[Quesnelia] [non-ECS] Multi-tenants DI
Overview
- This document contains the results of testing Data Import for MARC Bibliographic records with an update job in the Quesnelia release on qcp1 environments with Kafka consolidated topics and file splitting features enabled on a non-ecs environment.
- PERF-818Getting issue details... STATUS
Summary
- Three tests of DI with update jobs with 25k records were carried out on the qcp1 cluster.
- Comparing the Quesnelia and Poppy releases, we observed improved DI update job durations in the Quesnelia environment. On average, these durations were 1 minute and 23 seconds shorter for a single tenant and 2 minutes shorter for two and three concurrent tenants in the Quesnelia environment.
- Memory consumption for mod-data-import was 60%, mod-inventory - 128%. No memory leaks are suspected.
- CPU utilization for mod-data-import spiked to 70%, mod-inventory - 150%.
- DB CPU utilization was 93%.
- DB connections - 1273 on average.
Test Results
Profile | Test # | Tenant | MARC File | DI Duration Quesnelia (hh:mm:ss) | Results |
---|---|---|---|---|---|
DI MARC Bib Update (PTF - Updates Success - 6) | 1 | fs09000000 | 25K.mrc | 00:15:55 | Completed |
2 | fs09000000 | 25K.mrc | 00:17:06 | Completed | |
fs07000001 | 25K.mrc | 00:28:17 | Completed | ||
3 | fs09000000 | 25K.mrc | 00:33:57 | Completed | |
fs07000001 | 25K.mrc | 00:36:07 | Completed | ||
fs07000002 | 25K.mrc | 00:41:55 | Completed |
Comparison
Test # | Quesnelia | Poppy | DI Delta | DI Delta per test (hh:mm:ss) Quesnelia /Poppy | ||||||
Profile | Tenant | DI Duration Quesnelia per tenant (hh:mm:ss) | DI Duration Quesnelia per test (hh:mm:ss) | Profile | Tenant | DI Duration Poppy (hh:mm:ss) | DI Duration Poppy per test (hh:mm:ss) | |||
---|---|---|---|---|---|---|---|---|---|---|
№1 25K.mrc | DI MARC Bib Update (PTF - Updates Success - 6) | fs09000000 | 00:15:55 | 00:15:55 | DI MARC Bib Update (PTF - Updates Success - 1) | fso9000000 | 00:17:18 | 00:17:18 | -0:01:23 | -0:01:23 |
№2 25K.mrc | fs09000000 | 00:17:06 | 00:45:23 | fso9000000 | 00:23:11 | 00:47:27 | -0:06:05 | -0:02:04 | ||
fs07000001 | 00:28:17 | fso7000001 | 00:24:16 | +0:04:01 | ||||||
№3 25K.mrc | fs09000000 | 00:33:57 | 1:51:59 | fso9000000 | 00:26:36 | 1:54:01 | +0:07:21 | -0:02:02 | ||
fs07000001 | 00:36:07 | fso7000001 | 00:38:01 | -0:01:54 | ||||||
fs07000002 | 00:41:55 | fso7000002 | 00:49:24 | -0:07:29 |
Resource utilization for Test #1 and Test #2
Service CPU Utilization
Here we can see that mod-inventory-b module used 171% CPU, mod-di-converter-storage-b used 83% CPU and nginx-okapi used 66% CPU
Service Memory Utilization
Here we can see that all modules show a stable trend.
DB CPU Utilization
DB CPU in the average was 88%.
DB Connections
DB connections was 1139.
DB load
Top SQL-queries
# | TOP 5 SQL statements |
---|---|
1 |
|
2 |
|
3 |
|
4 |
|
5 |
|
Resource utilization for Test #3
Service CPU Utilization
Here we can see that mod-inventory-b module used 140% CPU in average, mod-inventory-b used 117% CPU and nginx-okapi used 109% CPU
Service Memory Utilization
Here we can see that all modules show a stable trend.
DB CPU Utilization
DB CPU was 93%.
DB Connections
DB connections was 1273.
DB load
Top SQL-queries
# | TOP 5 SQL statements |
---|---|
1 |
|
2 |
|
3 |
|
4 |
|
5 |
|
Appendix
Infrastructure
PTF - environment Quesnelia (qcp1)
- 10 db.r6g.xlarge EC2 instances located in US East (N. Virginia)us-east-1
1 database instances, writer
Name Memory GIB vCPUs db.r6g.xlarge
32 GiB 4 vCPUs - MSK ptf-mobius-testing2
- 4 m5.2xlarge brokers in 2 zones
Apache Kafka version 2.8.0
EBS storage volume per broker 300 GiB
- auto.create.topics.enable=true
- log.retention.minutes=480
- default.replication.factor=2