Reputation: 13342
I got an example, where I'm testing hazelcast performance for simple cases.
I have 4 cores machine, 1.2 GHz on each.
I'm sending 100 bunches by 100 "clients" (actually just lopp in one thread).
So in the end I have 10 000 bunches of tickets.
The performance is about:
putAsync
set
I consider this as low performance.
Questions: Is it? How could I improve it?
All code is there - configs. scala, sbt project
Update:
Running time= test time:
So "how long time" = test time.
Where: newer machine is: 8 cores 2.2 Ghz Older machine: 4 cores, 1,2 Hgz
Quite big difference in performance depending on machine.
Upvotes: 1
Views: 490
Reputation: 11307
Is it a single node cluster?
To give you an idea; on our old test lab with 4 dual socket sandybridge boxes with 24 ht threads and 1GbE card, we can easily do 300/400k inserts per second.
If we do readonly operation, the performance is +/- 700k ops/second since we don't need to deal with a backup.
This is member only setup; client will be a bit slower.
The putAsync should be very fast (although I don't know much about the client). The putAsync allows the IO pipelines to be filled up nicely and you can lift on top of a batching effect.
Upvotes: 1