Restore entire VM from DD2200

[Davide_S_]

Hi to all,

we have an environment that contains one DELL PowerEdge R310 Server (Backup Server) with Backup Exec 2012 Sp4 installed.

The virtual environment is on VMware vSphere 5.5.

The backups are saved on DataDomain DD2200 and on backup server OpenStorage Plugin 3.0.2 is installed for backing up via OST.

Throughput in backup phase is fine, about 5-6GB/min, but in restore phase (entire VM), the DataDomain restore data about 280MB/min in a local disk on backup server. The Duplicate Job from DD to TAPE is around 3GB/min.

We followed the articole "Tuning Windows Media Server for Performance" (KB000180974) on EMC site.

How can we improve restore perfomance? 4-6MB/s are poor.....

Thanks!!

-Davide S.-

[CraigV]

MIght be because the data/VM is being rehydrated from the DD. 

[Colin_Weaver]

Are you tryingto restore the whole VM or just GRT data (individual files) from inside the VM

Basically the rehydration has to happen whatever BUT the GRT restore also needs staging (which will slow it down again)

As such what you are experiencing (apart from optimizing your network and disk performance as much as possible) could be expected.

EDIT: based on your title I guess you are trying to restore the whole of the VM.

[Colin_Weaver]

Not that this will help directly with your current restore speed issue as you would need to update and posibly run a new full backup, but

1) Backup Exec 2012 support for ESXi stops at 5.5 without any updates, Backup Exec 2014 covers Update 1 and Update 2, and Backup Exec 15 adds in Update 3 support. as such if you are running any of the updates to ESXi 5.5 then you are not officially supported and it may be the support differences actually contribute to the performance issues (as the VDDK versions will be different) Note: you specifically need to open the SCL documents for all 3 BE versions to see that the updates require the newer BE versions

2) Backup Exec 15 FP3 has enhancements to increase performance of backup and restores of VMware VMs (the main part being not backing up the blocks that contain the windows swap files which for some servers can make a considerable difference)

So for long term I would suggest you need to move to a newer version of Backup Exec.

[Davide_S_]

Thank You.

The restore job is for entire VM, non for GRT. I don't understand why the duplication job to tape (that requires data rehydratation too) have a normal speed.

We restore entire VM not to VMware directly but in a Backup Server local storage. This disk does not have performance issue (copying files reaches about 200MB/s).

In this moment we cannot upgrade BE 2012 to any other version.

Any other suggestion?

Thanks!!

-Davide S.-

[Colin_Weaver]

Can you duplicate back from OST to a standard disk target and then restore form the duplicate into VMware?

Note you might have to setup a disk target on a volume with enough space to do this but it could be worth a try.

[Davide_S_]

What do you mean by standard disk?

The disk where I try to restore entire VM (vmx, vmdk...) is a 2TB (this is a RDM on storage). The VMs that I restore are about 40-60GB.

This 3d born because the GRT restore not working for incremental vmware backup on BE 2012.

[Colin_Weaver]

I mean create a traditional backup to disk storage device (NTFS) on the Backup Exec server and use that as the target for a duplicate job. (of course this may need some extra storage on your Backup Exec server to achieve)

[Davide_S_]

UPDATE: we upgraded today to BE 2015 FP3. GRT restore from incremental backup works fine.

Restoring on local backup server local disk an entire vmdk has the same speed... about 300MB/min. This makes me think that it is a "normal" at this point.

Have you ever tried this type of restore? How fast it is on your environment?