initVagrant | PROVIDER=libvirt bash
Once Vagrant is installed, it's possible to provision an All-in-One OKD cluster and measure its performance through the following instruction:
$ vagrant up
The provisioning process starts updating the Kernel version using the packages provided by the Extra Packages for Enterprise Linux (EPEL) project. Once the Virtual Machine has automatically rebooted with the new Kernel loaded the provisioning process will continue with the setup of the OKD cluster through the usage of the cluster_up.sh. After its execution, the setup.sh bash script will take place creating components required by the Sample Application, the performance benchmarks will be performed on this Nginx App. Finally, the test.sh bash script will run k6 workloads to measure the Router's performance.
This section shows the results that was taken from the creation of two Virtual Machines spawned in the same server. The following table contains the configuration values used during their provisioning process.
| Environment variable | Value |
|---|---|
| OKD_VAGRANT_CPUS | 44 |
| OKD_VAGRANT_MEMORY | 24576 |
| OKD_ENABLE_QAT | false |
/\ |‾‾| /‾‾/ /‾/
/\ / \ | |_/ / / /
/ \/ \ | | / ‾‾\
/ \ | |‾\ \ | (_) |
/ __________ \ |__| \__\ \___/ .io
execution: local--------------------------------------------------] servertor
output: -
script: k6-config.js
duration: 20s, iterations: -
vus: 10, max: 10
init [----------------------------------------------------------] starting
✓ status was 200
✓ transaction time OK
checks.....................: 100.00% ✓ 20948 ✗ 0
data_received..............: 422 MB 21 MB/s
data_sent..................: 6.4 MB 318 kB/s
http_req_blocked...........: avg=11.44ms min=3.81ms med=10.92ms max=40.18ms p(90)=14.96ms p(95)=19.08ms
http_req_connecting........: avg=276.87µs min=85.76µs med=268.5µs max=4.9ms p(90)=329.6µs p(95)=354.35µs
http_req_duration..........: avg=7.33ms min=2.11ms med=6.61ms max=29.1ms p(90)=11.12ms p(95)=14.05ms
http_req_receiving.........: avg=894.35µs min=224.34µs med=653.67µs max=11.4ms p(90)=1.56ms p(95)=2.22ms
http_req_sending...........: avg=100.45µs min=20.93µs med=89.06µs max=7.52ms p(90)=129.08µs p(95)=147.94µs
http_req_tls_handshaking...: avg=11.02ms min=3.4ms med=10.52ms max=39.44ms p(90)=14.48ms p(95)=18.62ms
http_req_waiting...........: avg=6.34ms min=1.45ms med=5.65ms max=26.36ms p(90)=9.93ms p(95)=12.39ms
http_reqs..................: 10474 523.697242/s
iteration_duration.........: avg=19.03ms min=8.17ms med=17.58ms max=52.19ms p(90)=26.15ms p(95)=31.48ms
iterations.................: 10474 523.697242/s
vus........................: 10 min=10 max=10
vus_max....................: 10 min=10 max=10| Environment variable | Value |
|---|---|
| OKD_VAGRANT_CPUS | 44 |
| OKD_VAGRANT_MEMORY | 24576 |
| OKD_ENABLE_QAT | true |
/\ |‾‾| /‾‾/ /‾/
/\ / \ | |_/ / / /
/ \/ \ | | / ‾‾\
/ \ | |‾\ \ | (_) |
/ __________ \ |__| \__\ \___/ .io
execution: local--------------------------------------------------] servertor
output: -
script: k6-config.js
duration: 20s, iterations: -
vus: 10, max: 10
init [----------------------------------------------------------] starting
✓ status was 200
✓ transaction time OK
checks.....................: 100.00% ✓ 29898 ✗ 0
data_received..............: 602 MB 30 MB/s
data_sent..................: 8.6 MB 429 kB/s
http_req_blocked...........: avg=8.82ms min=3.33ms med=8.59ms max=39.8ms p(90)=11.21ms p(95)=12.12ms
http_req_connecting........: avg=605.48µs min=90.87µs med=532.26µs max=14.67ms p(90)=909.4µs p(95)=1.18ms
http_req_duration..........: avg=4.17ms min=1.09ms med=3.98ms max=28.76ms p(90)=5.71ms p(95)=6.42ms
http_req_receiving.........: avg=989.79µs min=163.34µs med=798.9µs max=17.72ms p(90)=1.74ms p(95)=2.2ms
http_req_sending...........: avg=122.85µs min=17.19µs med=83.16µs max=14.14ms p(90)=181.04µs p(95)=299.39µs
http_req_tls_handshaking...: avg=8.05ms min=2.82ms med=7.84ms max=39.05ms p(90)=10.31ms p(95)=11.19ms
http_req_waiting...........: avg=3.06ms min=683.97µs med=2.92ms max=25.03ms p(90)=4.43ms p(95)=5ms
http_reqs..................: 14949 747.443662/s
iteration_duration.........: avg=13.3ms min=5.24ms med=13.07ms max=49.68ms p(90)=16.31ms p(95)=17.59ms
iterations.................: 14949 747.443662/s
vus........................: 10 min=10 max=10
vus_max....................: 10 min=10 max=10As we can see the major improvements when the QuickAssit Technology is
used are on the http_req_duration, http_req_tls_handshaking,
http_req_waiting and http_reqs values. Those fast responses on the
Web Application can result in better user experience or increase the
capacity to support major number of workloads by server.
Apache-2.0