Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I am supporting a company sells things to the public in three stores. At this time, the main site has two servers (file + AD server, and citrix server), and the other locations have one server (all windows 2003). For years they have gotten away without any failover capability, but it's obvious that if any of the four boxes died, they would not be able to deal with it well. The POS software would be gone etc. They do have backups, and raid 1 on all servers.

I would like to get them set up so that if a server died, they could keep transacting their business. They would need the least expensive solution that will get them there.

I've looked briefly at failover clusters and they look pretty complicated to set up. I am not sure that two servers at the main office would do it; as it is now they have two servers (file + AD server, and citrix server) and unless we combined the roles into one server, that'd be four boxes. But a failover cluster is the only way to get what they need, isn't it?

I don't think going to the cloud is an option for them, but I will look into it. If they had some amazon ec2 instances going, they'd be isolated from local machine failures, but still subject to internet outages.

If we do stay with local servers, an option I am considering would be to have the two satellite offices rely on citrix for access to the POS software. Then at least they would not need redundant servers.

This question may be hopelessly broad, but I would welcome any views you may have to offer.

share|improve this question
add comment

2 Answers

They would need the least expensive solution that will get them there.

Their mindset is already wrong and they are setting themselves up for failure. As someone who has worked predominantly with non profits that rely on donations from the public at large with no government support, I know quite well how budgetary restrictions make for some tough decisions. However, banish the word "cluster" or even "failover" from their vocabulary if they want to keeps costs down.

Unless they're okay with the least expensive solution being a few ten thousand USD.

I've looked briefly at failover clusters and they look pretty complicated to set up.

This is dependent upon the applications that are being protected and what level of OS protection is desired. You can cluster applications if they support it. However, to cluster the OS underneath the application is usually more of an ordeal and as such the tools that are able to perform this are quite expensive.

The complication involved is entirely dependent on the methods of creating the cluster.

You might want to look at an application like Marathon or Double-Take (I won't link to them, but you can find them if you search for their name plus the word "cluster"). Those products can cluster the entire server at the OS level and perform seamless (in theory) failover in the case of one failing. Also, be wary of clustering a server that runs Active Directory unless the clustering solution can guarantee that all atomic transactions are committed at both nodes.

The prices are steep, but that's just the cost of doing business in this realm.

If we do stay with local servers, an option I am considering would be to have the two satellite offices rely on citrix for access to the POS software. Then at least they would not need redundant servers.

This seems reasonable. You are now dependent on the internet connection between offices as well as the reliability of the servers, however. It would be less expensive to get double ISP connections at the branch offices and bond them than it would be to cluster to each server in each office.

Consider this as a possibility. Pour the money into clustering a central server that houses the POS virtual terminals and the rest of business logic. Then, if and only if the POS software support it, allow access to the server via a terminal services / citrix connection.

share|improve this answer
    
Thanks for that input Wesley. I doubt I will be able to get them to go for a solution that costs 20k or more, so failover and cluster will not be options. Citrix at the secondary offices would help solve one part of the problem. But what else can I do to help make the servers are the home office more durable? I could get them three identical servers with raid 1, and keep one as a spare, use the spare in case one of the other servers dies. The spare server could be restored using backups. Is that about the best I can do on a thin budget? –  meridian Apr 25 '12 at 16:13
    
@meridian Purchase a repair kit for the server and keep spare parts nearby, or get a 24/7 support contract that includes 4, 6 or 8 hour local repair response time. That seems like the best option with such a skinny budget. –  Wesley Apr 25 '12 at 17:33
add comment

As another halfway measure, I would implement a continuous, block-level backup program (I have used StorageCraft). If your server mainboard or RAID card were to die--or more likely--you got malware, or a Windows update wrecked your server, you could be back up in a matter of minutes. I think StorageCraft has a hot-spare sort of feature, but I never got to use it.

Oh and make sure you test your backups and disaster recovery plans regularly!

share|improve this answer
    
Thanks northben. Is there a 'term' for systems like StorageCraft, so that I can check out alternative? Probably "block level backup program", though StorageCraft does not appear in the first two pages of search results for that phrase. BackupAssist is another like this? It does sound like a good approach to eval. –  meridian Apr 26 '12 at 15:57
    
As I was falling asleep last night, I realized that should have just said "continuous data protection (CDP)". I thought block level backup was another term to describe that as well, but I'm not a backup guru. Basically, such programs capture changes at the block level (surprise!) rather than file level. Backups are very fast (hundreds of GB in minutes because they are making an incremental image compared to the last backup. It's the same concept as Microsoft Volume Shadow Copy. –  northben Apr 26 '12 at 16:18
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.