Thoughts on minimum-cost converged HA cluster storage

[wallewek]

Here's something to mull over.

I know that the standard recommendation for StarWind is RAID 10, and hey, I'm a long time fan of RAID 10 on heavy-duty servers. But over the last few years, as I've worked on VSAN HA environments, I keep thinking that maybe we're wasting money on multiple overlapping redundancies, particularly on budget installations.

Strictly speaking, maybe redundant storage hardware is unnecessary, once we have a VSAN cluster that is fully capable of continuing to run when one of the hosts goes offline due to failed storage hardware, and of rebuilding that storage redundancy once the hardware is replaced. Maybe we should even consider running totally non-redundant RAIDs like (shudder) RAID 0, and reap the cost-performance benefit.

Sure, it's nice not to have to worry how well the VSAN handles failures, or to worry about running in failover mode during recovery. But what's the defense? What's the down side? Is there something I'm missing here? Or is hardware RAID just an option for sites where the cost of down time justifies that extra layer of redundancy?

I looked for relevant discussion elsewhere on this forum, BTW, but searching for RAID got me 90 _pages_ of hits.

-- Ken

[anton (staff)]

You can go RAID0 locally (or RAID5 or RAID6 with SSDs as it's very close in terms of performance to RAID0 but you can lose one or two disks in a group and keep going) and do network RAID1 with StarWind on top of that.

Here's something to mull over.

I know that the standard recommendation for StarWind is RAID 10, and hey, I'm a long time fan of RAID 10 on heavy-duty servers. But over the last few years, as I've worked on VSAN HA environments, I keep thinking that maybe we're wasting money on multiple overlapping redundancies, particularly on budget installations.

Strictly speaking, maybe redundant storage hardware is unnecessary, once we have a VSAN cluster that is fully capable of continuing to run when one of the hosts goes offline due to failed storage hardware, and of rebuilding that storage redundancy once the hardware is replaced. Maybe we should even consider running totally non-redundant RAIDs like (shudder) RAID 0, and reap the cost-performance benefit.

Sure, it's nice not to have to worry how well the VSAN handles failures, or to worry about running in failover mode during recovery. But what's the defense? What's the down side? Is there something I'm missing here? Or is hardware RAID just an option for sites where the cost of down time justifies that extra layer of redundancy?

I looked for relevant discussion elsewhere on this forum, BTW, but searching for RAID got me 90 _pages_ of hits.

-- Ken

[wallewek]

Thanks, Anton. Fascinating. Good to know there's not some hidden gotcha!

BTW, I know StarWInd doesn't support "software" RAID, and does support "hardware" RAID, e.g. the kind with RAID controller cards from Adaptec and such.

But where does StarWinds stand on what I might call "firmware" RAID, the kind that comes with motherboards that support RAID1 and such only, and is configured in BIOS?

-- Ken

[Delo123]

Firmware Raid = Software Raid, Ken
Be very very carefull with those things. Bad performance, issues when needing to rebuild, unhappy users...

[wallewek]

Hmm. Thanks Anton.

Sounds like, if it's a choice between no RAID at all (just StarWind VSAN redundancy) vs. firmware RAID1, the better choice is no RAID at all.

It's certainly less complex. I'm always a fan of simplicity.

-- Ken

[Michael (staff)]

Hello Wallewek,
We are always happy to help you here!