HPE EVA Storage
1755609 Members
3856 Online
108836 Solutions
New Discussion юеВ

Vraid striping on disks

 
Digex
Regular Advisor

Vraid striping on disks

How does the Vraid actually stripe Vraid1 and Vraid5? Does it spread both of these types of raid accross all disks? So it is safe to say that Vraid5 is faster than Raid1+0 on direct attach via SmartArray since it has more disks in a disk group?

Im just trying to understand if we should get away from Vraid1 and go with Vraid5 when we get our EVA4000.
6 REPLIES 6
Rob Leadbeater
Honored Contributor

Re: Vraid striping on disks

Hi,

Yes. Data for both Vraid1 and Vraid5 is written across all of the disks in a disk group.

Whether VRaid5 will be faster than a locally attached RAID1+0 is difficult to say...

Cheers,

Rob
Ivan Ferreira
Honored Contributor

Re: Vraid striping on disks

>>> So it is safe to say that Vraid5 is faster than Raid1+0 on direct attach via SmartArray since it has more disks in a disk group?

If you consider read access this is true. For write access, probably because of the Storage model and drive type, but you will always will have the parity computation.

>>> Im just trying to understand if we should get away from Vraid1 and go with Vraid5 when we get our EVA4000.

You must run your own performance test. You must also consider the data availability. RAID 0+1 provides more protection as you know.
Por que hacerlo dificil si es posible hacerlo facil? - Why do it the hard way, when you can do it the easy way?
Peter Mattei
Honored Contributor

Re: Vraid striping on disks


RAID01 and RAID5 reads on an EVA do have about the same performance numbers.

The difference shows on the writes due to the RAID5 write penaty!
Write penalty means that for each write to disk you need to read parity data, calculate new parity and then write data + parity.

So to choose your RAID level you need to know your max required write performance and the read/write ratio.

Another fact beside performance is data availability. With RAID5 there is a slightly higher chance for a double disk error within a parity group.
See the EVA best practice whitepaper on:
http://h71028.www7.hp.com/ERC/downloads/4AA0-2787ENW.pdf

Cheers
Peter
I love storage
Digex
Regular Advisor

Re: Vraid striping on disks

Thanks all. Good document also. This is what I got from the document so far.

While Vraid5 provides availability and data protection features sufficient for most high-availability applications, some applications may require the additional availability and data-protection features of Vraid1. Vraid1 configurations can continue operation in failure scenarios where Vraid5 cannot. A statistical model of the EVA shows that, for an equivalent usable capacity, Vraid1 provides over four times the data protection of Vraid51.
Digex
Regular Advisor

Re: Vraid striping on disks

What is the default stripe size? Should we set windows stripe to match it for best performance?
IBaltay
Honored Contributor

Re: Vraid striping on disks

Hi,
see this thread to understand the RSS pls:
http://forums12.itrc.hp.com/service/forums/questionanswer.do?admit=109447627+1217872586099+28353475&threadId=886230
there you can find that the
RAID5 is written in each RSS in the 4D+1P pattern and RAID1 data is written across 2 RSSes in the pattern 4D (primary mirror) in RSS1 and 4D (secondary mirror) in RSS2, etc.
the pain is one part of the reality