Friday, April 15, 2011

George Crump, Senior Analyst

Overcoming Tape Detractors

Tape has three perception problems that it needs to constantly overcome. The first is error verification. It is hard to know when a tape has developed an error. This problem is being attacked by vendors like Quantum and Spectra Logic but Crossroads has been shipping an appliance that provides error verification for a while now. The Read Verify Appliance technology will be built into StrongBox.

The second perception problem is write efficiency. Tape requires a large data set to be sent to it for maximum performance and to reduce drive failure. You just can’t copy a file to a tape drive and expect any level of performance. Crossroad’s SPHiNX technology utilizes disk cache to buffer enough data so that when tape is written to it can be streamed. This technology will also be built into StrongBox.

The final perception problem is access. Usually you have to go through some backup or archiving software application to pull your data from the tape environment. It would be much easier to access tape like you would any other file system. That is the key capability that StrongBox delivers. By leveraging the LTO LTFS format and some very smart caching technology StrongBox looks like a CIFS or NFS mount point on your network.

When data is copied to the StrongBox share data is stored to disk first and then when the buffer is big enough (about 37.5 Gb) a stripe is written down the tape. This allows for excellent tape speed and quick movement from disk to tape. In addition time limits can be set so that data does not sit on the buffer for too long of a period of time. The resulting tape is LTFS, meaning that you can take the tape to any LTFS environment and read the tape. This also means that any, as long as they follow the standard, LTFS tape can be imported into the StrongBox system as well.

What about Retrieval?

Loading up the device is only half the solution, how about access? As data is moved from the disk to cache the first .5MB of the file is left on disk. When the application or user access the file that .5MB is used to trickle data to the requesting system slowly enough that CIFS or the application does not time out. If the file is under .5MB then of course the whole file is served from disk. If not the user may have to wait the 2-4 minutes for the file to be recovered.

The system also has intelligence to allow for the pre-staging of related data and the timing on how long the entire file stays in the disk cache before only being available from tape. Crossroads has implemented techniques to make this retrieval from tape time as short as possible.It is important to note that the retrieval is directly from tape to application memory. It is not a two stage process that requires a copy to the disk cache and then a copy from the disk cache to the application.

Storage Swiss Take

The StrongBox solution is exactly what LTFS was looking for when it was first rolled out a year ago. The ability to create an archive and then have the resulting tapes be LTFS eliminates vendor lock in. StrongBox takes that a level higher by offering what could be the simplest, easiest to use archiving solution we have seen. It could also replace some backup processes as well. Imagine using your favorite database dump utility or simply RoboCopy to move data to the StrongBox instead of an application. Even disk only products like Veeam’s virtualization backup solution could be used with this system.