Microsoft Licensing and hardware redundancy

[Flashster]

I am thinking of having total hardware redundancy by having two identical servers(win2k3, SQL Express). Is this the best way of providing totaly hardware redundancy?

The idea being that the second server would only be used if the first one developed a hardware fault. I haven't decided on how to backup the drives yet (which is clearly key to the solution - I'd probably use a disk mirror AND another internal disk for backup of the mirror), but I'd probably intend to restore any backup data to the second server, or perhaps rescue at least one of the mirrored disks from the first server and slot it into the second.

Anyway, the question I have is about Microsofts licensing. Clearly, the second server wouldn't actually be generally in use, and only one copy of Windows server would be actually 'live' at any one time.

If I was to opt for the method of pulling a hard drive from server 1 and putting it in server 2 , then I'd have to sort out activation for the OS as it would stop working as soon as it picked up a different motherboard/processor.

I am doing this the best way? Or is there a way around all this? I guess I need to know if its gonna be the cost of two complete systems, even of one is held in reserver for the other, with all its software installed (if that was the method).


Thanks,

JJ

[TDMWeb]

You would need licences for both servers I think. I'd suggest servers in two different datacentres (for effective redundancy) and a software-based automated failover solution -- we use Simple Failover and it works very well.

As far as SQL Server goes, you could use Log Shipping to keep both servers in sync, but you would need at least Workgroup as Express does not support log shipping. Otherwise you'd need to keep them in sync using your own code, which could be a pain but is possible. Remember too that Express does not support live backups.

In our experience, you are probably more likely / as likely to see datacentre/provider glitches (power, network) as server hardware glitches, so having 2 servers with the same provider could see you no better off.

Backup: daily backup to second drive and regular offserver backup works well. RAID is not backup. For RAID 1 and backup you need 3 drives; but many lower cost servers will only take 2 drives. However, if you are coloing rather than using dedicateds, the time it will take you to get to the dc and swap a drive out may mean that RAID is a necessity.

[JamieBeeston]

Are you simply after Hardware redundancy, or geographical resilience / redundancy?

how much of a switchover time are you prepared to suffer?
how much daily downtime for backup and syncing can you suffer?

and most importantly,

how much cash can you throw at the project if the answers to the above are low/none!

My Managed Serviced Division has installed and manages High Availability load balanced setups for many PLC / Corporates for reasonable budgets (mid - high 5 figures for the Load Balancing side)

Feel free to drop me a mail.

J

[Flashster]

TDMWeb - thanks for allyour info - really helpful.

I am on a very small budget (this is my own venture) so I need to start small. I am hoping to host a few ecommerce web sites so want some sort of hardware/data redundancy plan should the disks or other hardware fail on the 'live' server.

I cannot afford two live servers in two datacentres initially (I don't think). I don't want to rent servers as I want total flexibility.

My SLA will have to cover the datacentre failures and compensate the clients for any downtime, but the hardware/data is down to me and I need to think of ways of replacing the server/disks quickly with up to date data as quickly as possible (within reason).

My thoughts were to have two identical servers as that means two of everything that can fail - not sure if the second one needs to be 'online' I guess that depends on how I am going to replicate the data - and that is my problem.

If SQL express does not allow live backups, then I'm stuck. I'll be using .net 2.0 so mySQL is not available. I cannot afford sql workgroup at the moment (assuming the I am correct in thinking it is over 1k).

If anyone has any thoughts, please let me know....... :-(

Flash

[JamieBeeston]

the other option is just go for a provider with a rapid SLA.

Without wanting to 'advertise' in this section.. there are places who offer < 2 hour Full Server replacement SLA's

Whilst not perfect its certainly alot better for those on a budget.

J

[markcastle]

In order to "afford" MS software you could ensure that your colo provider has SPLA, that way you'll pay monthly for things like MSSQL WG etc.

With regard to redundancy, we provide multi site failover solutions and it normally works like this...

The greater the failover / recovery speed, the higher the cost.

[TDMWeb]

Remember that if you are charging for hosting you will need SPLA licences for Windows, SQL Server etc, not "retail" licences.

[jamster]

My advice would be as follows. I'm not sure if it's the best way, but I think it would work well:

Get 2 x servers with atleast 2 hard drives (perhaps even three if you want to use two for raid) eg
http://www.broadberry.co.uk/quote.ph...ise&spec_id=56)

Install your server on the one hard drive (or two if raid), then use drivesnapshot to backup a nightly image of the whole harddrive of the server on to the leftover drive (http://www.drivesnapshot.de)

Make a Barts PE disk with your raid drivers and the drivesnapshot client on it and in case of disaster you can just unplug the backup drive from the primary server, plug it in to secondary box, boot with barts pe and restore. If your provider provides a remote hands & kvm service you can even recover from a major disaster in the comfort of your office (or sofa)

If you're ultra paranoid you could even pick up a few extra drives in caddys and employ your provider to rotate them for you in a similar way to tape rotation. This would cost a bit extra but in my opinion would be worth it: if something is screwed up on your most recent hard drive image (or that hacker that just broke in helpfully deletes it for you you can revert to a previous one. Just make sure the disks are optimised for safe removal in windows.

(nb. I've not tried this method - we backup using drive snapshot to a SAN - and restore using bart's pe over network. But I can't think why it wouldn't work.).

Weigh this up against the cost of getting a provider to do it for you. Once you've paid for the servers and the rotation you may find it's a cheap to rent a 1U with a fast hardware replacement guarantee and use your provider's SAN to store the drivesnapshot image.

Obviously with your own solution at least you have the peace of mind of not having to rely on your provider to do it for you, but then again, you're reliant on them for everything else, and if you don't trust them you should probably find a new provider.