This message was deleted.
# general
s
This message was deleted.
b
Couple of things. We did some basic benchmarking with OMB and these are the values we were able to achieve. The test bench had the zookeepers to be separate. OMB with 3 pulsar nodes with 1 broker and 1 bookie. Each node has two NVME SSD. Journal and SSD must be separated to different disks. The throughput that was achieved is 7 million messages with 100bytes per message. AWS EC2 was used with 24vcpu.
cc: @Dave Fisher
Just incase you're curious what OMB is..take a look at this Short 5MAP video

https://www.youtube.com/watch?v=UTzOcdNCtzc▾

r
```Broker gets down randomly.
Bookie gets down randomly. ```
Have you checked the logs for why the broker or bookie is going down? If its OOM issue, you may have to tune your JVM settings accordingly
s
below is the log desc