Siusic Dot Com

I raise this question because after all these years of dealing SQL Server Cluster I feel it may not worth it. SQL Server Cluster is designed to be a high availability solution. The common set up is that two servers participate in a cluster and share the same network name, the same virtual IP address

and the same disk where all the SQL server databases located. The two servers (or the two nodes of a cluster) have to have their own system drive in which the operating system is installed. A SQL server cluster can have active-active setting in which both servers can answer database requests. One node having problem doesn’t affect the other node. Or it can have active-passive setting in which if one node fail SQL server will automatically failover to the other node. In this setting SQL server only runs on the active node which answers all database requests.

So you can see from here the share disk constitute a single point of failure in this architecture. This high availability solution is protecting against the failure of one of the nodes. This architecture make sense only when the reliability of the share disk is a lot high than the two nodes. In fact it make sense only when the reliability of the rest of the system including your network, your power supply and the other software components have lower failure rate than the two cluster nodes. Otherwise giving your system a cluster SQL server will not increase the over system reliability. For instance in an internet business system, the components of the system usually include the web server running a web application which will access a database through a network. The web server, the software application, the network, the database and the power supply are all failure points. If the database is already very reliable compare to rest of the system then increase its reliability even more may not be meaningful.

My experience is that with appropriate hardware a stand alone database can be more reliable than the network equipments (hubs, routers and domain controller) and web server which expose itself to hacker attack. In fact I have experienced failure on the share disk in very SQL server cluster I deal with before, in which case a cluster has no advantage at all. I think it is very common that in SQL server cluster the share disk has a higher failure rate then two nodes even the share disk is in the SAN. SAN can fail as well. I have seen this happen.

My conclusion is that if I were to set up a system for my own business I will not use cluster especially active-passive cluster. It is a waste of resource, increase the cost and difficulty of maintenance and doesn’t increase the over system stability a lot. After all I think most business can tolerate a few minutes downtime if it is necessary to reboot the database.

Related Posts:

How to Connect to SQL Server Running on Non Standard Port

MySQL Replication and MS SQL Log Shipping

Can Not Connect to SQL Server Named Instance Remotely?

Link MySQL Server Into MSSQL Server

Master.mdf is Compressed But Does not Reside in a Read-only Database or Filegroup. The File Must be Decompressed.