Good-bye — and good riddance — to spinning disk

A few weeks back I wrote about the need for widespread, ultra-high-speed broadband Internet and how unforeseen technologies, companies, and entire industries would exist if only we had a critical mass of 100Mbps or gigabit Internet connections. Sadly, many people don’t understand this. It’s the same shortsightedness that led to the thinking that 640K of RAM was enough for anyone or that the world needed only four or five computers.

Along those same lines, it’s becoming clear that within a decade, we may find ourselves in a post-storage world. By this I do not mean a world that lacks any form of persistent storage, but a world devoid of the painfully outdated yet ubiquitous spinning disk.

[ Also on InfoWorld: Let’s take the Internet back from the ISPs | Read about the year’s winning hardware, software, development tools, and cloud services in our slideshow, “InfoWorld’s 2013 Technology of the Year Award winners.” | Get the latest practical data center info and news with Matt Prigge’s Information Overload blog and InfoWorld’s Data Center newsletter. ]

I’ve been saying for years that modern servers should completely dispense with local disk. Imagine a completely redesigned server platform that has no facilities for local disk at all: no cages, no additional fans or cooling apparatus meant to deal with hot spinning drives. The option of booting from a 1.8-inch SSD or an SD card or via PXE or Fibre Channel or iSCSI is far better than relying on the dinosaurs of IT, the local hard drive. The rise of concepts like VMware’s software-defined data center will move IT in that direction due to the simplicity and reduction in cost and administration. As we lose the concept of fixed-purpose servers, we also dispense with the notion of local storage in its entirety.

Here in the first few weeks of 2013, we’ve already seen terabyte SSDs hit the market for under $600. This capacity and price point were unthinkable even a few years ago. The performance and reliability of SSDs continue to increase, year after year, and it’s not hard to envision extremely affordable, blazingly fast, and ultrareliable SSD storage arrays that all but eliminate many of the classic problems presented by spinning disk.

Yes, you can outfit traditional storage arrays with SSDs right now, but the capacities are much smaller than spinning disk, and the cost is prohibitive in many cases. This has led to the development of tiering technologies that make the best use of SSD, SAS, and SATA disks based on workload profiling. But those technologies are end runs around the problem of spinning disk offering large capacities for lower prices, albeit with a significant performance and maintenance reduction.

Rethinking centralized storage is a necessary part of this transition. We continue to use traditional RAID levels with all manner of storage media, including SSDs, because we’re in a transitional phase in which it’s easy to move between those technologies. However, traditional RAID may not be the best idea for the next generation of centralized storage, and certainly not RAID5 or RAID6. RAID10 may make more sense, but as solid-state storage technologies progress, those safeguards may not be necessary anymore, as they’re truly the remnants of the days when our storage had such odd components as electric motors, servos, and platters that rotate at thousands of revolutions per minute.

The whole notion of relying on such technology — technology that’s almost 60 years old — is beginning to seem strange. Soon it will seem as odd as watching movies made only a few years ago and noticing that the cellphones in use appear to be from the Paleolithic era.

For many infrastructures, I think we’ll see storage become more like switches and firewalls. Rather than a bunch of hot-swap disks eating up rack units and expelling a lot of heat, we’ll see a 1U device that provides sufficient storage capacity, along with high performance and high reliability, and functions as a set-and-forget appliance. Oddly enough, this will saddle only the largest infrastructures with the remnants of the spinning disk era — massive data warehousing projects will likely be more inclined to use what will then be extremely cheap and extremely large spinning disk drives to hold exabytes of data in one location. Cloud storage operations may continue to leverage spinning disk to a degree, possibly for archival storage, but will necessarily have massive amounts of solid-state storage on the front line in order to deliver data at expected speeds.

The post-storage era will be enticing for a variety of reasons, but none moreso than the fact that losing data will be a thing of the past: no crashed disks, no lost pictures or projects or reports, no hours of effort suddenly gone. Storage will be so available, cheap, persistent, and indefatigable as to be largely ignored and taken for granted. I, for one, am very much looking forward to that reality.

This story, “Good-bye — and good riddance — to spinning disk,” was originally published at InfoWorld.com. Read more of Paul Venezia’s The Deep End blog at InfoWorld.com. For the latest business technology news, follow InfoWorld.com on Twitter.