Recommendations for backup storage, backup target

[btmaus]

What would the recommendation be now in 2017 with Veeam 9.5 ... it seems it is Server 2016 & ReFS with local drives/DAS if you need more space. Is this a correct assumption?

[foggy]

Yes, this is a recommendation keeping price/performance/space reduction aspects in mind.

[btmaus]

After reading about ReFS in 2016 and the improved deplication, would using ReFS still give greater space saving benefits over deplucation? Has anyone done a comparison?

[response3]

What kind of overhead are is everyone seeing with ReFS? Coming from a raid world, it's pretty easy to calculate your usable space based on raid type. How do you do this in ReFS?

[Gostev]

If you are talking about overhead from file system structures, then it should not be dramatically different from stuffed NTFS (or be a concern at all for that matter, just like it is not a concern with NTFS). We're for sure talking less than 1% of total volume capacity or so here...

[flamb]

Goodmorning everyone.
I would like your advice to implement an storage unit dedicated to veeam in my infrastructure.

I currently have two HP ML350 server G8 vSphere 5.1 u1.
I have 8 drives in each server. Type SAS 1TB 7200 rpm in raid 5 divided into lun.
Servers with 4 1gbps NIC each, connected to two cisco switches sg200-26.

On two esx hosts run some virtual machines.
SRVDC01 and 02 (one on esx01 and the second on the esx02).
SRVAPP01 (critical server app)
SRVAPP02
SRVAPP03
SRVTS01
SRVBCK01 (Veeam)
vcenter

I attach a screen with the use of resources.

At the moment I'm doing incremental backups on a buffalo TS5200 nas connected as iscsi units to SRVBCK01 virtual server.
Nas is the low-end, with a volume of about 1.78 tb.
I also have a job replication of a critical VM (srvapp01) to the ESX02 every day at 1am and 1pm.

Can recommend me some storage units? One high-end unit and one average unit ?

Thank you all.

[pirx]

Any recommendations for backup storage for remote locations? We have some large NetApp filers in HQ for backup and backup copy but need something cheaper for our smaller remote locations. The size varies from 1-2 VM's with ~1-2TB used disk space to 10-20 VM's with 10+TB. For larger locations we will NetApps. But these small locations often don't even have the budget for a proper ESXi host, so we need a NAS (or maybe a server with local disks) that is not too expensive but can also be integrated in monitoring systems like SCOM, OPC or Fujitsu Server View. Backup over WAN is not possible because the bandwidth is too low (south america etc), so we need a local device. I thought about Fujitsu Celvin NAS (http://www.fujitsu.com/de/products/comp ... s/storage/) which is a relabeled Qnap device AFAIK.

[veremin]

As it's been mentioned here before, a server with a bunch of local disks might work better in this scenario. Keep in mind that you will still need a physical server in each remote location anyway, to play the role of a gateway server for those NAS shares. Thanks!

[Andreas Neufert]

Can recommend me some storage units? One high-end unit and one average unit ?

There are so many options here:

As you already live in the HPE world:
You can use a Standard HPE Server with a good raid controller and local disks as backup target. Or if needed a HPE Apollo Server. Potentially there are other servers/workstations that can be used in a minimal configuration but with good raid controllers.

If you want a Storage System, I would look at some mid range storage solutions as they are pretty reliable: A small HPE storage, NetApp E-Series, IBM/Lenovo Storwize, the list prices are sometimes high but discounts are high as well.

[flamb]

Thanks for the tips!

[BradI]

Building a DIY unit - sanity check?

Ok, I have a slow 1.5MBps connection to our co-location. The link and co-lo are only used for offsite backups. So I rationalize I don't need a super heavy hitting server on the other side. With that said, the Buffalo NAS unit we do have there just isn't cutting it. It's old, just 2 GB RAM.

I am thinking of taking an extra i7 Desktop with 20GB RAM we have and building that into a NAS. Load that up with 4 hard drives with JBOD and use a 60GB SSD for the cache. I know there are drawbacks to JBOD but with an SSD cache, maybe we make this work?

[Gostev]

Sounds good to me, except it is a little bit overpowered for a backup repository in terms of CPU to be honest, you can probably go with the cheapest modern CPU like i3. By the way, what drawbacks of JBOD are you talking about, and how exactly are you planning to deploy that SSD to fix those?

[BradI]

By the way, what drawbacks of JBOD are you talking about, and how exactly are you planning to deploy that SSD to fix those?

I didn't mean it that way exactly. I meant that 4 JBOD are going to be slower than say 12 HD set up in Raid 6 with a nice Raid controller. Maybe the SSD could offset some of that slowdown? I don't know.

I am having the problem currently on my old NAS where it gets about 5% in on GPT creation and then errors out due to an IO error. Support says that error is due to disk speed.

[JC80]

We need to upgrade our backup solution because of data increase. Weekly full is about 8TB (three separate jobs). There's a couple of massive Oracle database servers. Daily incrementals are about 300GB.

We are going to need about 40-45TB free space and I'm thinking of this Synology https://www.synology.com/en-us/products/RS3617xs with 8 x 8TB Seagate Ironwolf Sata disks. Then make a RAID 6 and would have about 48TB of space? Does that sound any good? Price would be about 5000 EUR. Oh, and 10GbE NIC for that Synology too. There is an option to add that.

We already have a separate physical server with Veeam installed. There is no space to attach local disks so that is why I'm thinking this Synology with SATA disks.

[Gostev]

There is no space to attach local disks so that is why I'm thinking this Synology with SATA disks.

Can you attach an external JBOD?

[JC80]

I don't have any free external SAS interfaces available in backup server. Also I'm not very keen on using JBOD as a backup repository. Would prefer some RAID level being used.

[nmdange]

Throw one of these SAS RAID controllers in the server which will let you do RAID on any SAS JBOD you attach to it:
https://www.broadcom.com/products/stora ... -9380-4i4e
https://www.broadcom.com/products/stora ... as-9380-8e

I use this RAID controller in conjunction with Supermicro SAS JBODs with great success.

[Dave23]

Backup Objective
Retain daily backups for 2 weeks
Retain end of week backups for 4 weeks
Retain end of month backups for 6 months
Retain end of year backups for 10 years

Environment
Veeam Backup and Replication 8.0 located at secondary (DR) data center
1 Veeam Proxy at primary (prod) data center
1 Veeam Proxy at secondary (DR) data center
6.6 TB of virtual machine data to backup

Question
I'm have been using Veeam for years but just for replication and would now like to utilize the product for backups to disk storage and move away from tape. I was able to get my hands on a HP StoreOnce virtual appliance on a POC and was trying to use this as my primary backup and backup copy (GFS) repository, only to find out that the backup copy was taking days and I had to stop it. Apparently this will not work due to the rehydration of data from being deduplicated on the device. My vendor recommended I use a server with JBODs as my primary backup for the 2 week's worth of daily backups and use a HP StoreOnce for my backup copy job with long-term retention. Does this sound like the "best practice" scenario for Veeam backups?

I should mention that we would like two copies of all backups so I would want to have two HP StoreOnce applicances replicating from one site to another and would also need some form of replication for my 2 week's worth of daily backups.

I would appreciate any feedback or what others are doing in their data center. My POC on the HP appliance is about to expire and I do not have additional storage to test with at this moment, but need to budget for something quickly for our upcoming fiscal year. Thank you in advance.

[foggy]

My vendor recommended I use a server with JBODs as my primary backup for the 2 week's worth of daily backups and use a HP StoreOnce for my backup copy job with long-term retention. Does this sound like the "best practice" scenario for Veeam backups?

Yes, this is indeed among our best practices, please review the first post in this thread for details.

I'd also recommend upgrading your Veeam B&R instance and utilize Catalyst, if you decide to go with StoreOnce, you will see much better performance.

[mdking]

Hi All,

Firstly sorry if this has been covered a million times already!

I have the opportunity to replace our existing backup solution with a new solution which hopefully will provide better resilience and performance.


Current system:

3 x QNAP 12 Bay NAS Server
All running WD Red Pro 6TB Drives
Raid 6 in each NAS
Using 2 x 1GB Ethernet Trunked ports per server
Mounted onto the Veeam VM using iscsi mapping.

Veeam backup jobs are all Reverse incremental


New solution:
Infortrend EonStor DS1000 24 Bay with Redundant controller + a 16bay expansion JB 2000 for extra storage.
2 x 10Gbe SFP+ ports which will go back directly to the main server room
6TB Enterprise drives.

Backup preferences currently:

24 VMs backed up in total:
Backup job 1
20 of these VMs are backed up every night with Reverse incremental and kept for 14 days (Monthly Active Full taken) Approx total full backup 900GB

Backup job 2
3 of these VMs are backed up every night with Reverse incremental and kept for 60 days (Monthly Active Full taken) Approx total full backup 400GB

Backup job 3
1 of these VMs are backed up every night with Reverse incremental and kept for 120 days (Monthly Active Full taken) Approx total full backup 5.5TB

What would be the best Raid type for these drives? Splitting them up into 3 Raid 10 for each backup above?
Also as the current nas mounting solution is using iscsi on to the veeam server, does this provide any draw backs / performance problems?

Also the large VM is very slow when performing FLR, I presume this is due to A: the Size of the VM and the server its sitting on with the 1GB links?

Thanks in advance!
Matt

[PTide]

Hi and welcome to the community!

What would be the best Raid type for these drives? Splitting them up into 3 Raid 10 for each backup above?

If you do not plan to switch from Reverse Incremental to Forward then I'd recommend you to stay away from RAIDs with parity (5 or 6), RAID 10 is good to go with. Splitting will eliminate all performance benefits of having many spindles, so it is better to stay with a single array. This article is also useful.

Please also check this thread.

Also as the current nas mounting solution is using iscsi on to the veeam server, does this provide any draw backs / performance problems?

iSCSI is fine. Since you plan to use 10 Gbit, the link is not likely to be a bottleneck in your case.

Also the large VM is very slow when performing FLR, I presume this is due to A: the Size of the VM and the server its sitting on with the 1GB links?

Which part of FLR process is the slowest? Browsing for files to restore or the actual restore process?

Thank you.

[mdking]

Thanks for the reply.

Ok so Raid 10 on a single volume...

On the FLR process, it's the mourning of the backup file. Also the veeam VM sometimes bluescreens, but I think this is due to a Antibvirus solution being on this server. I have spoken to support regarding it, they said to remove the AV entirely.
That feels slightly worrying todo!

[PTide]

Hi,

On the FLR process, it's the mourning of the backup file.

Kindly create a separate thread with a detailed description of your infrastructure (repo location, mount server settings etc).

I have spoken to support regarding it, they said to remove the AV entirely.

I think that adding Veeam services to AV exclusions should be enough. Please also post your case ID.

Thank you

[mongie]

Just wondering if anyone has any tips on buying backup repositories...

I need to buy some cheap storage for my remote sites, and a larger amount for my main office (replacing old hardware).

I'm thinking of going with Dell R730xd's with 12x8TB disks for the remote sites ($15K AUD).

Looking for something a little faster for my central office (I also have a 2nd dedupe tier at this site). I'd like to stick to a server if possible... Is storage spaces with tiering any good for storing backups? Are there other options? I'd like to use REFS if possible. Need around 80TB usable at this site. Budget is $45K AUD.

[foggy]

Hi Alex, this thread contains a lot of considerations regarding this similar questions, please review.

[Delo123]

Hi Alex, we had very mixed results with tiering for backups, never had the blocks we needed in the upper tiers Finally we chose decent JBOD's connected to decent servers. Server with a lot of disks slots usually cost "premium", JBOD's, however, are usually cheaper. All you would need additionally is raid controllers (if you want to do hardware raid) or some nifty hba's. For Backups i would alsways recommend Raid 10 (mirroring in storage spaces) since running vm's in a instant recovery settings is quite heavy on the storage. 80TB usable would require 160TB raw which would be 20 8TB drives and thus in a 24 bay enclosure / server this would give you 20 data and 4 spare drives. This could be done in a server of course. I believe the Dell R730 goes upto 16 x 3.5” SAS, SATA, you could then take another 16 bay JBOD. Larger capacity Jbod's would be something like HGST, they have very nice 60 bay boxes which are more affordable as you might imagine. The HGST 4U60 series are solid as a rock and available with lots of different drive options, we have 4 boxes and love them, they even give you 5 years warranty on the entire box including drives which is great!

[mongie]

Thanks Guido,

It looks like it might be hard to source the HGST appliances in Australia, but they look impressive and I will keep them in mind.

Dell have MD1280 (84x3.5" in 5RU) which might be suitable, but it only supports 6Gb SAS. They also have MD1200 (12x3.5" in 2RU) which I've used before. I'm meeting with them today to hear what they think...

[DaveWatkins]

Dell also have the 3060 which is the same sort of thing as the HGST unit, also only 6Gb SAS though

[Delo123]

Yeah, it's a shame one usually has to pay a big premium for the Dell enclosures... Supermicro also has nice Jbods or even Servers with a lot of slots. The Supermicro 6048R-E1CR60N must be the mother of all Veeam Backup servers, 60 disks, 6 slots for NVME and a nice mainboard Let me know (PM) if you need a good contact for HGST, i'm sure they will provide you with hardware if you like, but Avnet in Brisbane should be able to quote i guess

[avelinop01]

We just switched to Veeam from DPM and need to upgrade our Synology Repository.

Any recommendation between the RS3617XS vs the RS3617XS+?
We are upgrading the XS with a 10GB card but the price difference would still be about $1500.
Is the XS+ worth the extra money?