Do You Have to Sacrifice High Availability for High Performance?

Posted: 08/18/2011

There is no denying that you can greatly improve the performance of your database or virtual infrastructure by utilizing Fusion-io’s ioDrives.  Another well documented benefit of ioDrives is the reduced cost associated with eliminating the need for a SAN.  However, without a SAN, you might be concerned about limiting yourself in terms of the high availability options.  In the Microsoft world, when we think of high availability, the de facto standard is Windows Server Failover Clustering (WSFC).  WSFC generally requires a SAN; however, in conjunction with SIOS Technology, you can enjoy the high performance of the ioDrives without giving up the high availability associated with the traditional WSFC and SAN architecture.

One of the basic tenets of WSFC is that all application data must reside on shared storage, usually meaning some sort of SAN; putting the clustered application data on an ioDrive is generally not an option. SQL Server clusters require all database files, log files and even tempdb all reside on a shared disk. In a Microsoft virtualization world, if you want to take advantage of Hyper-V High Availability or Live Migration all of your Virtual Hard Disk (VHD) files must reside on shared disks.

So are we saying that in order to achieve the performance and cost benefits associated with ioDrives we must sacrifice availability?  Absolutely not!  One of the features introduced by Microsoft in Windows Server 2008 is enhanced support for multisite clusters, which do not require shared storage.

Multisite clusters are generally designed to support geographically dispersed clusters for disaster recovery purposes. Some of the changes in failover clustering were also included to support single site Microsoft Exchange Server clusters using native Exchange replication technologies such as Continuous Cluster Replication (CCR) or Database Availability Group (DAG). Exchange is the only application to include replication technology that supports multisite clusters. However, you can also build “shared-nothing” clusters for SQL Server, File Servers and Hyper-V Servers; however, you will need a third party host-based replication solution that integrates with Failover Clustering such as SteelEye DataKeeper Cluster Edition (DKCE).

DKCE is a Microsoft certified host based synchronous or asynchronous replication solution that integrates with failover clustering to enable ioDrive users to build “shared nothing” Windows Server Failover Clusters.  Whether all of the application data resides on the ioDrive, or there is a mix of local ioDrive storage and shared storage, DKCE’s storage class cluster resource enables failover clustering where clustering would normally not be possible; including SQL, File Server and Hyper-V clusters.

Join me and Jamila Gunawardena of Fusion-io on August 24 for a webinar, “SQL Server 2008 High Performance and High Availability Through Fusion-io and SIOS.” We'll discuss the benefits of the Fusion-io and SIOS solution in a SQL Server environment, including configuration optimization, performance results, competitive analysis and more.  I might even be able to give you a peak at the soon to be released DKCE v7.3, which is going to blow the lid off performance in a Fusion-io environment. You won’t want to miss this webinar if you plan on deploying SQL Server on ioDrives!