Is clustering, log shipping, or replication the best option for high SQL Server availability?

Printer friendly

Question

I am after a some advice for an installation that I am about to start.

The company I am currently working for has just invested heavily in a SAN. It is located in two sites with high speed connections. The solution is fully fault tolerant and each server is configure with 2 x 2GB fibre cards (HBA) which connect to 2 SAN fabrics (each with 2 fibre switches). There are 2 disk controllers in each of the HSG80's. It's a Compaq SAN which uses StorageWorks software to provide the failover for the multiple paths. As it is in the SAN, I can also perform snapshot backups. 

I have suggested a SQL 2000 cluster to meet the high availability and business resumption (DR) demands that our company must meet. (In fact, I am about to configure a test environment!)

I have had a heated discussion with the development manager who doesn't understand how a cluster works (but he thinks he does!). He cannot understand why clustering is better than replication. The fact that we can do both doesn't help. He can't see that a Cluster offers anything.

I have tried to explain the pros and cons of replication verses clustering, but haven't had much luck.

In your opinion, what is the best configuration? I need to offer high-speed database performance and data protection. I have thought of using log shipping to another disk on the SAN for data protection, but would clustering be better?

Answer

Here are your viable options, along with the pros and cons of each.

1) Clustering

• Offers hardware, OS, and software fault tolerance
• High reliability
• Automatic failover
• Very fast switchover
• No lost data
• Works with most application software
• Does not offer data fault tolerance
• High cost
• Coupled with SAN as shared array, can offer great performance
• Nodes have to be located physically close together

2) SAN Snapshots

• Offers hardware, OS, software, and data fault tolerance (assuming two
  servers, primary and hot spare), and some form of SAN snapshot/mirroring is used.
• High reliability
• Manual failover
• Potential for some lost data (time between snapshots). Transaction consistency is not guaranteed.
• Switchover is often 1+ hours
• High Cost

3) Log Shipping

• Offers hardware, OS, software, and data fault tolerance (assuming two
  servers, primary and hot spare)
• High reliability
• Potential for some lost data (time between log backups). Transaction consistency is not guaranteed.
• Manual failover
• Switchover is often 1+ hours
• Lower Cost

4) Replication

• Designed for data, not for schema modifications, because of this, hard to
  implement for stand-by server
• Replication is unreliable
• Manual failover
• Potential for minimal lost data (time between transactions). Transaction consistency is not guaranteed.
• Switchover is often 1+ hours or more
• Lower Cost

5) Disk Mirroring (SQL Server 2005)

• Offers hardware, OS, SQL Server, and data fault tolerance
• Automatic and manual failover options availables
• Failover can be fast
• Mirrored databases can be at different physical locations
• Does not require any special hardware
• Moderate cost
• Database-, not server-centric
• May not work automatically for all application software

You will have to weight the pros and cons of each and decide which option best meets your needs and budget. This is a question I can't answer directly for you.

Your Name
Email	(Emails will not be displayed on the site or used for promotional purposes)
Comment
Type characters in the image	(case sensitive)