Create a RAID group of as many nodes as you have. With our current RAID 0+1 recommended (RAID0 on every node and RAID1 done over network)
you'll have performance increase with every single node added to MPIO StarWind cluster. So 3 is a way to go now (4 is coming).
RAID0 with 3 nodes for sure. RAID10 with 2 node cluster (or have a decent backup to re-provision node from scratch, then maybe you'll be fine with RAID0 on a 2 node)
RAID0 or RAID10 on every node (see above)
Round Robin. StarWind is true Active-Active so scales out.
10 GbE for a backbone for sure. As many ports as you can afford (used for sync that's a bottleneck always).
That;s all... Network tweaks for HA and non-HA are the same.
OIMJB wrote:Question is simple: For HA target
What is the best scenario to support high performance read i/o operations for multiple HA targets - 2 node:
- Hardware RAID level
- MPIO mode and settings
- Network type and tweaks
In real world, read i/o seems to be the real reason of many poor vm performance and host performance stall.
We talk for about 30 HA targets and so many vm.
Best regards.
