-
Notifications
You must be signed in to change notification settings - Fork 0
Performance tests
Aleksey Vorona edited this page Aug 29, 2017
·
1 revision
This page will document performance tests I am executing.
A scalable solution should scale both ways. Let's see how UncleK performs on a t2.micro
AWS instance. All on one box.
- 1 instance, 4 topics, one of them is compacted, one has a short retention.
- CPU when idle: 0.1%
- CPU with 1 slow (300 records/s) producer: 0.4%
- RAM: 75Mb used, 312Mb cache
Not bad really. Considering that default Apache Kafka installation won't even start on this instance. After some tweaking (export KAFKA_HEAP_OPTS="-Xmx512M -Xms512M"
) one can get it to launch.
- 1 instance, 4 topics, 1 partition each
- CPU when idle: 2.6%
- RAM: 634Mb used, 239Mb cache