It Is Not A ‘Flash’ In The Pan

harish

Regardless of how evolved the human race becomes and how it masters the art of harnessing technology to its advantage there is one aspect that remains unchanged in the business world. And that is the perennial desire of customers to expand their business venture in a cost-effective manner. Not to forget the intolerance to risk in any undertaking that could include getting a vendor on board, hiring the right people or embracing a new technology.

Logically speaking, in the traditional IT world, a growing business means an equally flourishing data center that keeps on adding layers of equipment either to upgrade or meet the increasing data overload.

Today, when enterprises are looking at alternativesto existing storage data management technologies, flash is a front-runner. It just transforms the data center into a super-fast, high-performance IT environment that is lithe, reliable and affordable. In the wake of all this it wouldn’t be presumptuous to say that flash is a disruptive technology that not only required a new way of thinking as to  how flash will be applied today, but about how it will be applied tomorrow.

IDC estimates the all-flash array market will grow to $1.2 billion in revenue by 2015 and is being embraced by Indian companies across sectors. In the era of the SSD, it was all about accelerating specific workloads. As flash matures, new applications/new storage software stacks are being built “ground-up” for solid-state. Thus today, flash storage has value in a number of implementations within and across the storage infrastructure-be that in storage systems themselves, in servers, or in the network.

Should all the goodness of flash lead us to believe that flash is flawless? Think again.

Flash storage is defined by the number of bits (in enhanced MOSFET transistors) that are stored in a cell. Upside: Greater the bits stored, greater is the capacity offered. Downside: Hurts cell endurance as well as flash storage – often referred to as wear leveling.

A common issue that results in wear and tear is the writes. As most applications writes occurin large sized blocks (4KB, 8KB) you can very well imagine the scale of the wear and tear. Given this challenge, we will see traction around innovative solutions, speedy identification/diagnosis of a failure in this space. Additionally, the innovations will be around how to take advantage of faster IO speeds and especially, read.

On the ‘writes’ side of things, technologies like deduplication, compression are applied  before writes are written down to flash storage tohelp  get more capacity out of flash storage and  they also help in avoiding wear leveling.

While in HDDs, the disks were scrubbed (blocks written &verified) to check their health, this is a very damaging exercise In flash storage.  Smart counters and other indicators are used to conduct a health-check, instead.  Once an issue is detected or suspected, drives are replaced. This is where choosing the right RAID levels come handy.

Flash will transform the read-intensive databases.It has taken discussions on IO speed a notch higher from talking about raw power (IOPS) to sustained and fast speed (IO latency measured as sub milliseconds). While, virtualization of desktops and servers give mobility to applications, virtual machines can be hosted on faster/costlier storage or moved to cheaper/longer lasting storage based on needs.

Again, flash lends more punch to the virtualization narrative. Why? Because it offers a different class of speed! Additionally, (from HDD already and within flash storage) there are different tiers of storage like SLC, MLC and TLC to take advantage of. The outcome: innovations will force storage software to come up with innovative caching and tiering solutions.

For sure, Flash is not a mere ‘flash in the pan’ it’s here to stay – for a while, hopefully!