GitHub user bsandy999 added a comment to the discussion: 15k TPS with Payload size of 52kb?
@sijie Thanks for the feedback. So as per the diagram, and as we are running POC only, we have deployed our pulsar cluster on 3 GCP machines, where we are running 1 broker, 1 bookKeeper, and 1 Zk on each machine. So we have scaled our setup horizontally. What if we scale bookies vertically on each machine? So we tried that as well running 3 bookies on a single machine (3 machines * 3 bookies on each machine = in total 9 bookies running ), but with this approach, we faced memory issues and load issues on the system, which causes random bookies shut down an unstable setup. **Configuration of Pulsar Instance** - Machine 1 - 4 core, 8 GB , 100 GB Disk. **( 1 broker, 1 bookie, 1 zk)** - Machine 2 - 4 core, 8 GB , 100 GB Disk. **( 1 broker, 1 bookie, 1 zk)** - Machine 3 - 4 core, 8 GB , 100 GB Disk. **( 1 broker, 1 bookie, 1 zk)** This is as per our understanding, if anything seems wrong, then please suggest the best solution where we can utilize each system component in the best manner GitHub link: https://github.com/apache/pulsar/discussions/19578#discussioncomment-5073275 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected]
