The first time the term FLAPE appeared on the IT scene was in the summer of 2014. At that time the debate on whether tape is dead – or not so dead – was still very hot and lobbyists from all sides of the table – HDD, SSD, and flash, as well as storage system manufactures, were arguing over the Internet. Then analysts from Wikibon presented new research showing that over a period of 10 years, retaining data is best done when using a combination of flash and tape, allowing the highest functionality and bandwidth while maintaining moderate prices (with tape alone still having the lowest cost possible). Analysts stated that over the 10 year period enterprises can save 300% of their cost of their overall IT budget.
With these results at hand, researcher David Floyer from Wikibon argued that storage system manufacturers should now develop new FLAPE product lines for their future clients, even though he argued that some enterprises use some kind of self-designed flash/tape solutions to address their needs already. He stated that IBM or Oracle would be the most likely candidates who would be able to develop such a combined FLAPE system.
From his point of view there are six common use cases for FLAPE: archiving, backup, long-term retention, scale-out NAS, as a tier in multi-tier storage schemes, and for the cloud.
Now almost 4 years later, the topic begins to gain momentum. With flash getting cheaper, tape space getting bigger per media, and HDD’s seemingly having reached the end of further technology advances, FLAPE seems to be the solution to the ever increasing demand for storage in the age of Big Data.
According to Callin Sanchez, vice president of enterprise storage development at IBM, at an interview at the IBM Interconnect 2017, they currently having a global engineering team consisting of scientists, microcoders, and engineers deployed to figure out on how to better connect the two worlds in one single system. What the team came up with so far, is that they use a management software layer above the both worlds so that they can be better controlled. The reason for that approach is that applications can by this way better connect with the whole storage system.
Even though Sanchez points out in that interview that the main application area of a FLAPE system most likely will be archiving, it seems that IBM wants to change the storage world with a less expensive and reliable solution for long-term use in Big Data analysis. From IBM’s perspective, FLAPE has value for enterprises that are analyzing huge amounts of data, while some of that data is just needed for the (software) solution to be analyzed while others – mostly results – have to be kept for either further analysis or because of regulatory, compliance, or just internal company reasons.
This is an approach that reaches far beyond just archiving, which, as Floyer pointed out, is still much cheaper done solely with tape. FLAPE (or a Hybrid tape approach as it is also sometimes called) is a new way to handle complex workflows in a tiered storage architecture. Additionally, it helps cloud providers to handle zettabyte cloud storage to come.
The solution is to manage both flash and tape under one management software solution. This approach is very much related to the Software-defined Storage (SDS) approach. With this approach data is managed automatically and files will be transferred and placed by the FLAPE storage systems software. This software layer could be based, for example, on the IBM Spectrum Scale File System as well as IBM Spectrum Protect (for the Tivoli TSM tape backup solution). Using, combining, and further developing these currently available solutions would have the great advantages. For example, tape storage hardware could be easily added to such a system as a storage tier and data could be moved from one tier to another inside Spectrum Scale. However no information on the details of the management layer being used has leaked outside IBM.
Other reasons for IBM and Cloud providers like Amazon, Microsoft, Google and others – who are in desperate need to get their own storage costs down – to focus on FLAPE, is that there is not much development in the hard drive technology, while both flash and tape breaking on technology border after the other. Regarding sequential read speeds of big files, normally figures of around 80-160 MB/s with normal 7200 RPM for hard drives are achieved and the best results hardly exceed 250 MB/s. And those figures haven´t changed for several years now.
In contrast to that, scientists have reached another success in tape development just last year: IBM Research announced a record breaking success when they managed to store 330 TB of uncompressed data on a palm-sized tape cartridge. That equals an astonishing 201 gigabits per square inch, which was saved on a Sony prototype of “sputtered magnetic tape.” Current users can already buy the latest go-to-market LTO7 tapes which hold 15 TB of compressed data. The LTO consortium (IBM, HP and Quantum) released a new LTO roadmap extending to LTO 12, which will in the future hold up to 480 TB compressed data per cartridge.
The new LTO7-Generation of tapes, however, achieves speeds of up to 750 MB/s and Flash-based consumer drives like the Samsung 850 Pro or the SanDisk Extreme Pro already achieve speeds of 550 MB/s read and 520 MB/s write – which is more than double the speed of a normal HDD. In addition to that, flash chips are getting cheaper every year, so mixing the two technologies, which have both their own unique advantages, makes a perfect combination in the future for a variety of use cases.
Find more information on FLAPE here: https://www.lto.org/2015/09/flape-flash-and-tape-archival-pairing/
Picture copyright: Kroll Ontrack GmbH, Germany