Comprehensive data protection for all workloads
Post Reply
skumflum
Service Provider
Posts: 33
Liked: 1 time
Joined: Jun 13, 2016 6:51 am
Full Name: Søren Emig
Contact:

Build the perfect low-cost repository server

Post by skumflum »

Our use case is as following: About 120 TB front-end data and a circular backup requirement of about 15 restore points in average (some jobs fewer others have more than 15).
We are looking to keep cost down (who is not) and scaling is important since the growth historically has been around 20% each year.
Currently we use a mix of old decommissioned SAN and new HPE MSA systems. Not very sensible and some of the old SAN’s use ridiculous amount of power.
Commercial dedup appliances like HPE storeonce are too expensive since we do not need many restore points.

Two options comes to mind:

1. Expand the MSA systems and retire the old SAN storage – and continue to buy MSA systems
2. Replace all with an entirely new solution

Instead of using general-purpose SAN storage like MSA, I am thinking dedicated servers with direct attached disk.
What would be best in a scale/performance/price balance? My gut tells me that I should stay away from the extremes (100+ DAS).
I could stay in the HPE world and by Apollo 4510 (up to 68 LFF). This single server would be enough to hold my entire backup for now.

Is this suggestion sensible?

Now for file systems…. I had imagined that I would install Windows Server 2016 and use the REFS filesystem. Recently I have heard that it will not handle large backups well. Would you guys play it safe and go for traditional periodic full backup (my suggested system would not have enough IO to handle large merges otherwise).

I had problems with REFS but I formatted it with 4K I am yet to try 64k

What about Windows Server Storage Spaces? Should I go for that or buy an “old fashioned” RAID controller?

Sorry for the long post :)
Andreas Neufert
VP, Product Management
Posts: 6747
Liked: 1408 times
Joined: May 04, 2011 8:36 am
Full Name: Andreas Neufert
Location: Germany
Contact:

Re: Build the perfect low-cost repository server

Post by Andreas Neufert »

Just my 2 cents:

1) Server with DAS. You can use a Apollo or Cisco C3260 (or similar). Price per TB is low and performance is quite good. You can scale out by buying additional servers and use Veeams Scale out Backup Repository, as it will scale out capacity and performance.

2) Repository Server + Storage Systems: NetApp E-Series, Small 3PARs, IBM Storwize, ... can be used and the controllers have likely enough power to scale out with additional servers. Sometime direct attached Servers can be used with SAS HBAs to keep the invest and complexity low.

3) Software defined like Windows Storage Spaces or Storage Spaces direct. You build your own Storage System with Microsoft Software. You need special hardware that is on Microsofts HCL, but the price can be quite low. You need to invest into your own knowledge and partner selection for this much more than at the others.

We have a lot of Gold and Platinum Partners that select storage systems for Veeam backups daily and can help you.
Post Reply

Who is online

Users browsing this forum: Bing [Bot], Google [Bot] and 120 guests