Hi,
It all depends on the amount of load that the services and/or data/compute
have.
You would need to experiment and see what stats the nodes give in terms of
cpu/memory utilization.
Look for these messages:
^-- CPU [cur=1.93%, avg=0.76%, GC=0%]
^-- PageMemory [pages=200]
^-- Heap [used=128MB, free=86.88%, comm=981MB]
^-- Off-heap [used=0MB, free=99.98%, comm=80MB]
^-- sysMemPlc region [used=0MB, free=99.21%, comm=40MB]
^-- default region [used=0MB, free=100%, comm=0MB]
^-- TxLog region [used=0MB, free=100%, comm=40MB]
^-- Outbound messages queue [size=0]
^-- Public thread pool [active=0, idle=0, qSize=0]
^-- System thread pool [active=0, idle=6, qSize=0]
^-- Striped thread pool [active=0, idle=8, qSize=0]
Pay attention to thread/cpu/memory/heap/off-heap usage.
You can use Ignite's benchmarking tool to prototype your app:
https://apacheignite.readme.io/docs/perfomance-benchmarking
Details on capacity planning:
https://apacheignite.readme.io/docs/capacity-planning
Ignite has new metrics available:
https://apacheignite.readme.io/docs/new-metrics
Also take a look at: https://apacheignite.readme.io/docs/memory-metrics
and: https://apacheignite.readme.io/docs/cache-metrics
These show you how to use JMX to monitor your apps.
*If services and data/compute are deployed on separate pods, is it a good
practice to collocate the pods on the same node*
This again depends on your use case. Usually it is best to have the
Kubernetes controller decide where to place the pods, but your situation
might be different.
Here is one example:
https://apacheignite.readme.io/docs/collocate-compute-and-data where you
might think about collocating things together.
Thanks, Alex
--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/
