Block level replication environments

X

xxan0xx

Limp Gawd
Joined
May 12, 2006
Messages
153
I'm currently looking for what people are using for data replication in their environments that are replicating on the block level instead of file level such as Steeleye LifeKeeper does.
 
All of our block level copies are handled at work through our StoreAge (now LSI) SVM/DPM front ends, or through VolumeCopy feature on IBM DS series arrays on the controller side of things. In a real pinch, we'll do DD copies in linux on a mounted LUN. Neither of the first two are cheap..as in at least $20k to get started, and that's without HBAs, switches, disk, or cables.
 
k1pp3r said:
This mainly,

Does doubletake to block level?
Click to expand...

Yes it does, but from my experiences, DoubleTake cannot handle scaling storage environments that grow past 1TB. I try to steer clear of DoubleTake when at all possible.

StoreAge (now LSI) SVM/DPM front ends
Click to expand...

The SVM/multiMirror/multiView concept looks very appealing to me. How much data have you seen the SVM handle in a single node environment?

I ask this as I'm potentially going to have figure this out for a scaling 10PB SAN. Which is looking like EMC hardware is going to be used.
 
it's more a limitation of how many IOPS you need. Each DPM can handle 2million IOPS, and ofcourse, you'll have 2DPMs and 2 SVMs per fabric. You also have to license each feature, and your managed storage in 10TB chunks. We recently deployed it in a National laboratory project that is 1PB. Their IO requirements were very low, as they were using it to move from existing mixed storage to their new storage array. They only have 2 boxes on each side of the storage to faclitate their needs.

The nice thing about this technology is that the more boxes you add, the more linear scaling you get with performance. We see a lot of sales on this product with customers that have large datacenters, and are looking to standardize their storage, since it will work with any FC attached disk, regardless of vendor. To the National Labs, being able to find a solution for a few hundred thousand K that allows them to do all this data moving is a very cheap price tag.
 
Our issue is that the software that is going to be utilizing the storage is very I/O heavy, and we're having a hard time finding a solution that can handle it and keep up with it and not corrupt data on the replication end. - It looks like investigating the SVM solution is going to my best bet from what I've found so far.
 
It's a really nice solution. I spent 3 days in Irvine with the engineering team getting hands on time with the product. Our LSI partnership level from years past (since they no longer sell directly to VARs, and only do OEM deals) gained me access to those resources. Havign 3 guys that have spent the last 10+years each doing SVM development was a bit overwhelming at first, but very helpful. Plus, I got to get the hell out of Albuquerque for the week. At any rate, even on a small SAN, plan to take about a week to figure out what you want to do, and how to tackle your issues, and another week to deploy it. A realistic procurement to integration and testing schedule would be close to a month, especially given the size of the data you're talking about.
 
Yep I agree on the timeframe. It's going to be a fairly drawnout project I can invision.

If you're curious at all what is going to be doing all the storage writing, is NovaNET-WEB which is a WAN based backup utility (a lot like MozyPro, but you control it end to end)

This is actually the software that the Company I work for develops in house,
 
xxan0xx said:
Yes it does, but from my experiences, DoubleTake cannot handle scaling storage environments that grow past 1TB. I try to steer clear of DoubleTake when at all possible.



The SVM/multiMirror/multiView concept looks very appealing to me. How much data have you seen the SVM handle in a single node environment?

I ask this as I'm potentially going to have figure this out for a scaling 10PB SAN. Which is looking like EMC hardware is going to be used.
Click to expand...

Sorry to bump an old thread but if it's going to be an EMC SAN why not just use the EMC arrays based replication tools?
 
I was looking to see if there was any other options out there that people where having success with, and I wanted a valid rebutle to "We don't want to spend the money on that portion of the EMC configuration... " or whatever the big wigs come back with.
 
xxan0xx said:
I was looking to see if there was any other options out there that people where having success with, and I wanted a valid rebutle to "We don't want to spend the money on that portion of the EMC configuration... " or whatever the big wigs come back with.
Click to expand...

Fair enough, but you know the majority of major financial houses and thousands of others are using the EMC array based software for their replication. It's pretty much what made EMC as a company - Timefinder and SRDF.

You'd be hard pushed to find a third party that comes close.

Then you have the fact that support is easier with just one vendor.
 
You must log in or register to reply here.
Back
Top