Friday, April 30, 2010

Virtualization, beyond the hype...


My team has been doing virtualization in the data centre since late 2001 using IBM's Power based technology. We were recognized by IBM and the Common User group with the very first ever Innovation award for Infrastructure Simplification.  I was lucky enough to help IBM in 2005 kick off their Power systems Analyst event in New York at IBM's Palisades facility.  Based upon what we were doing at that time, many analysts praised us privately for the work we accomplished and it was even indicated to us at that time that we were using IT to gain a leadership position and were amongst maybe the top 4% world wide for what we were accomplishing.  In 2005 we took that work and advanced it even further to consolidate systems and storage into a single IBM Power system, at that time running i5 OS, AIX, and Enterprise Linux from Novell as operating systems in their own hardware virtualized space in a single system (about 15 virtual machines using those 3 operating systems). We also had 4 Intel based blades in that system (IBM called them IXS  (Integrated X Series Servers) cards back then) running windows workloads with the disk storage virtualized into the internal disk in our IBM Power System. This internal disk was nicknamed "SAN in a can" because it was internal disk but for and IBM power system it treated it as though it was an internal SAN without all the hassle of fibre cables, switches, and licenses, etc.  It is from this perspective that I share this experience of inside virtualization beyond the hype.

Planning for Tomorrow

For the past 18 months we have been working on the 3rd phase of this journey. Taking into account all the leading edge gains we have made using IBM Power technology, looking at industry best practices in order to maintain a leadership position, looking to address all past concerns that were either marginalized previously in solution design or became new issues during the time since our last system was implemented. And to do so in a cost effective manner that was affordable to our business. That last one is always a challenge in solution design and usually creates trade off's between technology versus cost.

Today's World

Todays solution we are currently implementing consists of the following components

- IBM Power 7 server - inherent hardware based virtualization, 64 bit, supporting 3 operating systems (AIX, i OS, Linux)

- IBM (netapp) 3600 SAN - FC, SAS, & SATA disk arrays

- A pair of 3850 servers running VMware Vsphere 4 ESX, using Vmotion failover to each machine

- A single 3650 server acting as Manager/Console for all servers/tools

- All connected via Fibre 4gbs/8gbs on a Brocade switch fabric

We are using IBM director to manage the entire solution and monitor energy efficiency. This solution provides virtualized capabilities across all server platforms including network and storage. So now to the focus, what does this really mean to the organization having the environment virtualized.

Beyond the Hype

As I don't consider myself as a technical person this is not meant to be a technical debrief, but rather a review of the hype surrounding Virtualization and what it looks like when you begin to achieve the much talked about Panacea.

> Virtualization is Best Practise - while virtualization may be best practise it does not equal doing IT simply in every case. There is truth to the saying that not all solutions are created equal. IBM Power systems treat disk as internal disk but deliver all benefits of a SAN, while a traditional SAN we have learnt about fibre switches (SFP's, Licensing, Zoning), The SAN setup (Aggregates, Volumes, LUNS, etc.) to name but a few.

> Just because you can virtualize multiple operating system's into a single system, does not mean you do not need to know how to work with each of those operating systems. Management tools while better, still are not a panacea to mange an entire solution. So if reducing operating staff for each OS is a goal, you may make some limited headway by combining teams and sharing skills sets and responsibilities, but the real savings comes from reducing physical foot print as that is real savings through consolidation.

> Patching, a nightmare for most os us before, does NOT get any easier with virtualization. What we are finding that the more we leave the pure IBM power solution which truly embodies the art of simply regarding virtualization,and move to "mainstream" Intel based virtualization including storage virtualization through use of a SAN, that the task of patching is even more complicated. We now have to worry about managing multiple patching requirement's:  firmware, os, patch, fixes, Service Packs. This is not the world of an IBM Power solution, but it is today's answer to best of breed approach to data centre virtualization.

So while we have made forward movement with 100% data centre virtualization as an SMB, we are learning that the fine line between doing IT SIMPLY and Best Practise is a very fine one. You may be thinking that hey this guys hates virtualization, far from it, what I do hate is vendors creating complex solutions that take more effort than I know is possible. To finish this up let me share this one tidbit with you, I personally could in 5 minutes in a pure IBM Power environment, carve disk, create server, and boot. Now I need my SAN administrator to carve disk and create LUN's. Give those LUN's to an environment. Have my environment administrator carve me a virtual server space and associate the disk to the server, and then boot. Not a 5 minute task without multiple administrators involved, why has it gotten harder to achieve virtualization, I suggest it because vendors haven't learnt from those who lead, that's my iBM Power server and that's no hype.

Thanks for listening . . .

No comments:

Post a Comment

Twitter Updates

    follow me on Twitter