Caching is quickly becoming the method of choice for storage managers to deploy solid-state disk (SSD) for broad use by a variety of applications. Caching analyzes data accesses to and from the storage system and serves the most active data from memory-based storage. Caching offloads and greatly accelerates reads and provides more headroom for traditional storage I/O paths, which in turn accelerates write performance.

All caching systems, though, are not created equal. While many claim to be non-disruptive, most require some change to the environment. The amount of change that the caching system requires for installation and operation should be a critical decision point in the selection process.

What is transparent caching?

Transparent caching means, quite literally, no changes to the environment, infrastructure, applications or servers. Upon implementation, a transparent cache should immediately begin analyzing and caching the most frequently accessed data to achieve a measurable performance improvement.

The benefits of transparent caching

The first benefit of transparent caching is its ease of implementation. If properly designed, a cache should be capable of deployment in the middle of the production day with no risk of downtime. It should also leverage the high availability design the infrastructure is likely to already have by installing into one segment at a time. With the device plugged in and power applied, analysis should begin and performance should immediately improve. De-installation should be just as straightforward as unplugging a connection at a time and resuming normal operations.

By comparison, caching solutions that require changes to the storage system, network, servers, or applications themselves require a significantly higher level of planning and testing prior to implementation. If for some reason testing is unsuccessful, de-installation can be equally complex and time-consuming.

The primary benefit of easier implementation is a quicker path to return on investment (ROI). This is because transparent caches require less up front design considerations and reduced implementation and failure condition testing, meaning less concern about the installation. As a result, transparent SSD caching systems like those available from Cache IQ, can be installed in under an hour and begin providing performance benefits immediately thereafter. This means that the success or failure of a project can be determined within a few days of implementation, instead of after months of testing with simulated data and environments which are never as effective as the real thing.

Transparent caching systems provide an operational benefit as well. The administrator can add new systems, servers, applications and storage to their infrastructure as required without modifying the transparent cache. This success can expand the use of cache in the environment, thereby widening the benefit of the SSD investment.

Another operational advantage of transparent caching systems is that they require no data management changes. Backup/restore and other operational applications continue without modification. In a non-transparent caching system requiring new mount points, new file systems or new IP addresses, it’s likely the data protection process will need to change as well. Again, this is not only a problem during implementation but it also impacts ongoing operations. As a result, any change in the environment or to the caching system has a ‘halo effect’ that will impact other processes throughout the data center.

Transparent caching also provides a greater degree of resiliency. While most caching systems have high-availability features, which are necessary, these systems have another advantage. They allow data access from the original storage system even after a complete caching system failure. In a non-transparent cache, administrators may be able to find their way back to the data, but the effort required can be considerable. Any changes to the storage systems, network infrastructure, servers or applications will need to be reverted to their original settings.

While the manual steps required to reverse these settings is concern enough, a greater challenge may occur if these changes need to be made years after the initial implementation. As is often the case, the personnel that were around when the project was first completed may no longer be available two or three years later when the system needs to be removed, upgraded or replaced. And, there may no longer be the documentation available to quickly re-point applications to the original storage system.

Of course, high availability on the caching system, transparent or not, is still critical and should factor into the decision. When caching systems are implemented, application design moves from satisfaction in seeing improved performance to the expectation that this performance is now a requirement. Transparent caching systems make the design of high-availability easier because they can leverage so much of the existing network infrastructure for most of their needs.

Transparent caching systems can also provide easier scaling since servers and/or applications don’t need to be manually changed to take advantage of upgraded or increased cache size. A user simply adds another caching appliance and the global pool of available cache storage grows larger. There is no need to manually rebalance which application is being served by which caching appliance.

However, just because a transparent caching system can scale without application changes doesn’t mean that it can’t be fine-tuned. For example, Cache IQ’s caching appliances provide detailed analytics that examine data and allow storage managers to set specific policies and even manually pin certain data sets to the cache area. The difference is that this fine-tuning is done at the caching appliance level, with holistic intelligence provided by its analysis of the storage, network, server and application.

The primary impact of selecting a transparent caching system is that no changes need to be made to the application in any of the above situations. This may be the most realistic way to implement a cache since the application and its owners are the ones furthest removed from the physical storage, but most likely to complain about a performance problem. Transparent caches provide storage managers a way to quickly solve storage performance problems without having to redesign or customize their storage infrastructure or involve personnel from outside the storage team.

Cache IQ is a client of Storage Switzerland 

Previous Entry: “Scaling vs. Retention in Backup Systems”