Improvment of Backup Infrastructure

[habibalby]

Hi All,
I have a budget to improve the backup infrastructure and I'm open for your suggestion around that. I'm listing a high level detailed of how I'm backing up my environment;

1. Physical Backup Server Dell R720 with 16GB RAM and 2CPUs Quad Cores.
2. Dell TL2024 Tape Library with two LT5 Tape Drivers, 1 tape Drive dedicated from Veeam To Tape Backup and second Tape drive dedicated from Symantec Backup.
3. Same server act as Storage Repository and Proxy.
4. Two separate Proxy (VMs with 8 gig and 8 vCPUs) Hosted on vSphere Environment
5. Same server connects to VNXe Shares using two pNICs (LACP)
6. VNXe CIFS shares presented to Veeam Backup.
7. Backup to Disk Jobs to CIFS Shares. (Friday Full Active Backup, Sun,Mon,Tue,Wed,Thu as Incremental Backup)
8. Keep Restore Points on Disk 5. (Which Results, every Wednesday the old Backup Chain removed by Veeam Job and Thursday last Incremental Backup)
9. Every Thursday after last backup job, Backup Veeam Files to Tape using Veeam that starts to backup the whole Backup Chain.
10. Since the Active Veeam Backup is scheduled as Friday, I run Incremental Backup Files to Tape Job to copy .vbm and latest .vbk file as of Yesterday (Friday).

The above setup works really great, but the only drawback I have is the Backup Window. For some jobs the waiting period is too long to free the tape drive for a the next job. As well as the backup to tape speed, 78MB max with Veeam. Anything around that..?

Since I have a commitment to Dell to spend few bucks as part of other project, they are trying to push high-end-dream backup solution using their own software's which in my opinion it will not work for a number of reason.. One and suffice the argument, is the retention period where is not easily to rip a software and go with another software just easily and I forget about a 3 years retention period..

I would highly appreciate your opinion on this.

Regards,
Hussain

[Shestakov]

Hi Hussain,
In general your infrastructure looks good, but I have a couple of suggestions.

Every Thursday after last backup job, Backup Veeam Files to Tape using Veeam that starts to backup the whole Backup Chain.

Why not to use Backup to Tape instead? It restore point-awear and works faster then file to tape.

The above setup works really great, but the only drawback I have is the Backup Window.

What`s bottleneck statistics for the jobs exceeding the desired backup window?
Btw, backup method you use(incremental with active fulls) is the fastest one.
Thanks!

[mcrape]

With regards to point 7 (CIFS Shares), you might see some improvements there? Typically, speeds to CIFS tends to be slower (and transfers can be more prone to error). Alternatively, you can try and use iSCSI or NFS (via a Linux machine) for the repository. Check out this thread regarding VNX and Veeam - there might be some useful info in there for you.

[habibalby]

With regards to point 7 (CIFS Shares), you might see some improvements there? Typically, speeds to CIFS tends to be slower (and transfers can be more prone to error). Alternatively, you can try and use iSCSI or NFS (via a Linux machine) for the repository. Check out this thread regarding VNX and Veeam - there might be some useful info in there for you.

Thanks for your suggestion, as far as I'm aware of the iSCSI LUN in Windows is supported up to 4 TB in Windows 2008 R2. But defiantly will check it out.
Regards,

[habibalby]

Hi Hussain,
In general your infrastructure looks good, but I have a couple of suggestions.
Thanks!

Do you mean to move the Friday Active Veeam Backup to Tape instead of Disk?

This is the result of one recent backup files to tape... Primary Bottleneck always the Source...
3/27/2016 10:03:16 AM :: Load: Source 77% > Proxy 16% > Network 44% > Target 69%
3/27/2016 10:03:16 AM :: Primary bottleneck: Source


Regards,
Hussain

[habibalby]

With regards to point 7 (CIFS Shares), you might see some improvements there? Typically, speeds to CIFS tends to be slower (and transfers can be more prone to error). Alternatively, you can try and use iSCSI or NFS (via a Linux machine) for the repository. Check out this thread regarding VNX and Veeam - there might be some useful info in there for you.

Its my thread ...

[Shestakov]

My suggestion was to use backup-to-tape job instead of file-to-tape one.
As far as I understood, the primary backup job takes too long. What`s the bottleneck of that job?

[habibalby]

Hi Nikita,
My Primary jobs are all Backup-to-Disk (CIFS Shares)

I have tried backing up directly to tape and the speed was almost same.

Bottleneck on Backup-to-Disk is very which depends on the time that I run the backup and the workload on the infrastructure. Sometime is source which is Dell Compollent SAN.. and sometime the Proxy Servers and that also depends on the number of jobs running in parallels.

Thanks,

[Shestakov]

Backup to tape gives more flexibility for backup and restores, but that`s just my suggestion.
Regarding backup to disk, having Direct SAN mode and proxy as a bottleneck, the improvement you can do to shrink backup window is adding more RAM to the proxies or increasing a number of them.