Score:2

How does VMWare vSAN's RAID5/6 work

mx flag

BLUF: How does VMWare's implementation of RAID5/6 work?

Below is VMWare's chart for RAID5/6 in vSAN:

enter image description here

It seems VMWare is altering RAID's terminology. When I hear RAID6 I think two parity of something; usually two parity drives and subsequently we can tolerate two failures. I'm having trouble making sense of this chart for a couple of reasons:

  • What is the difference in row 2/4 between RAID5 and 6? They seem to be identical.
  • Why is RAID 1 (mirroring) listed twice, but with different failures to tolerate and different capacity required? It seems to follow that additional copies of the data must be made, but if that's the case then it wouldn't be RAID1 (at least not as I understand RAID1)
  • How is it that they have a RAID6 that only has a single failure to tolerate

As I wrote the above question it dawned on me that what the might mean is that they are running a RAID6 with VMWare's FTT set to 1. However, if that is the case how does that work? How are they getting data size 100GBs with 133GBs required?

Score:2
ca flag

Row #1 describes two-way mirroring or traditional RAID1, where your data are duplicated to another disk. Writing 100 GB means using 200 GB in total. You can tolerate one disk failure for each mirror.

Row #2 is about traditional 4-disk RAID5, where you lose one disk worth of capacity (ie: 100 GB written are expanded to 133 GB due to parity). Again, you can lose a single disk for each RAID5 set.

Row #3 describes triple-way mirroring or RAID1E. Writing 100 GB really means 300 GB are going to be written, but you can lose two disks for each mirror.

Row #4 really is about 6-disk RAID6 (including RAID5 here is a simplification, the numbers are going to differ), where 100 GB expands to 150 GB due to double parity. Again, you can lose up to two disk for each array.

Grant Curell avatar
mx flag
I'm sort of following but row two for example says RAID5/6 - what does it mean then by RAID6 on row 2 and RAID 5 on row 4? Your explantion of RAID5 on row 2 makes sense to me and 6 on row 4 but I don't understand the 5/6 bit on either row.
yagmoth555 avatar
cn flag
@GrantCurell A RAID 6 can have two disk failure, so row #2 with only one failure possible and wrote RAID-6 seem a simple error to me.
Grant Curell avatar
mx flag
That doesn’t explain line 4 though - you can’t have a RAID5 that tolerates two failures.
Grant Curell avatar
mx flag
Updated - you’re right - the chart is wrong.
Score:0
mx flag

Short answer: the chart in my OP is wrong.

See this page for a much better explanation.

I am marking @shodanshoc’s answer correct as their explanation is on point. Basically ignore RAID6 in row 2 and RAID5 in row 4.

enter image description here

I sit in a Tesla and translated this thread with Ai:

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.