Server-based flash has arisen to provide rapid read access to data that is close to the application that needs it.
You can add server flash as an extension to onboard memory or as an add-on to the server that acts as direct-attached storage (DAS). But how does the hottest data get put into those instances of server-based solid-state drives (SSD)?
That’s the job of flash caching software, and in this interview, storage editor Antony Adshead speaks with Chris Evans, an independent consultant with Langton Blue, about the key use cases of flash caching software and how to assess and implement it in your environment.
Adshead: What are the key use cases for flash caching software?
Evans: I’d like to set a little bit of background before we go into use cases, because I think it’s good to understand what flash caching can do for you.
Traditionally we either store our data in memory, in the server – as non-volatile memory, which is expensive but fast – or we write it out to disk; and if we do, that’s a permanent medium we can put it to.
We’ve seen the evolution of tiered storage that has allowed us to make that physical permanent medium more static, and obviously SSD is a great way of doing that, but one of the problems with it is how we use it. We’ll come on to talk about that.
Flash caching software gives us the ability to write data to another tier of storage that’s in the array. So, applications with high I/O demands, such as databases, email servers or even highly active web servers, would see the benefit if they could put their data on faster, more permanent storage in the server. Those are the apps we see flash caching being useful for.
Flash your cache
- EMC upgrades VFCache, shows off 'Project X' flash array
- Dell storage products roadmap includes server flash, primary dedupe
- Server PCIe flash cache trend catching on with storage vendors
- Enterprise flash: Three implementation options for virtualisation
- SanDisk releases FlashSoft flash caching software for vSphere
- NetApp embraces server flash cache, reseller deal with Fusion-io
- Texas Memory Systems, Nevex partner for flash cache management
Adshead: What are the key steps in assessing and implementing flash caching software in my environment?
Evans: So, let’s just think about where we might want to use this, and we’ve mentioned some examples. People might also want to put flash caching within their [virtual server] hypervisor, and obviously there are the main ones, vSphere, Hyper-V, XenServer and KVM that are out there, and those tend to be supported by the software platforms that the vendors offer.
So, the first thing you want to look at is to decide what do I need to accelerate; is it my application, my virtual environment, and in what way would I prefer to do that?
Clearly, you could just put in some SSD, but you’ve got to find a way of accessing it. That’s one of the benefits that flash caching software offers you – it gives you a way of putting data to the SSD in a more granular and effective way than if you just put storage into a server.
So, what am I looking to accelerate – application or VM? That will have a bearing on what product you choose in the marketplace; some are targeted at servers, some at virtual environments.
The other thing to look at is whether you want to tie that software to a piece of hardware that’s out there. A lot of vendors tie their software with their own hardware, and a lot of the time we’re talking here about PCIe devices. The question is whether to go down the route of a combined hardware and software solution or buy the software and use ordinary SSDs?
Then really you need to look at the price and the applicability and ask how is it going to fit for you, is it going to deliver all the features you use in your environment? So, if it’s a virtual environment do you need software that supports, for example vMotion?
You need to check applicability of features. You’ll find that you’ll be able to work out if SSD fits your environment better than, say, onto standard disks, and you’ll get a cost saving there, but really you need to look at the whole marketplace and at features and functionality you expect it to deliver for you.
This was first published in November 2012