Building a private cloud? Better sort your storage first
Establish a bridgehead, parachute in SSD stacks
The final HP Regcast of our short series on how to build a private cloud covered the usual afterthought of any IT project: storage migration. Much of the conversation dealt with the problems of translating good intentions into action.
The research conducted by Freeform Dynamics at the end of 2012 showed that those IT departments that had innovation most under control tended to build a bridgehead and gradually upgrade the surrounding infrastructure using the bridgehead as a template.
As Freeform’s programme director Tony “Dr Stats” Lock says: “The alternative approach, attempting to replace everything in a big bang, involves too much cost and risk for all but the most well-funded and confident organisations to contemplate.
“Modifying things piece by piece with little strategic direction, or ignoring everything, have been shown to lead to frustration and minimal benefits.”
Lack of direction
No IT department is likely to prepare a PowerPoint deck for the board entitled: “Why ignoring everything is a good idea.” But if we continue to think of storage as an afterthought, private cloud may expose a lack of strategic direction.
Integrating storage in a cloud environment cannot be an afterthought. Private cloud infrastructure has virtualised storage but that is a necessary condition, not a sufficient one. There must also be flexibility so that storage can respond to changes in demand.
Freeform Dynamics reports that Reg readers, even those who are giving the most serious thought to cloud, have been slower in tackling storage virtualisation than server virtualisation.
Risk aversion plays its part, along with awareness that storage performance is one of the most important constraints on application response times.
Also, there is companies’ perception that their storage might be working well enough as it is. Even if utilisation is stuck at around 30 per cent, the budget has been signed off and allocated, and thanks to management silos it is not shared.
That was true in the past but there is no reason to repeat the pattern. Look closely at advocates including storage architecture in any blueprint for a bridgehead.
This means focusing on integrated stacks, integrating the storage with server capabilities inside the stack. It simplifies management, and needs to be done to recognise the performance requirements of specific applications.
If half of Reg readers prefer, as Lock says, to "put stuff in that they know is going to last and that they can expand from", the question is what that stuff might be.
HP favours pre-configured stacks as a first step. Phil Lewis, pre-sales manager in the storage division for EMEA at HP, explains that funding a stack for one application in one department encompassing both servers and storage is best for the bridgehead.
"You don't get a couple of pallets turning up in your data centre that you have to cable," he says. "You can start small.”
One of the most significant innovations has been the solid-state disk
Lewis points out that Gartner has said integration problems might delay a first exploratory step into private cloud by up to a year and half. With a pre-configured stack "you're talking weeks”.
One of the most significant innovations in integrated stacks has been the solid-state disk (SSD).
Many traditional disk-based architectures are unsuitable for cloud deployment because the I/O performance is constrained by the storage performance.
Response might not just be be slow at times but unpredictably slow, frustrating the IT department’s ability to deliver quality of service.
While SSD storage can be more expensive than spinning media by an order of magnitude, using a tier of SSD can improve performance dramatically – but only if it is available to all the apps that require its benefits.
Disk pooling has to be designed efficiently to apply the benefits of shared resources to those applications where storage performance is important. Adding a virtualised tier of high-performance storage means that many more applications benefit from it.
Therefore, build a bridgehead but not a silo. Lock warns: “The key is to build in openness between stacks."
“Much of this comes down to ensuring that suitable management tools and processes are deployed alongside the integrated stacks, as well as encouraging suppliers to build in openness.”
Will vendors resist? Lock thinks not, on the grounds that their historic desire to impose proprietary storage architecture is trumped by the commercial need to be a good citizen.
“Openness will be just as valuable to the vendors as to the enterprises buying integrated solutions. It greatly reduces the fear of vendor lock-in, an issue which has the potential to become a significant inhibitor,” he says. ®
Sponsored: Hyper-scale data management