Input needed: ESX 3.5, Exchange, etc

[Frosty]

Hi all,

I've just started work for a new organisation and am considering moving them towards a Veeam backup/restore solution for their 2 ESX servers. At my last workplace though, we had a physical Exchange server; at my new workplace it is a single, virtual, Exchange 2007 running an ESX 3.5 platform. They currently use a mix of VMware's backup (VCB?) and BackupExec to backup the Exchange server and the Exchange databases respectively.

Ideally I would like to have a single backup approach for our virtual servers. I know from reading here that it is possible to restore individual mailboxes from a Veeam Exchange backup, but am wondering how robust that process is? Can any users with personal experience of this post their thoughts? Specifically interested in whether I should aim to retain BackupExec, or whether I can proceed as hoped an remove it from our infrstructure.

[Gostev]

Hello Stephen, congrats on your new position. Frankly, right now the process is not very robust and requires multiple steps (we have a white paper you can get from our support). Improving item level recoveries is one of the features of our upcoming version 5 (next quarter), this version is going to make the process fast and simple. Thanks!

[Frosty]

Thanks Gostev ... that sounds perfect for our needs ... I think that prompts my next question: If I "buy now" v4.1.1 do I get an automatic upgrade to v5.0 when it is released, or will I need to wait for v5.0 and then buy a license?

[Gostev]

Yes, with Veeam upgrades are always "free" for customers on maintenance.

Important thing to note though, is that the granular restore functionality will only be available in the new Enteprise Edition, which is going to cost extra than v5 Standard Edition (which is what you get as a free upgrade from 4.1.1)

However, good news is that the upgrade fee will be waived on licenses purchased on or before June 18, 2010
http://www.veeam.com/go/free-enterprise-upgrade

[Frosty]

Sounds like a winning plan. I think my last questions are these then:

-- my employer currently has vRanger Pro ... is there a competitive upgrade path, or do we have to buy Veeam outright; and

-- we are a not-for-profit organisation (www.cbm.org.au) ... is there any discounted pricing for charities?

[Vitaliy S.]

Stephen,

Yes, we do have some competitive upgrade offer. Please refer to our sales rep for more information.

Thanks!

[Gostev]

Yes, I believe we have a special pricing for non-profit organizations.

[Frosty]

Pleased to report that we purchased a 4-CPU licence for Veeam and we're looking forward to installing when the upgrade to 5.0 becomes available! In my new role I am still working through the organisation's requirements for Disaster Recovery, however I hope to have that nailed down in the next couple of weeks. Once I have those requirements, I will be making some decisions about our existing VMware infrastructure.

We are currently running ESX v3.5 on two (2) hosts (Dell 2950's) and we have a 1Gb/sec iSCSI SAN in place. The SAN is definitely going to be replaced. Not sure about the Dells, will depend on DR plans.

But I now need to re-engineer the backup regime. We're using a horrible mix of technologies at the moment. I think I want to implement a 2-tiered approach:

Tier #1 ... backup to local disk onsite (we need to backup about 2-3TB daily) ... to be used for most file/system restore situations
Tier #2 ... copy those backups to portable HDDs to take offsite ... these to be used only in the event of a DR situation, or file recovery going back more than 1 week

I was thinking of putting in a new stand-alone backup server with about 10TB of disk storage to accommodate Tier #1, and using a pair of external SATA HDDs for Tier #2. But my concern is that getting the data from the SAN to the backup server will hit a bottleneck, given that it will likely be a 1Gb/sec network connection between the two, so the Veeam backups will have to run over this 'slow' network link.

Can anyone make suggestions as to how I might approach it differently? I've thought of scaling up the new SAN with a load of cheap SATA drives, so that Tier #1 could run totally within the SAN, but then I run into a problem when trying to get the data off the SAN and onto the portable HDDs don't I ???

[TrevorBell]

Frosty,

First thing you need to do is upgrade to V4. then you can take advantage of the new Api which will make backups faster. If you do this then you can make a VM and install Veeam on that, you could then backup your hosts and the .vbk file and subsequent incrementals will reside on your SAN. They you can backup to tape and keep offsite or just setup a batch script to copy *.VBK to removeable drive once a day or week whichever you require and take the data offsite. The possibilites are endless and really simple to achieve

There are alot of posts on here how people do it so have a search around the forum.

I would recommend upgrading to V4 is your first step as mentioned above then you will have more scope moving forward

thanks

trev

[Frosty]

Thanks for the feedback Trev. I do like the idea of keeping everything on the SAN. The only issue I have is how to go about getting the data off the SAN and on to the portable HDDs ... if the data has to traverse the 1Gb/sec LAN then its going to take a long, long time.

That's why I thought I should have a backup server with its own storage. That way I can use ROBOCOPY to mirror the data (changes only) over from the SAN to the backup server, which ought to be a lot more efficient use of network bandwidth. From the backup server, I can physically attach a pair of external SATA drives to take the data offsite.

[TrevorBell]

What is the size of your backup ? i see you say 2-3Tb daily is that the compressed size ?

[Frosty]

2TB-3TB is the raw size. As we've not yet run Veeam, we don't yet have an idea of the compressed/deduplicated (?) size.

The good news is that we have settled on an architecture design and a hardware solution. We've also (we think) solved the issue of how to get the data quickly off on to the external SATA drives. The design currently looks like this:

Production ESX Hosts:
-- ESX1: new Dell R710 server (2 x 2.4GHz quad core Intel CPUs, 48GB RAM, RAID1 local storage only for ESX install)
-- ESX2: new Dell R710 server (2 x 2.4GHz quad core Intel CPUs, 48GB RAM, RAID1 local storage only for ESX install)
-- ESX3: old Dell 2950 server (16GB RAM)

Disaster Recovery ESX Host:
-- ESX-DR: old Dell 2950 server (16GB RAM)

We plan on installing Veeam Backup on a guest O/S on ESX3. That host will have 2 x PCI External SATA cards installed in it. We'll use VMDirectPath to make them directly available to the backup server. Backup will consist of two phases:
-- phase #1 ... Veeam (via vStorage) will backup all guests on ESX1/2/3 to SAN storage
-- phase #2 ... Robocopy will copy that data on to external SATA drives for offsite storage

As for the SAN, as budget is constrained, we are going with:
-- Promise E310s (dual controller) loaded with 12 x 450GB 15,000rpm SAS drives; and
-- Promise J310s (dual controller) loaded with 12 x 2TB 7,200rpm SATA drives

The ESX hosts will have 2 controller cards in them which will talk direct to the E310s controllers, so there will be failover capability. Each E310s controller has 4 ports; each ESX host will use 1, so 3 will be in use, with one free. I'm told throughput of the E310s peaks at about 8Gb/sec for reads and about 5Gb/sec for writes, which is substantially better than our current EMC 1Gb/sec iSCSI solution.

That's the design as things currently stand. It will give me about 2TB of high performance SAS storage, and about 16TB of lower performance SATA storage ... plenty of room for us to grow.

The old EMC SAN will be reorganised and set up to work with the DR ESX host, then stored offsite at Iron Mountain in case its ever needed.

[TrevorBell]

2TB-3TB is the raw size. As we've not yet run Veeam, we don't yet have an idea of the compressed/deduplicated (?) size.

The good news is that we have settled on an architecture design and a hardware solution. We've also (we think) solved the issue of how to get the data quickly off on to the external SATA drives. The design currently looks like this:

Production ESX Hosts:
-- ESX1: new Dell R710 server (2 x 2.4GHz quad core Intel CPUs, 48GB RAM, RAID1 local storage only for ESX install)
-- ESX2: new Dell R710 server (2 x 2.4GHz quad core Intel CPUs, 48GB RAM, RAID1 local storage only for ESX install)
-- ESX3: old Dell 2950 server (16GB RAM)

Disaster Recovery ESX Host:
-- ESX-DR: old Dell 2950 server (16GB RAM)

We plan on installing Veeam Backup on a guest O/S on ESX3. That host will have 2 x PCI External SATA cards installed in it. We'll use VMDirectPath to make them directly available to the backup server. Backup will consist of two phases:
-- phase #1 ... Veeam (via vStorage) will backup all guests on ESX1/2/3 to SAN storage
-- phase #2 ... Robocopy will copy that data on to external SATA drives for offsite storage

As for the SAN, as budget is constrained, we are going with:
-- Promise E310s (dual controller) loaded with 12 x 450GB 15,000rpm SAS drives; and
-- Promise J310s (dual controller) loaded with 12 x 2TB 7,200rpm SATA drives

The ESX hosts will have 2 controller cards in them which will talk direct to the E310s controllers, so there will be failover capability. Each E310s controller has 4 ports; each ESX host will use 1, so 3 will be in use, with one free. I'm told throughput of the E310s peaks at about 8Gb/sec for reads and about 5Gb/sec for writes, which is substantially better than our current EMC 1Gb/sec iSCSI solution.

That's the design as things currently stand. It will give me about 2TB of high performance SAS storage, and about 16TB of lower performance SATA storage ... plenty of room for us to grow.

The old EMC SAN will be reorganised and set up to work with the DR ESX host, then stored offsite at Iron Mountain in case its ever needed.

Setup for the dell r710 seems fine not sure about vmotion capabiltity with the older Pe2950 as we have all same ( 3 x PE 2950 ) but you could alway put the Veeam Backup VM on one of the R710`s so you get vmotion capability...

esx-dr box sounds good we have a an r710 34gig ram and dual quad 2.4ghz so that will work well for daily replications or just as a spare box

For your san i use Dell Equallogic nice and easy to setup we currently have 2, 16x 146 gig sas disks raid 5 giving 1.5TB fast storage and 16 x 1 TB sata disks raid 50 11TB storage....i can run all 34 of our vm`s on the latter array with medium iops, so with your funds available why not speak to your dell rep you will be surprised what deals can be done... if you have xxx to spend they would surely take that than nothing! See what they say you could be surprised. Our Equallogic Ps4000e 16 x 1TB with 3 yr 4 hour gold support was £12,000 retail price was £28,000

[Frosty]

Thanks Trevor,
Yeah, not too worried about whether vMotion will or won't work between the R710s and the 2950. We're only going to be running test VMs on the 2950, plus the Veeam Backup VM. If something ever went wrong with it, we'd just have to wait a few days for a fix. The Test VMs can be cold migrated. The Veeam VM will be using VMDirectPath, so we can't migrate it anyway; it will be host-specific.
I will be chatting to our Dell rep in the next couple of days, so will see what they can or can't do. But with the pricing I currently have, the entire Promise SAN solution (including disks) looks like its going to cost about A$25,000 which is about GBP 15,000. I doubt that they'll be able to offer me anything comparable.

[Frosty]

Well we've made our purchase ... and I must confess I was wrong about Dell ... they came through with an amazing pricing offer on their new MD3200 storage array. We have given them the business, and we will be purchasing:

3 x Dell R710's with 48GB RAM each
1 x MD3200 dual controller storage array with 12 x 450GB 15000rpm SAS drives
1 x MD1200 dual controller storage array with 12 x 2TB 7200rpm nearline SAS drives

We're saying goodbye to iSCSI and it will be interesting to see how it performs. The MD3200 offers 4 x 4 x 6Gb/sec data (i.e. 4 ports per controller, 4 x 6Gb/sec per port). Should last us 4-5 years, given we only have 12-14 guest VMs to run on it! The MD3200 has just been added to VMware's HCL and is supported under ESX 4.1 so that was the final piece of the puzzle for me.