In a previous post, I invoked Thomas Jefferson in predicting the future of storage.
As I prepare to conclude and summarize this series of Computerworld posts, I'd like to invoke another, eminently-quotable American statesman, Dan Quayle. Our 44th vice president famously described the future by stating:
"The future will be better tomorrow."
This seems especially appropriate given the storage news earlier this month. Researchers from IBM announced that they had successfully stored a single bit of information in an array of 12 atoms. As it currently takes approximately one million atoms to store a bit of information using conventional technology, this announcement points to an exciting potential approach to achieving far denser storage (and, presumably, far less expensive and far greener storage) in the future.
Of course, the new approach (leveraging anti-ferromagnetic materials) presents a number of technical and manufacturing challenges. As one of the IBM researchers said, "Even though I as a scientist would totally dig having a scanning tunneling microscope in every household, I agree it's a very experimental tool."
While the future of storage may be more exciting tomorrow, I don't believe we need to look too far out or invoke exotic new technologies to achieve radical improvements in storage. Instead, we can apply well proven techniques in what I call the 6 "S" model of storage.
As I have argued throughout this blog, much of the mainstream storage industry continues to promote an approach that was appropriate for the IT workloads that were predominant when Quayle was vice-president.
Back then, the design center for IT was the delivery of high reliability, high performance transactional systems. In such environments, the amount of data to be stored was relatively small (a 500 GB database is HUGE), and there was a premium on performance and reliability. Not surprisingly, the IT environment tended to consist of monolithic applications, running on proprietary databases and operating systems on highly-engineered SPARC servers. Unsurprisingly, storage systems were also proprietary, highly engineered and monolithic.
Over the last two decades, computing has changed dramatically. Applications are developed in frameworks and delivered across virtualized servers running Linux on top of large numbers of commodity x86 boxes. But, storage has remained largely proprietary and monolithic.
The trends that have dominated computing must inevitably catch up with storage. There are several reasons for this, including:
The massive growth in unstructured data (over 40% CAGR).
The inexorable move towards virtualization (over 50% of all IT workloads by the end of 2012).
"Big Data"/Map-Reduce workloads where compute, data, and analysis are distributed across large numbers of devices.
Public, private and hybrid clouds, which require that storage and computing be delivered as multi-tenant, multi-purpose clouds.
All of these trends will pose insurmountable challenges to the proprietary, monolithic model of storage. Instead, the new model of storage will start to conform to the 6 S model:
1. Software: Value in storage will move to software and away from highly-engineered hardware.
2. Scale out: Higher levels of performance, capacity, and availability will be delivered by aggregating more commodity servers and disks (the little box approach) rather than deploying larger monolithic systems.
3. Simplicity: Complicated, storage-specific technologies will yield to standard Ethernet, Linux, etc.
4. Sourced openly: Storage will be delivered as open source software on commodity boxes.
5. Superior economics: The combination of all of the above will cause the trend lines for storage pricing to come back in line with Moore's law.
And the sixth "S"? Super Sexy. Embrace all of the above, and the next couple of years will be very exciting years in storage.
Of course, I may be wrong. But, as Dan Quayle also said, "I stand by all my misstatements."
Ben Golub was CEO of Gluster, Inc., which is now the Storage Business Unit of Red Hat. He is on Twitter @golubbe. Source:http://blogs.computerworld.com/19672/dan_quayle_anti_ferromagnetic_bits_and_the_future_of_storage_embracing_the_6_s_model
No comments:
Post a Comment