Clustering for high availability: but don't forget about your backups! (Storage Networking).

Remember how much of a pain the old-fashioned Christmas tree lights were? As soon as one light bulb died in the chain, all of the other lights went out as well. The worst part was that there was no way to tell which light needed to be replaced. You actually had to test each and every one until you found the right bulb. Of course, these days you don't have to worry about that. If a bulb goes out now, the rest of the lights continue to shine. So what exactly does this have to do with computing? It's all about clustering.

Clustering is the use of multiple computers and redundant interconnections to form what appears to be a single, highly available system. When a computer unexpectedly falls or is intentionally taken down, clustering ensures that the processes and services being run switch to another machine, or "failover," in the cluster. This happens without interruption or the need for immediate user intervention. It provides a company with a high availability solution, which means that critical data is available at all times.

Another key benefit of clustering is that it helps make systems more scalable because it allows for incremental growth. The number of users who need to access an application can be increased at any time. Also as more users work on the computer cluster and the servers begin to slow down, additional resources can easily be added to improve performance.

How a Cluster Works

In the clustet; shared resources can be seen from all computers, or nodes. Each node automatically senses if another node in the cluster has failed, and processes running on the failed node continue to run on an operational one. To the user, this failover is usually transparent. Depicted in Figure 1, is a two-node cluster, Node "Left" and Node "Right." They both have C: and D: local disks and shared resources G: and H: The Heartbeat connection allows the two computers to communicate and detects when one fails. if Node "Left" fails, the clustering software will seamlessly transfer all applications and resources to Node "Right." All shared resources are available to both computers.

The Need for an Effective Backup Strategy for Clusters

Although a company is able to achieve a high availability rate with this type of architecture, it's still possible for entire systems to go down. For example, a new virus could enter the system that deletes all of the data. In order to prevent this from happening, a sound backup strategy needs to be developed. Many backup and restore software products are not able to support clustering. In this case, if a failover occurs during a backup job, the entire backup will fail and manual intervention will be needed.

With clustering support, if failover occurs during a backup job, the task will fail, but will automatically retry and succeed on the other physical node. An example of this is shown in Figure 2. The processing is being routed through physical Node "Left." if lightning strikes and Node "Left" becomes unavailable, the backup software should automatically failover and continue the backup job through Node "Right."

San Diego County Credit Union

One company that has been able to implement a clustering environment with a sound backup strategy is San Diego County Credit Union. It is a $1.7 billion organization with 21 locations throughout San Diego County and 20 remote offices and administrative buildings. The IT support staff is responsible for managing and maintaining the entire network and deploying applications, policies, and network and backup solutions. They were recently charged with the task of making the switch to a Windows and NetWare environment, which included clustering technology.

Sean Azhadi, vice president of Networking Services, helped make the move to NetWare 6. "Because it was a brand-new product, the backup solutions that we had in place did not work, so we had to get beta versions of new backup software from different vendors. We tried to make the beta versions work, but they didn't. And in the process of the whole thing, we changed from a centralized storage to a storage area network (SAN). We chose XIOtech as our SAN vendor and were able to immediately implement network clustering services. We were able to put two of our servers on NetWare 6 in a cluster right out of the box with no additional cost to us."

By migrating to a SAN and cluster environment, the company is able to easily handle rapidly increasing data volumes while improving performance and maintaining high availability. Azhadi explained, 'The issue with a SAN environment is that we don't have any more local storage devices. In a traditional way, if you have a drive "Sys" and a "Vol 1," your data is static--it always remains on the same volume. Therefore, you just tell your backup software to go pick up the data that resides on the Sys Volume or Vol. 1. In a cluster environment, this data moves. Any of the clusters could be at any given point with this data and running it. And that's really the whole purpose of the cluster. If you have four or five servers in the cluster at any given point, you could have a users volume, you could have an Apps volume, you could have a mail volume, you could have an NDPS Print volume. Any of these servers could be running any of these services. You no longer have the set devices running on set servers, on a set device s basis."

Azhadi continued, "The issue with the backup came about because as long as I implemented a backup solution on the server that could see the existing volumes, such as the user's volume, the mail volume or apps volume, the backup was fine. But at any given point these volumes could move from one server to the other, and the backup would fail."

Now, the company had to find a new backup and restore solution that was compatible with their new environment. Azhadi added, "We tested Backup Express from Syncsort Incorporated, and it allowed us to centralize our backup management. Off one box we implemented an Overland Data Neo 4000 jukebox and the backup solution worked perfectly with the NetWare clustering. It allowed me to tell my NetWare volumes on the cluster to go out and find the backup device."

In the main site, the company had three NetWare 6 cluster machines in the SAN, and they were looking for LAN-free backups of those NetWare 6 cluster machines. If a NetWare volume should fail, the backup software is now cluster aware and could follow that shared volume. In addition to their main site, the company also has 20 remote sites with NetWare servers, and a local, stand-alone tape drive.

Azhadi was able to back up those machines locally at the remote sites, but manage it centrally with one backup catalog or database at the main site. He added, "We have basically implemented our entire cluster to go on the jukebox. We have implemented many of our servers to write the data set on their own local devices, or sometimes on a remote device or other servers, devices. Then we just tell the backup set to find the device and write it on there."

Clustering technology offers the promise that companies will be able to achieve high availability. In today's e-business environments, this is critical since downtime can translate into millions of dollars in losses. But just having clustering software isn't enough. As San Diego County Credit Union realized, an effective backup strategy is also necessary in order to protect a company's data and system configurations. Once this is accomplished, the full potential of clustering can be achieved.

Craig Abramson is a technical analyst at Syncsort (Woodcliff Lake, N.J.).

www.syncsort.com

www.xiotech.com

www.overland.com