# Write performance in multi-node clusters?

**URL:** <https://discuss.aerospike.com/t/write-performance-in-multi-node-clusters/1048>\
**Category:** Operations\
**Created:** [March 11, 2015, 9:27pm UTC](https://discuss.aerospike.com/t/write-performance-in-multi-node-clusters/1048 "2015-03-11T21:27:40Z")\
**Posts on this page:** 1\
**Showing post:** 11

<div class="post-metadata">

**Author:** ![kporter](https://sea1.discourse-cdn.com/flex019/user_avatar/discuss.aerospike.com/kporter/32/515_2.png) [@kporter](https://discuss.aerospike.com/u/kporter)\
**Post date:** [May 1, 2015, 1:13am UTC](https://discuss.aerospike.com/t/write-performance-in-multi-node-clusters/1048/11 "2015-05-01T01:13:17Z")

</div>

Sorry for the delay 😱

> [@Joshua\_Buss](#):
>
> I was using clients on the same servers and different servers. Moving the clients to just remote servers lowered my throughput substantially

> [@Joshua\_Buss](#):
>
> No, performance was consistent regardless of whether migrations were happening

These two would indicate that your clients were the bottleneck and the fact that they are python makes that really likely. I suspect that you would have seen the TPS increase if you increased the number of clients indicating that the server could handle more than the clients could push.

Migrations should definently have an affect on performance especially peak performance.

I recommend using the java benchmark client to see how many transactions per second the server can handle. On internal machines the java benchmark can push upwards of 300,000 TPS per second.

---

_[View the full topic](https://discuss.aerospike.com/t/write-performance-in-multi-node-clusters/1048)._