Archive for December, 2006
December 29, 2006 at 4:27 pm · Filed under Consolidation, blades, rackmounts, vmware
http://www.itbusinessedge.com/item/?ci=22835
An interesting summary of how HP are seeing companies want to be more efficient with their IT using either vmware, grid or both. Check it out.
Also check out: http://www.internetnews.com/ent-news/article.php/3573731
It’s got a very interesting and relevant point that consolidating your server estate from 400 servers to 200, isn’t the key thing, it’s consolidating to work smarter, to do more with less. Very cool, and good to read. Consolidation and virtualization should be used hand in hand with effective IT strategy, of knowing this is where we want to go, what we want to do and doing it.
Â
December 29, 2006 at 4:21 pm · Filed under blades, vmware
http://www.itbusiness.ca/it/client/en/ComputerCanada/News.asp?id=41577
A recent study at the University of New Hampshire found that 38 per cent of the U.S.’s power supply was being absorbed by data centres. According to IDC, businesses will need an additional 12 million square feet of data centre space by 2009.
An article about IBM’s new enclosure, it’s lower power footprint in comparison to HP’s, I highlight it mainly because of the quote above, and the general comments made about blade power consumption, very interesting. At what point do we stop blade buying due to power?
December 29, 2006 at 2:32 pm · Filed under blades
Was in a meeting with one of our blade deployment and virtualization consultants we had in for the day. The datacenter guy and myself were stating the problems with buying more blades for this particular application to meet their additional capacity (we basically lack datacenter space and cooling is starting to be an issue). His response below: (Anyway my questions/comments are not in quotes.)
We’ve got a problem with hosting more blades, the datacenter capacity is ok (for the moment), but cooling is becoming a problem physically and financially.Â
“Hang on, how long are you keeping these blades?â€
Well they’re depreciated over 3 yrs, typically out the door by four, because by then they’ll want more powerful servers, the early Intel Xeon bladess are getting on, they want dual core/quad core blades now for better performance.
“Why not just turn the cooling down by 1 or 2%â€
Well they’d run hot and they wont last as long, they’re production – we can’t let them overheat and go offline.
“No, no, not, overheat, just think about this. The standard x86 blades are factory rated to operate at up to say 45 degrees C, you’ll be running your data-enter at say 25 right?â€
Yes we try to maintain low 20’s for ideal an operating environment.
“Turn the temperature up by 2 degrees, you’ll see you can either fit more blades in, or reduce your power consumption and operating costs by a significant amount. We approximately one hundred thousand a year in electricity costs.”
You’ve said yourself you wont have the blades/servers in say four years, why be concerned if they will only last 4 years instead of 8?â€
Wont the disks fail more quickly?
“They’re batch servers for grid right, you’ve chosen IDE notebook style drives? They’re going to be the weakest point of the blade anyway, you know yourself that they’re a support issue as you discussed. Since they’re grid, they’re meant to be disposable, replaceable on demand.”
Anyway a very interesting perspective,
Some unique ways of thinking about things, but in reality saying what you might think but might not want to say……
December 29, 2006 at 2:17 pm · Filed under Consolidation
http://www.processor.com/editorial/article.asp?article=articles/P2852/23p52/23p52.asp&guid=
“IDC estimates at least 75% of companies with 500 or more employees currently deploy virtual servers, and 45% of all new servers purchased this year will be virtualized.
The InfoPro’s survey of Fortune 500 and 1000 companies found that 85% of respondents cited virtualization technology as critical.
IDC found Linux is the fastest-growing platform for virtual machine deployments, accounting for 40% of the dollars spent on virtual machine software last year. Windows followed with 28%, and the remainder went to Unix and mainframe environments.
Gartner predicts in 2007 that power and cooling spending will exceed server spending. ”
Very interesting summary from this editorial which discusses vmware, and it’s impact, it’s set to change datacenter planning and will have an interesting impact on the traditional server operating system engineer role, I might ask for ESX certification rather than linux/windows….
December 29, 2006 at 2:13 pm · Filed under Consolidation, vmware
http://chucksblog.typepad.com/chucks_blog/2006/12/grid_and_virtua.html
The question was simple enough: “when should I be thinking grid, and when should I be thinking server virtualization?”.
“I think it makes sense — in the long term — to think of abstracted computing environments, and not virtualization and grid specifically. To do so could result in multiple IT stacks being built, followed by the inevitable reconciliation and reintegration.”
A very interesting perspective on grid and vmware, how they are both relevant ways of providing a way of running, deploying and supporting applications.
What is interesting is that we’re continuing to move away from a typical kind of physical box, defined operating system, application, to a grid/vmware and application, the operating system impact is becoming less significant.
Yes you have an os, but it’s mainly because your application depends on a tool it can interact with to manage the processing cycles, the memory and the storage to run. Once we virtualize the physical, box, the storage, the next move whether its grid or another product, is to virtualize the application, to move to an era where there’s vmware, virtual storage, and application or virtual application.
December 29, 2006 at 1:48 pm · Filed under Consolidation, blades, vmware
Ask any of your IT management teams and they will admit (all be it possibly in a defensive tone) that the server estate at any point in time is 90% or so idle. That means that as a result of limitations with the operating system, application code, and due to the windows where your applications are in use (like 9-5 weekdays) your servers are sitting there running the system idle process. To a business sponsor, that can mean I’m paying for something and not seeing the benefit.
This is where virtualization and (or) grid comes into place.Â
With vmware you’d take those servers running at low utilization and virtualize them onto a vmware session, allocating the required number of processors and memory, putting them on a vmware box with similar operating systems, and configurations to share the processing and reclaim/reuse/recycle the physical asset.Â
With grid you could decide ok, all servers join the ‘generic’ batch grid which any business line can use. A rule is established to scavenge processing power from any server below 45% utilization, the server notifies the grid, grid submits a batch and the server does some grid processing.
We know vmware can do it, we know grid can do it, it’s a matter of agreeing a window, whether this requires more groups for different windows or better configuration of the ESX server, it’s still an interesting test I’d like to see tried.
My final ramble on the subject is about blade farms. I’ll give you a case study example.Â
We manage a blade farm comprising several thousand batch calculation servers, they’re busy, they run at 100% for about 8/12hrs a day, and they are running windows 2003 with datasynapse.
They’re dual processor dual core blades with 8GB ram. I’d love to try the following on two blades:
Install ESX on each blade
Install two windows 2003 virtual machines on each blade and install datasynapse
*Â I have the same amount of processors, but is this a more efficient form of providing processor engines?Â
   2000 versions of the system idle process and explorer can’t be efficient
*Â I would have in effect one copy of the common windows processes
* I wouldn’t have the layered software drivers running etc.Â
Something to think about, but I certainly wonder if in the near future someone thinks about instead of deploying 2000 blades with windows, 2000 ESX servers and 8000 calculation virtual machines. (ESX could even have datasynapse installed on it?)
License wise maybe more, but if the performance/throughput is better, this might not be a problem.
December 29, 2006 at 1:08 pm · Filed under vmware
There’s been a lot in the press of success stories with vmware; there was the water company that found the unique benefits of a virtual infrastructure.
I thought I’d introduce a conversation that I was part of with respects to vmware for production. Before I begin, let’s define what is meant from production. Many of the banks operate in a development/support and production type of infrastructure. That is, my developers write the application, request a box from infrastructure, infrastructure look after this box and assist development/support in making sure it all works. The key being infrastructure say they support it in production, they have to meet agreed SLA’s and service requirements, (in theory anyway).
So the debate surrounded the following, one of the business teams wanted to place their virtual windows IIS server into production.
There were horrified gasps, vmware in production?
*Â But how do we do the performance management?
*Â What about downtime?
*Â The ESX server is production, but I need to occasionally take the box down for service updates, patches etc,
*Â How do we go about capacity?
*Â How to deal with a mixed production/development environment
Needless to say it appears we’re not quite ready to facilitate such a request, not for technical reasons, more comfort and process based ones.
Yes you need to take the esx server down for maintenance, but you need to take the session down for patching anyway at some point, and very few applications don’t have a downtime slot.
The way handle this is to first prove the concept within IT, to the dba’s the middleware, windows/unix and networks teams, so that when it’s time for a full sales pitch, the majority are involved everyone’s ready to play ball.Â
This is where vitalising the infrastructure comes into play, (coupled with getting your vmware performance right), those back office IT systems, that officially are production, but in terms of user experience, if we were without them would not be user affecting.
December 29, 2006 at 12:35 pm · Filed under vmware
http://searchservervirtualization.techtarget.com/tip/0,289483,sid94_gci1231666,00.html
http://feeds.feedburner.com/SearchservervirtualizationServerVirtualizationTechnicalTipsAndExpertAdvice - the xml version allowing you to see all their postings on vmware -v ery good information
A great article detailing how to improve your vmware session performance. One of the key issues technically with vmware was performance and the article goes over some easy ways to resolve this.
The initial vmware implementation we had was initially successful but it slowly started to suffer from poor performance, much of this was issues with configuration and optimization as well as the way that vmware was deployed.Â
Vmware as with anything needs maintained and effectively managed, it needs an owner for strategic - where do we go with it? Is it production ready? As well as day to day, server needs more memory type issues…
December 29, 2006 at 11:02 am · Filed under blades
There’s a perception with IT that the busier the system, the more value for money the business is therefore receiving from it, (until that is there is a delay to the end user). This remains the same with grid, the perception is that if my grid is not near 90% then there is wastage. There will always be an element of processor cycles lost, whether because there isn’t a need for batch processing (holidays/security patching), or that the processing doesn’t need all the engines.
 What’s important is looking at your site, at what you use grid for, if it’s used to offset calculation from the client pc out of office hours for risk reports, and relatively idle during the day, your overall usage report for your grid might be in the low percent range. What’s important is whilst your implementing grid is that it works as expected, gives the performance you need, once the basics are covered, then you can work out the cross charging, the aim for more effective usage.
December 29, 2006 at 11:01 am · Filed under blades
There’s a perception with IT that the busier the system, the more value for money the business is therefore receiving from it, (until that is there is a delay to the end user). This remains the same with grid, the perception is that if my grid is not near 90% then there is wastage. There will always be an element of processor cycles lost, whether because there isn’t a need for batch processing (holidays/security patching), or that the processing doesn’t need all the engines.
 What’s important is looking at your site, at what you use grid for, if it’s used to offset calculation from the client pc out of office hours for risk reports, and relatively idle during the day, your overall usage report for your grid might be in the low percent range. What’s important is whilst your implementing grid is that it works as expected, gives the performance you need, once the basics are covered, then you can work out the cross charging, the aim for more effective usage.
Next entries »
