The New, the Old, the New Again By Alex Bakman published: Tuesday, June 05 2007
You know how people say "everything gets reinvented over and over again." That’s exactly how it feels to me when it comes to VMWARE ESX. Now that we’ve tried distributed computing for about 20 years and have felt the pain of managing 1,000+ plus small servers, the world has decided that the mainframe concept was not so bad after all!
So here we are again shifting from lots of small dedicated servers back to big machines with lots of memory, fast SAN storage and network. Call it what you like, I call it a mainframe built around the X86 architecture.
Having seen this "movie" before, let me share with you some lessons I’ve learned from the past that will await you in this "new" environment. The first thing I want you to do is to focus on a new word – shared. Yes, in this environment your applications or departments get to share the resources with other applications and departments. Just like on a community playground--all the kids in the sandbox had to learn to get along and share it. The same is true of a shared production ESX environment. To facilitate the sharing, the owners of the new sandbox will have to find a way to understand who is using what resources to make sure that all the "kids" are playing well together. No hogging of CPU, memory, network and storage is allowed! For this, the owners of the ESX environment will need a new set of tools that can quickly show if all the kids are playing by the rules and how various applications and departments are consuming resources.
If history is any indication, and I believe it is, what the tool will show is that not everybody is playing by the rules. Some are using way more resources than others. The owners of ESX farms will quickly rediscover the age-old rule of computing which states that programs will promptly fill all available resources. It is only a matter of time. Given the high cost of ESX farms and the cost of purchasing more SAN storage, etc., the owners of ESX farms will quickly realize that IT departments can’t shoulder the cost alone. They will need to charge back business units to recover these costs.
So just like in the mainframe days, the owners of ESX farms will require software to chargeback business users based on their actual consumption of physical resources (CPU, memory, storage, network). I love it when the history repeats itself!
Alex Bakman
27-year computer-industry veteran. Founder and CEO of V-Kernel Corporation, creator of the world’s first Capacity and Chargeback Virtual Appliance. www.vkernel.com.
Everything old is new again: Virtualization in 2005, Part I Published on VSM in 2005
Everything old is new again: Virtualization in 2005, Part II
|
