Data Centers

IT business

In an effort to speed data centre replication over wide area networks, Cisco Systems (NASDAQ: CSCO) has put its WAN optimization technology for branch offices into its top-of-the-line acceleration appliance.
The company said a new version of the Wide-Area Application Services (WAAS) software that runs its WAE-7341 and WAE-7371 Wide Area Application Engine appliances for data centres now come with a replication acceleration mode.

“In terms of return on investment, you’re now able to minimize the WAN bandwidth that’s used for data replication,” said Marcus Phipps, Cisco’s senior marketing manager for data centre solutions.
Most large organizations have a separate data centre for backup protection, he said, or they might have several data centres that feed into a central centre. If big enough, replication could take days, even over an OC-3 or DS-3 link. The replication acceleration technology can “dramatically shrink” replication times over a WAN, he said. As a result, organizations can schedule more replications in a given period, which will make the data more resilient.

This does sound interesting, the closer we can move to data center virtualization, where my application follows the sun, or even the cheapest place to run is an exciting concept. That I could fail workloads/data centers between sites in line with business or IT requirements could bring dramatic improvements in the ability to do preventative maintenance, be more efficient with our resources, and be more responsive or on demand to the business need.

Building.co.uk

The next time you use Google, think of the vast amounts of power data centres use – 20 times that of a normal office. Stephen Kennett looks at what companies are doing to prune their consumption

Can a data centre ever be green? Although from the outside these nerve centres of today’s digital world might appear like little more than glorified industrial sheds, they in fact consume colossal amounts of energy. Stuffed full of the computer servers that provide the lifeblood of financial institutions and our demand for instant internet access, a typical data centre requires a power load in the range of 20-40MW, 15 to 20 times that of a typical office. But last year worldwide emissions from data centres topped 170 million tonnes of carbon and there’s little sign of it peaking. By 2020 they are predicted to quadruple, outstripping those of the airline industries.

This insatiable thirst for energy is beginning to attract the attention of US and European governments, with steps being taken to introduce legislation to curb consumption. However, some clients and operators are now developing what they claim to be green data centres. One is an Arup Associates-designed data centre in Frankfurt for financial giant Citi, which could be the first to achieve a LEED Platinum rating.

Check out this article about data center power utilization, it’s an interesting read and a topic of conversation that’s ongoing in Europe amongst other regions. Related to data center power is your data center carbon footprint and your operating costs, granted you as an end user might not be concerned about the amount of carbon generated or power used to provide your IT infrastructure, but consider that these costs are billed directly or in-directly to you. Simply adjusting the infrastructure setup, refreshing hardware, looking at your service level agreements or how you procure and provision your IT can reduce your environmental impact, the ability for your IT to deliver and your operating costs.

Health IT Insight

Virtualization investment is on the rise across the healthcare marketplace. National and regional healthcare providers and payers are deploying both server and desktop virtualization solutions to manage their IT operations with more efficiency, security, mobility and transparency. For virtualization vendors, now is the time to get in front of the IT executives driving virtualization deployment in healthcare environments—and “Health IT Insight” is the place to do it.

Virtualization can bring different benefits to different sectors. One might all be about hardware abstraction about, getting the application away from that DL580 in cabinet s19. However, there are growing opportunities for virtualization when we come to the application, to the desktop/the client experience. That I need specific requirements/services online to me - do I need a client pc, with the administration it brings with it, a physical asset with data security issues to consider - or a thin client running virtualized applications on demand?

Could through a simple more efficient organization of the way we provision and supply the IT, mean that the end user only gets access to what they need on an on demand basis? Where I can move capacity around to suit the business requirement - that billing need capacity for the overnight run, but during the day, the ‘client pc’ service needs the capacity for checking in patients, for patient record searches etc. With these different requirements bring commonalities and opportunities in terms of hardware, software and service, the more we discuss our differences and similarities, our requirements, the more we can exchange knowledge, illustrate how the technology was brought in to enable our business - to work around my business, not me work around it.

I was speaking with one of the medical colleges which had recently deployed virtualization with great success reducing their physical hardware, but increasing their ability to provide more services to its end users. It’s all exciting stuff.

I had an interesting discussion with Danny the other day who’s started working for an multinational. He’s been brought on to their windows server team to help with their existing server estate and help application teams keep everything running for the end user. He previously worked in the finance sector.

I’ve removed any references to Danny’s organization and my questions are below.

Describe the server estate.

It’s a mixture of legacy Proliant and early DL servers running a mixture of Windows NT and Windows 2000, for various applications including web and in house ones. We have a spreadsheet with the servernames and the application team responsible, but it’s very much driven by the appliction team.

What’s your core objective

There are two, the on call here is quite harsh due to the number and age of the systems involved, as the estate ages you get the issues with disk space, with hardware failures, disk drives/memory etc. We have to do the first line issues relating to the operating system and escalate to the application team if we need to reboot the server or ’cause an outage’.

What are the key failures you’ve had in the last three weeks?

Hard drives particularly the 4.3GB ones on the old biege style Proliants, the 1600s/2500s which we really need to get rid of. Memory chips and power supplies on some of the newer style proliants, the DL380s, but then they’re around 6 years old, so it’s to be expected.

What actions have you taken to try and improve operations?

Audit the data centers - I know this sounds silly, but understanding first of all, what we have, what’s where and who owns it, us, the business line, or external vendor. I need to know how many DL380’s, how many 6500s we have before we can decide how we move forward.

Establish a more effective management tool for servers - we needed to build a rapor with the application teams, try and synchronize information, and make it easier for on call. So we took our spreadsheet and published it in a wiki type format. If your server isn’t on the list, it’s not supported - getting the application team buy-in to the support/ownership process. To make a server supported, they need to log a call to get it added to the ‘production list’. The existing servers were added, but we emailed everyone down as an owner to provide more information. So by the end we had something like: servername, cabinet, application and DL380 Windows 2000.

Audit the server estate - we needed to know who many servers we had with the different operating systems, and get some basic information, from there we could decide what the priorities were.

Establish priorities - with the audit done, we then established the priorities, clear down the disk space we could on those servers running NT4 and establish a programe to upgrade to Windows 2000/2003. Hardware refresh programe based on platform, so all servers below a specific specification to be replaced with buy-in from the CIO - the end to the 2500’s.

What strikes you as the main support cost?

A mixutre of things. Crucially is age of the infrastructure, through lack of investment, by not upgrading the operating systems and the servers, we get more calls “the system is slow”, the server is down, or even, “how much is it to upgrade the memory in my Proliant 6400r?” the answer is, it’s uneconomical.

What are your key plans over the next year?

The business have kicked off a Windows 2008 project, which will involve upgrading all the application code and refreshing the servers, I’d like to see if we can include virtualization? A lot of the servers are ticking over and don’t necessitate their own physical hardware, the Micorsoft one looks cool and would not require a lot of training, but then I’m sure VMWare would be just as good.

Some interesting comments from Danny. One thing to consider though is that a legacy infrastructure need not necessarily be a high maintenance one. Granted as the servers start getting older, they become more prone to failure, but even things like a re-install of the operating system coupled with firmware upgrades and optimization of the OS to the server can bring real benefits to stability and performance.

That Compaq 2500r with Windows NT4 might be getting a bit unstable with its original NT installation, but re-load the os, load a new driver pack, and tidying up the shares/permissions and you’ll find that the server will be more responsive and reliable than it would do otherwise. Consider investing in the infrastructure you have, optimizing that so that business as usual can be made easier, from there you might find the transition might not be as complicated.

Long term we need to move off the legacy kit in order to work with the new operating systems, the application requirements in terms of memory/storage, as well as from an energy efficiency, service delivery and cost standpoint. Refreshing the hardware alone to one or two standard configurations, can reduce your cost of ownership, your support costs and possibly the energy requirements, switching from 12 different model types to say a DL385 and a DL585, or to blades and VMWare could condense the physical infrastructure need to provide the actual service/workload you need to run.

Your infrastructure is only as good as the amount of time and effort not to mention finance you invest in it - thinking long term is of value, but what about the here and now, what minute changes could be made to improve daily service - clearing down C:\temp on all the Windows servers? Even arranging a reboot, or applying security patches/service packs? Even the new driver pack for better system monitoring?

BBC News

Hundreds of jobs and heat for Prince Charles’ organic business in Caithness have been identified as spin offs from a proposed green energy project.

Tidal power developer Atlantis Resources Corporation confirmed it was considering a site near Castle of Mey for a computer data centre.

The plan - still in the early stages - would see the centre powered by a tidal scheme in the Pentland Firth.
US financial giant Morgan Stanley is a major shareholder in Atlantis.

In an exclusive, Atlantis told BBC Scotland there were potential benefits from the project in terms of employment and Charles-inspired organic and local produce business, Mey Selections.

As we become increasingly interested in the cost of electricity (and cooling) to power the data center, as well as the concept of the corporate carbon footprint, using renewable energy could be ideal as part of a range of technologies and best practice to being more efficient in the data center. Using technologies like fresh air cooling, solar panels on the roof, allowing the data center to run a little hotter, with a lights out configuration could reduce your carbon footprint, and your operational costs. Check it out, an interesting read.