I read with interest Chris Mellor’s recent article on Oracle’s latest tape drive and it got me thinking. In the mid 1990’s I was doing some work for StorageTek (which is where this tape technology comes from) and was lucky to be asked to speak at the StorageTek Forum, that year held in Albuquerque, New Mexico. At the time, STK was about to release the 9840 drive, codenamed “eagle” and we had a number of press conferences and discussions about the new technology. StorageTek were always ahead with their tape technology and their own proprietary format, but the tradeoff was cost – STK was more expensive. Looking at Chris’ article, it made me wonder two things; first how are people calculating the ROI on such monsters and two, whether these kinds of tape drives have a future in the Enterprise data centre.
Return on Investment
Firstly, there’s cost. Imagine purchasing a drive that gets amortised over three years. Irrespective of the absolute cost of the drive, the benefit is in the volume of data that can be read and written from it. Large capacity media certainly helps as it reduces the number of media swaps and the down time that involves. Fast transfer is also essential and the Oracle T10000D is apparently 57.5% faster than the competition. However how many of us have actually seen tape drive being driven at full throughput? What does it take to keep a drive, capable of 756MB/s fed with data?
The reality is that the drive will never run at that speed and to even try to achieve it would require masses of disk cache to keep the media continuously fed and not continually in stop/start mode. That makes it difficult for anything other than the large enterprise customers to use and even then, they may well find it hard to justify over LTO-6.
Archive & Cloud
Now, the world of cloud and in particular archive represents a perfect opportunity for these types of tape drive. I can see these devices being incredibly popular as the back-end storage medium for large cloud ISPs who need to read, write and copy large volumes of data. Effectively these are environments where economy of scale allows them to be fully utilised, reading data from large numbers of clustered server nodes. Perhaps there will be a few other use cases where organisations with large media catalogues can use them too.
What about indexing? How will the data be stored, accessed and retrieved? The industry would have us believe the answer is LTFS. I’m not sure that this is the solution and what we really need is for the cloud providers to develop their own techniques and formats for using sequential media. This could require a significant rethink and perhaps some mainframe smarts from many years ago. In a few weeks’ time I’ll be attending an IT Question Time event in London, then the SpectraLogic analyst event in Colorado, where I’m hoping to have some good discussions and find some answers. I’ll post some more details on both events once I have them.
Comments are always welcome; please indicate if you work for a vendor as it’s only fair. If you have any related links of interest, please feel free to add them as a comment for consideration.
Subscribe to the newsletter! – simply follow this link and enter your basic details (email addresses not shared with any other site).
Copyright (c) 2013 – Brookend Ltd, first published on http://architecting.it, do not reproduce without permission.