This post is more than 5 years old
52 Posts
0
1072
Distributed VPLEX Local engines within a data center
I've seen a few references to using VPLEX Metro within a data center to provide better availability guarantees than what VPLEX Local can provide. However, I was thinking why can't you just do this with VPLEX Local by scaling it out to 4 engines and putting the engines in different racks? If I'm within a data center, it would be relatively simple to have these connected via Fibre Channel, which I assume is how they interconnect anyway? Is there some requirement for VPLEX Local that a multi-engine cluster has to be within the same rack or within 10-15 meters of each other like a Symm cabinet?
Thanks,
Dan
ankur.patel
286 Posts
0
March 9th, 2012 07:00
Today we cannot split a multi-engine between racks. But your thoughts make sense, another cross connect benefit is that since you have 2 seperate clusters you get increased scale
Slowrider5
52 Posts
0
March 9th, 2012 07:00
Darn, that would be nice! In my case, I have a scenario where there is well under 10TB of extremely critical data to protect, so the gap in pricing between Local (where first 10TB are included in base price) and Metro is pretty significant. But, if Local engines all have to be in the same rack that doesn't meet the availability requirements.
Thanks for the quick reply!
ankur.patel
286 Posts
0
March 9th, 2012 08:00
You can still do local HA by creating a R1 mirror with volumes from 2 arrays. I do see your point. Understand also the VPLEX engines are the same hardware as our VMAX so that alone should speak to the reliability of it. Of course that doesnt protect you if someone decides to yank the power![](https://prod-care-community-cdn.sprinklr.com/community/687062f5-603c-4f5f-ab9d-31aa7cacb376/ReDistributedVPLEXLocalengines-05d0ef02-6fa8-4b10-96e2-8365acdc69dd-1575803322)
RickBrunner1
117 Posts
0
March 15th, 2012 14:00
As Ankur points out VPLEX does utilize the same hardware engine complex used in our Tier 1 VMAX arrays today so even a VPLEX-Local configuration is highly redundant and reliable. In your use case if this is not satisfactory, then the only supported way to add redundancy and resiliency would be to utilize VPLEX-Metro with each VPLEX cluster residing in a separate rack within the common DC. We do not support splitting a VPLEX-Local into separate racks as this creates serviceability and fault risk that we are nor willing to trade off as VPLEX is being used in front of multiple Tier 1 arrays.