You are browsing the archive for Disaster Recovery.

DPM 2010 launch week @ MMS 2010: Part 4: Virtualization and Data Protection, better together

7:29 am in Uncategorized by mikeresseler

Hey All,

Here’s part 4 of our DPM 2010 launch week overview

For the full set:

DPM 2010 launch week @ MMS 2010: Part 1: Technical Introduction

DPM 2010 launch week @ MMS 2010: Part 2: Protection Applications

DPM 2010 launch week @ MMS 2010: Part 3: Protecting Windows Clients

DPM 2010 launch week @ MMS 2010: Part 4: Virtualization and Data Protection, better together

DPM 2010 launch week @ MMS 2010: Part 5: Disaster recovery and advanced scenarios

DPM 2010 launch week @ MMS 2010: Part 6: Partner announcements

This was the last session of DPM Wednesday, and given by Asim Mitra and Vijay Sen, 2 program managers within Microsoft and responsible for the Virtualization protection within DPM.

On the agenda of this session:

  • Protecting your hyper-v environment
  • Hyper-V Recovery Options
  • Recovering from a disaster
  • Sample Customer Deployments

They started by outlining the top priorities for CIOs in 2010


If you look at the screenshot, you will see that Disaster Recovery / Business Continuance and Server Virtualization comes in 2nd and 3rd.  First one is cost reduction, but I guess that will be so for the next x years :-)

I know that virtualization is more “sexy” then disaster recovery for an IT Pro, but it is of course pretty important to think about backup / disaster recovery whenever you deploy a new solution into your environment.  So why not do this hand in hand?  DPM is designed to protect hyper-v fully and if you have read one of my previous posts you know that it is also capable of backing up vmware virtual machines… if you tweak a bit :-)

So what are the features of DPM 2010 for protecting hyper-v?

  • Host-level backup of Hyper-V on WS 2008 R2
  • Cluster Shared Volumes (CSV) support
  • Seamless protection of Live Migrating VMs
  • Alternate Host Recovery
  • Item Level Recovery

Sounds interesting?  Let’s continue to have a look.

First, they started with a discussion on what to protect.  Should we protect on the host and backup entire VM’s?  Or should we protect inside the guests and take the data?  Now this was the sign for many people in the room to shoot the profile of their environment at the two and ask what the solution would be for their specific case.  Luckily these guys were smart enough (or well trained :-)) to leave all options open.  Why?  I think they share the same opinion as I have.  You never can take this decision without first assessing an environment thoroughly.  There are so many questions you need to ask first before you can decide on what strategy you are going to use.  And even then, in many cases, you will be using both.  I actually had a discussion that evening with a guy that could not believe that at a certain moment you would only choose for the host-level backup for a certain virtual machine.  I actually do think there are cases when this can be done.  Imagine a webserver that is running in production and where the configuration only changes once and a while.  A daily backup of the guest should be enough.  I think a lot of servers that are running and running and don’t contain user data or business data can be protected that way.  I mean, who cares that you lose log files if you are not compliant to something?  If you can recover the server quickly when he’s out, that’s more important then those log files right?  And if they are important, I’m sure that the business then have a solution to archive these logs into an auditing system.  But for the conclusion, this really should be looked at on an individual base and here under are some points that can be used to make that decision

  • Host
    • Protect or recover the whole machine
    • “Bare Metal Recovery” & “Item Level Recovery” of every VM
    • Protect non-Windows servers & LOB applications that don’t have VSS writers
    • No granularity of backup
    • Single DPM license on host, all guests protected
  • Guest
    • Protect or recover data specifically
    • SQL database
    • Exchange
    • SharePoint
    • Files
    • No different than protecting the physical server
    • DPML per Guest

Next topic, how does it work.

As always, you start with an initial replica.  After that, this is never done again.  What happens is the following:

  1. DPM initiates the backup process
  2. Using the VSS framework, an application consistent snapshot is created inside the guest virtual machine
  3. A snapshot of the VM is created on the Host (Important mark, use a hardware VSS writer is you are using a CSV)
  4. Then there is a checksum comparison of the VM snapshot with the DPM Replica
  5. Finally, only the changed blocks are replicated to the DPM Server

Seamless protection of Live Migrating VMs

Yep, you’ve read it correctly.  The backup administrator (I would like to introduce a new title for this job, I would like to call him a Business Continuity and Protection Engineer or Officer… what do you think? :-)) doesn’t need to care where the actual virtual machine resides.  With live migration, pro-tips in SCVMM and virtualization admins you can imagine that the placement of a virtual machine is never fixed.  And you can also image that the virtualization admins won’t like to update the backup guy every time a machine has moved.  With all the automation you can create these days (SCVMM, Opalis, SCOM…) they will probably don’t have a clue either.  DPM will know where the virtual machines is, and protect it from there.  If a machine is moved, then DPM will follow it to its new path.

What about Storage Migration?  Will that work also?  Yep, it will.  Again, DPM will follow the path

All nice and well, you are protected.  But issues happened, and you need to recover.  What are your options?

  • Restore VM back to original host or cluster

Probably the most expected option, system went down, recover to the same location and you’re up and running again.

  • Restore VM to a different host or cluster

A little less expected.  Restore the server to another cluster or individual host.  Now this opens options.  Take a backup of a production server, and restore it to another host for testing purposes.  Just make sure that your test environment doesn’t have the capability to talk to your production environment.  Not sure about the latest patches or service packs?  Restore to another environment, deploy the patches and see if the server starts nicely again.

  • Item Level Recovery (ILR) to file share

And this will become a very much used feature in the future.  Mount the virtual machine, get inside the virtual machine or guest and get the items out of the disk.  This can be extremely handy if you decommissioned a server but forgot to copy one or two files.

What they also discussed is disaster recovery and how to prepare for it, but this will be much more highlighted in the next part.

Finally they showed some real-life implementations.  I’ll add the example of a mid-sized asian hoster in here

CSV Production Environment

This customer has multiple 3-5 node CSV clusters with 30+ VMs on each.

Each CSV has Fiber channel SAN – Dell EqualLogic with H/W provider

Maintained a ratio of 1 CSV per cluster node & VHDs for a VM are co-located in a CSV.

Backup Configuration:

The VM workload mix comprises of almost all Microsoft workloads (Complete Microsoft Shop).

The average size / VM is ~70 GB.

All VMs are backed up at the host level with DPM 2010 on a daily basis.

35% of servers which require which require granular backup and near continuous RPO continue to get backed at guest level using DPM 2010, just as earlier in a physical environment.

Typical DPM 2010 Server Configuration

Number of Processors on DPM Servers: Intel 2×4 cores

Amount of RAM on DPM Server: 8 GB RAM

DPM 2010 protects a fan in of 3 such CSV clusters


Till next post



System Center Data Protection Manager: Disaster Recovery

9:07 am in Uncategorized by mikeresseler

Hey All,

One of the most difficult things there is with backup solutions is Disaster Recovery.  Certainly when you want to do a bare-metal recovery of an old physical machine that has died on you.  Although you can have many possibilities to recover from such a failure, most of them will cause tricks, workarounds, hoping that you can find a server with the same (or almost the same) hardware configuration and lots of time.  Matthijs Vreeken ( has written an interesting post about this  (


Here’s the theory.

You do a P2V of an existing physical server but leave it off afterwards (Assuming Hyper-V here, so that DPM can take a full backup of the VHD and config files).  Every day, you take a backup of the system state and daily data of the physical machine.

At the moment the physical server dies on you, you start the virtual server, restore data and you’re back up and running.


He also mentions some possible issues that you can have during the start of the virtual machine. (Read at his post for more information)


Although this sounds great in theory (I was already on my way to my boss to discuss this ;-)) I still have a few questions about this.


1) What will happen if the physical server has a different patch level?  Will it be necessary to patch both servers @ the same time?

2) You need to keep the system state of the physical server in case you want to go back to the physical server.  But if you already have resetted the password for the computer account of the virtual machine, what will happen then?

3) How will an exchange or SQL server will handle this?  After all, this will be an online restore.

4) How will the DPM agent react on the virtual machine?  What are you going to do when the DPM agent has been upgraded in the meantime.


The idea sounds really great, and I don’t want to throw it away although I still have a few questions about it.  So if you have idea’s remarks or solutions for the questions, don’t hesitate to reply :-).  I’m going to continue to think about it because if we can find a rock-solid solution or process for this, it could become a very popular solution for a disaster recovery plan.  Share your ideas….