Throwing a Defragmentation Grenade

​All SSD-enabled storage devices are fast. Real fast. But are they as fast as advertised? Not likely. Not in a live environment.

But, they're probably fast enough ... or are they?

​While at a conference in Toronto several years ago, I had the opportunity to sit down with renowned storage analyst, Marc Staimer, It was great geek talk’ about emerging storage. I remember him talking about all-flash arrays (AFA) and how they were running north of 100,000 IOpS. Well north. 

 

I told him I don’t know of any of TriAxis' customers that would require even half of that. He smiled knowingly, and advised me to watch that market carefully.

​Years after that, I joined BWG Strategy’s Tech Sector Advisory Board and had the opportunity to be a panelist on many fascinating topics. I began to see very different aspects of my sector from a variety of industry giants and their perspective. I gained insights that I hadn’t had previously and, armed with it, knew I was more valuable to our customers.

​I saw that the hybrid array market was here, essentially removing the discussion about 15K drives and striping for performance. Right on its heels, the AFA market took hold and, as prices dropped, made hybrid arrays almost superfluous. Meanwhile, SSD-enabled hyperconverged infrastructure appliance were replacing their predecessors and eating into the server market. Again, very fast. And so it goes -- a rapidly changing market that gets faster and faster.

​But not as fast as the sales rep said it would be. Certainly not for the Windows servers. Why is this?

​They call it, the “I/O blender’ effect.

​According to searchstorage.com, the I/O blender effect “… occurs when multiple virtual machines (VMs) all send their input/output (I/O) streams to a hypervisor for processing. Under heavy work loads, I/O processes which might otherwise have been relatively sequential, become random.  This increases the read/write activity on disk heads, which in turn, increases latency. This randomization can occur with all virtualization platforms …” This is further exacerbated by the inherent fragmentation that occurs with Windows OSs.

​This means that pretty much anyone that has invested in SSD-enabled storage arrays is experiencing this effect. Because these are also the same organizations that are typically heavily virtualized at a server level.

​Did you know there is a fix for that?

​About a year ago, I was introduced to a company that is synonymous with defrag: Condusiv Technologies. Well, I mean their previous name, Diskeeper Corp., was. If the name rings a bell, don’t be surprised. This is the company that, in 1995, was asked by Microsoft to develop products for its upcoming Windows NT system. They developed the defrag product for Windows NT. Since that time, defragging disks for Wiindows-based system is simply part of the regular maintenance routine. 

​But it’s also part of the problem.

​Especially with SSD-enabled virtualized Windows environments. Condusiv, not long ago, gave Diskeeper a ‘big brother.’ It’s called V-locity and it’s a software product that solves the two big I/O inefficiencies in virtual environments that rob 50% of bandwidth from VM (virtual machine) to storage with I/O characteristics that are much smaller, more fractured, and more random than they need to be.

​Condusiv’s Brian Morin calls this "death by a thousand cuts," penalizing performance of both flash, hybrid and spindle storage systems. V-locity solves this problem with large, clean contiguous writes and reads, so more payload is delivered with every I/O operation. And it further reduces I/O to storage by caching hot reads from idle, available DRAM.

​Removing the I/O inefficiencies will help you get more performance – likely twice the performance – than you are currently getting from that SSD-enabled investment.

​If you’d like to check it out yourself, Condusiv allows you to try it for free.

​V-locity offers other tangible benefits that you’ve probably already deduced. If you can get more performance out of each Windows VM, you don’t have to allocate as many resources. This gives you better density on your host servers, But more about that later.

​But that's a grenade -- here’s the real bomb: a recent discovered SSD vulnerability that a hacker can use to decimate your entire AFA has been discovered. And, a non-traditional defense is available. ​

This we’ll be talking about on September 26th webinar. Click here to sign up to learn more about what this under-reported yet more potentially devastating threat is – and what you can do about it.

Tom Mumford is Co-founder, CEO & CTO for TriAxis, Inc.