TCO issues in disk technologies part 2: ROI.

Last month's article in CTR about Total Cost of Ownership discussed how to determine the costs of a storage solution. The TCO of a storage solution is important to know, as it can be much higher than anticipated; but it can usually be offset by a Return on Investment. ROI is how much profit or cost savings are realized over the life of the solution and can be very aggressive for a storage solution.

Straightforward Replacement

Many storage solutions purchased today are replacing systems that are no longer covered by the initial warranty. Monthly maintenance costs for old storage can easily exceed a lease payment of a new purchase. For example, if the maintenance cost for an old system is $3,000 and it costs $2,500/month for a three-year lease to replace the system with a new system, that $500 difference goes directly to the "profit" column. Sometimes, just replacing older storage with newer, less expensive solutions will decrease the monthly cost to operate and provide an immediate ROI for that application.

Performance: With a replacement storage system, it is almost guaranteed that there will be better performance with the new solution. Storage solutions, a few years ago, were based on 1Gb Fibre Channel. Today's solutions are mostly 2Gb (i.e. double the potential bandwidth--MB/sec). Storage controllers are also becoming faster in processing the number of transactions (i.e., I/Os per second). Most Fibre Channel disk drives are now spinning at 15,000 rpm. This faster performance allows databases to process more data in a shorter period of time. Disk-bound applications will benefit from the faster storage devices. For example, using a storage device purchased today versus a storage device purchased three years ago, streaming video applications deliver the content much faster, allowing for more content to be transferred; an application that generates revenue based on the number of transactions processed (i.e., credit card adjudication systems) will process far more requests per minute with controllers and disks able to deliver more I/Os per second.

"Advanced" Replacement

Many older storage solutions were purchased when "premium features" (snapshot, remote replication, etc.) were new and cost prohibitive to implement. With recent market conditions bringing about aggressive pricing, an outdated "basic" storage system can now be replaced with a more robust, feature-rich solution where ROI can be realized in many areas.

Snapshot

Replacing a "bare bones" storage system with a solution that has snapshot technology can dramatically increase the ROI of the purchase. Since the new storage system is likely to perform twice as fast as the existing solution, implementing snapshots on top of the new storage will not negatively affect performance. Here are a few examples of how the snapshot feature can increase ROI:

Backup: Snapshot technology can be used to essentially eliminate a backup window. There are two methods to performing a backup: cold and hot. A cold backup is when an application is taken off line, which means there's no user access to the data, and the data is backed up. A hot backup is when the application remains online and user access is retained while the backup is performed.

A cold backup is usually the optimal solution for those applications that can tolerate multi-hour downtime to perform the backup. Some applications that used to be backed up cold have now grown so large that the backup cannot be completed during the allotted window. If a cold backup is still desired, implementing a snapshot allows the application to be taken offline, a point-in-time "snapshot" of the data is taken, and within a matter of seconds, the application is brought online. The snapshot can then be mounted back onto the application server, or mounted directly to the backup server, and backed up. Total downtime for the application is the time it takes to stop the application to perform the snapshot and then restart the application.

Since many applications must perform with zero downtime, a hot backup must be used. A hot backup is when the data is quiesced (i.e., Oracle database is put into hot backup mode) and backed up in that state. While the data is quiesced, the application usually experiences a performance penalty.

When first implemented, this lag in performance may have been acceptable. As the application data grows, however, the length of time needed to perform the backup also increases--which means the length of time the data is quiesced is longer. The longer an active application's data is suspended, the longer it takes to reactivate the data, which also impacts performance. If a snapshot is implemented, the backup would consist of suspending access to the data (typically, seconds or a few minutes), taking a snapshot (seconds) and releasing the data (seconds). As with a cold backup, the snapshot can be mounted to the application server or backup server and backed up.

For either type of backup, implementing a snapshot will practically eliminate the backup window for the application. This will allow the application to stay online and running optimally for the maximum amount of time, which allows for more transactions to be processed in a day. If a value can be assigned to a transaction, then this can be used as part of the ROI equation.

Using snapshots to improve the efficiencies of backups can also eliminate or delay the purchase of a more robust tape library. Since snapshots can be backed up at any time without impacting production, an existing library that was incapable of meeting backup windows now has extended life and use. The money saved by not purchasing a new library (and media and tape migration services, etc.) can be significant.

Data Mining: Many OLTP databases also need to be queried or "mined" to extract the useful data within them. Many databases that were originally in production mode during the day and had reports run at night have grown into 24X7 systems. OLTP transactions and large reports (queries) typically have different and conflicting I/O profiles; the performances of both environments usually suffer. Since most reporting systems can use data (a few hours old or a day or more old) snapshots can be used to create a complete copy of the database. Much like with a hot backup, the OLTP database can be placed in hot backup mode, snapped and then released. The snapshot could then be mounted on a dedicated reporting server where queries could run at any time without impacting the performance of the OLTP system. As with backups, snapshots allow the OLTP system to run at peak performance 24X7 and allow the reporting database to also run far faster than it would otherwise. Again, if value can be assigned to the increased performance, then that value can be applied to the ROI.

Development/Debug: Many application programmers benefit greatly by developing software, and debugging issues, against a live copy of a database. Utilizing a snapshot to create a copy of the database (and mount it to a development server) to work on can greatly reduce development timelines. Using "real" data helps guarantee that when a new version of software is released, it will work against the live database. Debugging production problems becomes much more efficient when the problem can be recreated against a full copy of the real data. The use of snapshots will, in many cases, reduce development times and bring applications to market much faster than trying to work with simulated, or old, data.

Remote Replication: Today's enterprise-class storage systems support remote replication. There are two ways to replicate data remotely: synchronously and asynchronously.

Synchronous replication is when every write to the local storage is also written to the remote storage device before the I/O is acknowledged to the application. With asynchronous replication, an I/O is acknowledged to the application from the local storage device and sent to the remote storage at a later time.

* Synchronous: There is a certain amount of performance overhead associated with synchronous replication. Synchronous replication is primarily used as a disaster recovery (DR) solution (if the production site goes down then the servers and applications at the remote site are brought up and use the replicated copy of the data). Calculating ROI of a synchronous replication solution is difficult since, if a disaster never happens and the replicated copy is never used, then no value is ever realized--reducing the replication solution to "insurance." If there is a disaster, then the ability to be back in production on the replicated copy of data in minutes or hours vs. trying to recover from tape during a period of many hours or days could pay for the solution immediately and perhaps be the reason the company stays in business.

* Asynchronous: Asynchronous replication allows the data at the remote location to fall behind the data at the primary location, but reduces the impact of replication on the performance of the primary application. Asynchronous replication can also reduce the cost of the communication lines between the local and remote sites, which can deliver significant savings. An asynchronous replication solution can be used for DR, but it can also be used for other things that affect ROI. For example, an asynchronous replica can be snapped and used for backups, data mining, reporting, etc. One added advantage is that all these functions are now being performed on a separate storage device from the production system, which eliminates any potential performance impact to the production applications. Also consider that if corporate policy is to store backups offsite from the production data then backing up the data off a replicated, remote copy satisfies this requirement. This process potentially eliminates the cost of a courier, offsite storage facility, extra tape media, etc., and can add up to a substantial cost reduction.

Conclusion

Justifying the purchase of a new storage solution with an aggressive ROI can be very simple. Cost savings can be determined by exploring just a few areas of the solution. Other areas that can also be considered are, among others, higheavailability ("five-nines" uptime), reduced management costs, reduced physical space requirements, reduced maintenance costs (three years is typically included with a new system) and server consolidation. While this article mainly discussed replacing an existing storage device, the content is also relevant for an initial storage purchase.

www.lsilogicstorage.com

Jim McKinstry is senior systems engineer at Engenio Information Technologies, Inc. (Wichita, KS)