Recommendations for backup storage, backup target

[dacky7]

Just debating what to do for our setup as we are approaching near full on our backup host.

At present i have a DL180G6 with 12xTB SATA drives. Whilst backing up from our Hyper-V cluster i`m geting around 40mb/s on a backup job which is getting slower as time goes on. We have made some improvements by running some optimisation on our Netapp filers but something has to change. Also our replication source is set to backup, that in turn is dumped to another netapp filer connected to a hyper-v host (dr site) but the speeds are terrible e.g. 4mb/s! Either way i need to increase the capacity of our backup host. Any recommendations? Do i get an external appliance e.g. a NAS???

Cheers,

[foggy]

Peter, you can find the recommendations on backup storages in this thread above, everything depends on what you can afford. Regarding the speed of backup and replication jobs, please examine the bottleneck stats they provide to define the vector of optimizations.

[dacky7]

No worries,

Source appears to be the highest between 50-99%, i presume this is the backup server storage (local storage) that the replication is reading from.

regards,
pete.

[foggy]

Right, in case of replication from backup, source is the source backup repository. Check if it is busy with some other (backup?) tasks at the moment when replication runs.

[DusanM]

A lot of interesting and useful info here. I would like to hear opinion about my plan.

We have 5 ESXi servers with 25 guest OS. Total amount of data to be backed up is close to 3TB.

My idea is to get only one server.
1 CPU
16 GB RAM
24TB storage
Windows 2012 R2

This server would be the main server, proxy server, database server and repository server. In other words, it will do everything.
I would like to attach a tape drive to this server too.

Question: Is this an OK configuration for what I want to accomplish and is adding a tape drive too much for this server to handle or should I consider a separate server for tape?

Thanks.

[DusanM]

BUMP

Anyone can help me with the question above?

Is it OK to attach a tape drive to the server that stores the entire Veeam configuration?

[Shestakov]

Hello Duscan,

Sorry for the delay in our response. Of course it would be better if you use several devices, e.g. a dedicated server as a proxy, since 1 CPU may not be enough for your infrastructure. You can leverage Virtual Appliance mode to speed up the backup processing.

How often are you going to backup your infrastructure?

[DusanM]

Thanks Shestakov.

We will do daily incremental backups, 3 weeks rotation, full weekly backups, 5 weeks rotation and monthly backups to keep them permanently. This might change, but it is like this for now.

I can add another processor to the server and more resources, that is not a problem.

My main concern is the tape drive. Should I get another server just for the tape drive, or it is OK to put it together with the main server. How resource intensive tape backup processes are?
Tape backup would be performed after disk to disk backup is done. We have a big backup window, from 9:00PM until 6:00AM.

[Shestakov]

Tape server is not that exacting as backup and proxy servers. You can review related system requirements part.
Btw, our system engineers have made a handy tool for rough estimation.

[DusanM]

Thank you Shestakov.

I will put everything on one server. If I find it not to perform well, I can always move proxy server to be a virtual server.

The tool you provided is great. I would be useful to have some text on the side, to explain what the boxes mean. Some of them are self-explanatory, some of them are not.

[Shestakov]

Thanks for the feedback Dusan!
It was originally made by our SEs who were asked by users about their infrastructures size frequently. We are going to enhance and embed it to Veeam ONE in next versions.

[pirx]

Hi forum,

~1200 VM's, 200TB used space inside VM's (FC storage, no array integration), 8h backup time frame, reverse inc. backup, 2-4 proxy/repository server at 2 locations (for primary backup and copy). What storage/disks would you use for this environment? Would large SATA disks be sufficient? Or maybe JOBD's attached to the proxies?

[foggy]

Some recommendations can be found above, however only actual tests will show whether you're able to meet you backup window with this or that hardware/setup (too many variables to predict).

[pirx]

It's a bit hard to test this. Even if it works in a POC with 100 VM's and some storage, this doesn't mean it works for 10x VM's. Some numbers and recommendations would be heplful.

[foggy]

My intention was to say that numbers vary from environment to environment, so wouldn't be too much helpful. While tests using your particular hardware and data sets will. Though this is not to prevent others from sharing their experience here.

[Shestakov]

As was mentioned above, you can have an educated guess with this instrument:

Our system engineers have made a handy tool for rough estimation.

[quinnjovm]

@GarethUK Do you mind sharing what you were using for replication in this scenario?

Are you referring to Veeam replication or another windows server 2012 r2 replication method/software?

Since your original post was awhile ago, just curious if you're sticking with the RAID60 (or moved to RAID10) and/or still recommending the one large pool of all disks in the array? If you have since then added a jbod to the primary server, do you create another large pool using all of the disks in the jbod too?

Thanks!

[mongie]

Particularly interested in large repositories - e.g. 50TB+

We currently use three older Equallogics (25TB) and an R720xd (27TB) as primary storage and a fully populated Dell DR4100 (72TB RAW) in our primary office, with select jobs replicated to a second DR4100 (27TB RAW) at our DR site. The DR replica storage is pretty tight, but performance for primary dedupe appliance is also sometimes an issue.

Not sure whether I want to look at a Veeam integrated dedupe appliance (HP/DataDomain) or go back to plain storage.

What storage are you using for backups?

[mkretzer]

What storage are you using for backups?

Our repo is about 240 TB all in one NTFS filesystem, spanned with storage spaces over several RAID 6 coming from 8 Hiatchi HUS110 boxes. It works very well.
As soon as W2016 is there (some of our backup files are quite large) and per VM backup chains works as it should we plan to remove the storage spaces, use SO repo and add windows dedup.

Markus

[LaithIT]

Hi,

I'm intrestted to know what are you using for long term backup solution.
Ive used StoreOnce both VTL (Backup to tape, Files to tape and Jobs to tape)
Used StoreOnce Catalyst with Copy Backup.

What is your hardware that are used for long term? and what is the dedup ration that you are having.

Regards,
Laith

[ahsankhan]

Hi everyone,

We are looking for a solution to eliminate Veeam restore performance issues. We have the following infrastructure in place for our Veeam Backups.

Backup Server: Veeam B&R v9 as a VM in VMware vSphere
Primary storage for Backups: EMC datadomain 2500
Secondary storage for DR(OffSite): EMC datadomain 2500
LAN Network:10G
We are using EMC Datadomain DDBoost feature for our backups as we have around 100+ jobs that we need to get done within a specific windows, previously we used to use a CIFS share from Datadomain and do backups there, Though Backups and synthetic Backups were slow but restoration was comparatively fast enough. Since we switched to DDBoost storage units to do Backups faster our restoration speed has drastically decreased and we did a Mailbox server restore few days back from a Full Backup and it took 18 hours to do it with a restore speed of 3Mb/s. I have been told by Veeam that data hydration and re-hydration is the cause for this performance hit so now we looking for a solution to keep 7 days of Backup in a Fast primary storage and then archive it using Backup copy job to DataDomains.

As per the Data that is being written in a week with synthetic jobs is around 99048 GB which is compressed by Datadomain to 1255.2 GB with 78.4 Compression factor.

So I need some suggestions and advice regarding this Fast storage solution to keep 7 days of Backup on it, so that in case of an emergency we can quickly recover a VM rather then sitting for the whole day.

[foggy]

so now we looking for a solution to keep 7 days of Backup in a Fast primary storage and then archive it using Backup copy job to DataDomains.

Right step, indeed, inline with our reference architecture. You can find some backup target recommendations in this thread above. Thanks.

[ahsankhan]

HI all,

So which solution can provide good compression along with High restoration speed. we are severely getting hit by the restore speed when we try to restore a 500Gb VM from DataDomain.

[nmdange]

I find a physical Windows server with lots of internal storage to be the best option as a backup repository, which can also double as an off-host proxy server. There are many vendors these days that sell 4U 60-disk SAS JBOD enclosures. Fill that with 4TB+ SAS drives and get all the storage you need! If you choose a vendor that uses standard Seagate or HGST drives, it'll be even cheaper than someone like Dell or HP, who heavily mark up those drives when they resell them.

[bubbagump]

We are looking for a cheap and deep solution to store long term backups. I work in health care and we have ridiculously long retention policies on some things. We're currently using Veeam and Exagrid for short retention times (2 weeks to a month), but Exagrid gets expensive in a hurry. I have considered running ZFS on Linux using compression/dedupe as a Veeam repo, but RAM requirements or SSD requirements for L2ARC seem they may undo any savings.... plus there is the pain of management.

What have folks found relatively cheap, not a pain to manage, and HDD based in this situation? I am looking at 100s of TB of storage and tape is simply too slow to recall from.

[Delo123]

Hi,

we use "cheap" HGST 60 Bay Jbod's connected to physical Veeam server running windows & MS dedupe. I know some (even Veeam) have mixed feeling regarding ms dedupe but we have been using it in production over 4 years now and once we knew how to configure some of the culprits we had no issues since. Also using backup copies and replicas with the repositories as source we feel quite safe the data on the repositories is actually "good"...
Our HGST's are configured with 6TB SAS drives, after Raid we get 270TB usable raw. Just checked on of them. It has 197Tb free and currently holding 263TB of backup data. So without dedupe it would be as good as full... Also warranty on the disks is 5 years if i remember correctly, anyway we love them

[bubbagump]

This is similarish to what I was considering. How did you configure RAID on the JBOD? 10 disks per RAID 6 and then create a giant dynamic disk? Or?

What MS Dedupe problems did you have to work around?

[albertwt]

Our repo is about 240 TB all in one NTFS filesystem, spanned with storage spaces over several RAID 6 coming from 8 Hiatchi HUS110 boxes. It works very well.
As soon as W2016 is there (some of our backup files are quite large) and per VM backup chains works as it should we plan to remove the storage spaces, use SO repo and add windows dedup.

Markus

how is that possible ?

Veeam Backup job itself is already deduplicated ?
Does placing it under Windows Server 2012 R2 or 2016 deduped NTFS volume can reduce the backup size even further ?

[bubbagump]

Veeam has a Dedupe-friendly setting which lets repos perform further deduplication. This is the setting you use with Exagrids, DataDomain, etc. Windows Dedupe is doing the same thing as these appliances.

[albertwt]

Ah I see,

So it is possible and suppported by Veeam Backup to place the VBk and VRB on the on Windows Server 2012 R2 deduped volume for further space reduction ?

Does the restore process will be much slower or the backup process is taking longer ?