Peter's Technology Trumpet : Virtualisation

Microsoft To Mainstream Containerized Data Centers

SmallCountry — Wed, 09 Apr 2008 20:02:57 GMT

Trends of consolidation can be seen in the migration of servers to fewer, larger more automated datacenter locations, and also in consolidation of the units within those datacenters.

With the availability of advanced management tools and virtualisation techniques it's inefficient for engineers to be fiddling around commissioning individual servers, storage and racks. In a consolidated datacenter world, Prefabricated containers containing maintenence-free redundant units of datacenter fabric and processing power are more practical.

Rackable, SUN and others have helped drive container datacenter designs, and this commodity-based approach is now entering the mainstream. Microsoft's annoucement at the Data Center World conference last week was of a substantial 100+MW datacenter facility filled entirely with Microsoft C-Blox container datacenter.

Although this use of containerised datacenters in a Microsoft mega-facility is the first of its kind, the modular approach is also interesting for Scotland, as proposed in an earlier post on the renewable Energy Datacenter.

As James Hamilton, of Microsoft's Live Platform Services team points out when considering where to locate computing power, "multiple smaller datacenters, regionally located, could prove to be a competitive advantage". Scotland's natural advantages around cheap power potential, addressing heat density and geopolitical position should make it worthy considering as a location for Europes datacenters.

Microsoft To Mainstream Containerized Data Centers With C-Blox -- Data Center -- InformationWeek

I Kidaro you not....

SmallCountry — Thu, 13 Mar 2008 13:02:34 GMT

The consolidation of Virtualisation vendors continues with Microsoft's annoucement of the acquisition of Kidaro . This follows Microsoft's acquisition of Calista Technologies, Citrix acquisition of XenSource and VMWare's acquisition of Thinstall.

Kidaro's Managed Workspace toolset will add an interesting capability to the Desktop Optimisation Pack - it delivers a seamless experience for users making use of Virtual PC. Wherever apps require a legacy OS, or have specific needs for isolation, Virtual PC can be a good solution. The Kidaro tools add a unified admin console, optimised delivery of the Virtual PC images, and the app is all delivered into a seamless window to the users desktop. In the words of the Microsoft Virtualisation team

Kidaro's Managed Workspace product allows enterprise data and applicatiosn to run within a "transparent virtual machine wrapper." Kidaro's product is built upon Microsoft Virtual PC, and the wrapper provides "enterprise class" management, deployment and a clean user experience. You can read more on Kidaro's site. With this acquisition, the wrapper becomes Microsoft Virtual PC.

Microsoft's Calista acquisition in Jan was aimed at solving a different problem - it focuses on the model where the Vista desktop OS has been virtualised. Centralised desktops are delivered to users via Terminal Services and RDP - which till now has resulted in an eroded user experience for graphics and video intensive apps. With the Calista tools...

users will enjoy the same rich user experience as with a locally executing desktop: Full 3D graphics with support for DirectX, Vista Aero and WPF applications, full frame rate video with 100% coverage for all media types, and fully synchronized audio.

Added together with the Softricity toolset, also part of the Microsoft Desktop Optimisation Pack, Microsoft have lined up easily the most comprehensive solution set for enterprise centralised desktop management to leverage Virtualisation in all its flavours.

Windows Virtualization Team Blog

£600m green investment for Scotland : Hi-Tech Scotland

SmallCountry — Mon, 21 Jan 2008 16:35:00 GMT

Internet Villages International announced that they will constructing datacentres on their 140 acre site in Lockerbie. The announcement of the creation of more datacentre space is hardly newsworthy, but there are a few unusual aspects to this project:

it will bring hi-tech employment to an area of Scotland not associated with the IT industry (500 jobs was mentioned), stimulating the local economy
it will take advantage of renewable energy sources including the local 44MW Biomass power station and various wind turbines in Dumfries and Galloway
the waste heat from the datacentres will used by local housing and horticultural businesses

Perfect.

Article, £600m green investment for Scotland : Hi-Tech Scotland

Head in the "Cloud"

SmallCountry — Fri, 12 Oct 2007 13:47:41 GMT

At the age of 24 I rode my bike across the 6.5km Columbia River bridge between Oregon and Washington state. Big river, big bridge. I remember thinking it would make a damn :-) good power plant.

Upriver, past Portland and in the shadow of Mount Hood, I hear Google have constructed a 30 acre server farm near the Dalles Dam, just one of several 2 gigawatt hydroelectric schemes up the Columbia gorge.

High bandwidth reliable networks, and the migration of computing power to "web services" in the cloud, mean that one should generally put servers next to cheap power. As long as the datacentre has the appropriate connectivity, the huge capacity and speed of global optical networks mean processing is best placed close to the power, rather than close to the people.

If such dam-envy seems dangerously close to train-spotting, consider that your average punter is now an enthusiastic user of web services like search engines, Internet mail/IM, and web mapping services.

Likewise IT Professionals are increasingly drawn to the irresistible logic of consuming services like email security and management (Exchange Hosted Services) and online Meeting services (Live Meeting) from "the cloud". Its far more efficient and effective to run these generic services centrally in a massively scaled, highly automated datacentre. This tallies up with the traditional business view of outsourcing "non-core" activities like payroll, security, catering, and PC desktop support.

Based on the success of companies like Salesforce, and huge investment from Microsoft, Google and others, it seems that in the next 2 to 5 years will see significant uptake of IT services that are remotely hosted and internet delivered. With data spread across the Internet, and applications running across the browser, desktop and remote datacentres, black-shirt wearing dot-com entrepreneurs have come up with new terms to describe the social and IT phenomenon's like "Web 2.0" and "cloud computing".

Microsoft's Software + Services mantra is a reflection of these new architectures, with some people factors thrown in - users can only use applications which are simple and familiar, and IT people want choice of where they put their servers (in their offices, centrally hosted in a datacenter, or completely outsourced to the cloud).

What this means in practical terms is that we will increasingly depend on 100,000-system datacentres, where automation and economies of scale drive down the cost of these services compared to the in-house IT approach. Microsoft's has a dual role in this - firstly to provide energy and effort efficient platforms and automated management tools, and secondly to provide the tools and building block services which enable this new S+S world.

Which brings us back to the Columbia River. One issue with the above is the growing power consumption of datacentres globally. George Gilder in Wired magazine recently estimated that the datacentres of the 5 leading search engines consume around 5 Gigawatts of power, counting servers, storage, cooling, and the inefficiencies of the grid's power transmission. Compare that to the 60GW peak demand on the UK national grid, and global energy requirements of datacentres can only grow as the global interweb's consumption of these services increases. Power accounts for around 40% of the overall running costs of the typical datacentre, a proportion which will only increase as energy prices increase and improved technology and automation reduce staff costs.

The moral of the story is that building your datacentres next to sources of cheap, clean power, makes good fiscal sense, and gives you a nice warm feeling about doing your bit for the planet.

Scotland, and in particular Orkney, Shetland and the Western Isles, has extraordinary renewable energy generation potential. Scotland has an estimated potential of 36.5 GW of wind and 7.5 GW of tidal power, 25% of the estimated total capacity for the European Union and up to 14 GW of wave power potential, 10% of EU capacity. Despite this the Scottish government have set modest targets for 6GW from renewable schemes by 2020.

Sadly, the costs and planning issues of grid connection can have an impact on their financial viability and social acceptability. But to overcome these grid-connection issues, why not consume the power at source by placing computing power close to point of generation? The Highlands and Islands region is blessed with lower temperatures and freely available water supplies (to aid cooling), cheaper than average labour, and plenty of tidal, wave, hydro and wind gen capacity.

Such a scheme could use self-contained, modular and portable container-based datacenters to overcome the logistical and skills issues, and allow the Highlands and Islands area to quickly capitalise on its sources of cheap and renewable power.

Container datacentre solutions allow a fully functional datacentre to be dropped in remote locations close to inexpensive power. Rackable systems unveiled such a solution recently, and James Hamilton of Microsoft, who frequently speaks about his ideas for the commodity datacentre, reckons that this approach can help overcome the myriad of political and logistical issues that surround datacentres and their power needs.

Power is currently about 40% of the cost of running a datacentre. With rising fossil energy costs, decreasing hardware/software costs, and decreasing staff costs through automation, this proportion will only increase unless abundant cheap power is nearby.

Coupling renewable energy with the commodity datacenter makes sense to meet our renewable energy goals, create a new source of employment and income for Scotland, and help bridge the predicted 20% " UK energy gap" by 2015.

Perhaps its time for Scotland to look towards the Renewable Energy Datacentre.

Virtually Normal

SmallCountry — Tue, 28 Nov 2006 14:15:00 GMT

Virtualised hardware technologies represent a fundamental shift in the way customers plan and deliver Windows server infrastructure. In the past 6 months, the majority of infrastructure projects I am aware of have included Microsoft Virtual Server or VMWare in their plans.

This trend holds true from SME, with partners like Castle Computer Services building consolidation business on VMWare virtualisation with workshops for their smaller customers, to the Enterprise, with companies like Barclays Capital building software developemnt and test environments on Microsoft Virtual Server.

This means that any new Windows Servers being installed, or existing servers being consolidated, are as likely to be deployed on a Virtual platform as direct on the physical hardware. Therefore an understanding of the Virtualised environment, Virtualisation product and licensing issues, is crucial for anyone who makes their living around the Windows platform!

Mr. Moore's law still holds true, and x64 hardware and OS platforms are now in common use. Virtualisation is seen by IT dept's as the easiest route to efficient and fully utilised servers - and consequently reduced maintenence, power and management costs.

In the past, it was one application per physical server (usually because an isolated application is a reliable application). By virtualising this application server to a virtual server image, many customers achieve impressive consolidation ratios, with >n virtual images running on a modern n processor server.

So the ongoing cost justification is clear. What is often unclear is the startup costs, which include licensing the Windows OS, the installed Server software and the Virtualisation platform . There are reams of white papers and web pages written on the subject, and knowing that such weighty tomes can be a real turn-off for most sane people, please find below a gross over-simplification of the situation.

Here goes.....

To support a virtualised environment, you 1st need the Virtualisation software (VMWare or Virtual Server), the Windows OS, and the applications running on the WIndows OS.

Virtualisation software: VMWare typically works out around $2000 per processor. Virtual Server R2 Enterprise edition is free.
Operating System (option 1) - 1 copy of Windows Server R2 Enterprise edition entitles the user to run 4 virtual Windows images on a single server. Typically $4000.
Operating System (option 2) - IT Manager with many server images to consolidate will choose Windows Server R2 Datacenter Edition. This is a per-processor license (around$3000) which entitles the user to run an unlimited number of virtual Windows images on the server. Thats a lot of Windows for not a lot of money.
Its worthy of note that the crafty IT manager can utilise the Datacenter edition licensing even in a VMWare environment - i.e. there is no requirement to actually run Windows Server Datacenter edition as the host OS
Server appliciations - Choosing the example of Microsoft SQL Server - both per processor and per server licensing allow for unlimited virtual instances of SQL on a Virtualised platform. There are some differences for other server applications, but I lost the will to live whilst reading the white paper.
Virtual Environment management tools: As consolidating on a Virtualised platform is the IT equivalent of putting all your eggs in one basket, it is key to ensure that the hardware platform, the virtualisation software, the OS and the server applications are all well served by management tools. VMWare would offer their Virtual Center and VMotion management tools, at a cost of around $2000 per processor. The Microsoft Virtual Server platform (plus OS, apps and hardware) can be managed from Microsoft Operations Manager, at a cost of around $400 per server.

...phew

So, to give an illustrative example - for a customer needing 30 8 way servers for server consolidation with high availability, a VMWare ESX based solutionwould cost >$500,000 for software licenses alone. The equivalent Microsoft Virtual Server +MOM solution would cost in the region of $100,000.

In my experience, when customers start on the route to a consolidated, virtualised environment, they are not aware of the typical costs for the entire infrastructure. Clearly its worth understanding these costs from the start, as choice of Microsoft or VMWare can have a significant effect on acquisition costs.

Great Train Journeys of our Time

SmallCountry — Tue, 10 Oct 2006 00:07:00 GMT

Living and working in Scotland we are privileged to have such wonderful countryside on our doorstep. As Mrs. Ferry usually annexes our car for short trips to "ladies who lunch" dates, I often find myself on public transport. This is no great hardship, as Scotland boasts some of the most picturesque train journeys in the world.

It was with this thought that I boarded the 0914 from Haymarket to Dundee, travelling over the Rail Bridge and up the Fife coast along the Forth Estuary. It’s a beautiful line, marred only by my destination – Dundee, city of Discovery and the wee jakie capital of Europe.

After dodging the junkies at the station, I was lucky enough to sit through a presentation from SUN during a visit to a customer. Although Microsoft and SUN have settled many of their differences this was the first time I have had the opportunity to sit in on a talk from our partner in co-operatition.

Scotland has long been a stronghold for Microsoft competitors such as Novell and SUN. Perhaps its something to do with the gulf stream, the low winter sun at these latitudes or the dogged focus on outdated ideas like independance, but Scotland’s IT decision makers till now have favoured lost causes. I’m sure SUN and IBM plants in Linlithgow and Spango valley have something to do with it too.

SUN Sparc-based servers running Solaris have remained the server platform in many larger Scottish enterprises long after the rest of the world have switched on to the better price/performance of other x64 OS/hardware combinations. A large financial institution I worked with until recently continues to deploy 32 bit Windows despite being entitled to deploy Windows x64 for applications which would make give them far greater performance, at the same price point and with a minimum of user re-training.

Clearly AMD Opteron x64 multicore processors and budget vendors like DELL have been nibbling away at SUNs SPARC business. So what I heard from SUN was total focus on the SUN x64 server range. SUNs ranges of discrete servers and blade clusters sound like beautifully engineered machines, but must be difficult to shift in a world where DELL are shipping much cheaper 4 socket AMD multicore servers. SUN tout their x64 range as a great platform for Solaris, Linux (red hat and SUSE), VMWare and Windows (in that order!). Their server consolidation story was familiar, referencing the 15% processor utilisation of typical servers, and the opportunity to consolidate these to bigger scarier SUN boxes using their proprietary Solaris partitioning technologies (if you only want to consolidate Solaris) or VMWare (for multiple OSs and versions).

Server consol through virtualisation sound great for low perf and legacy servers, but it only addresses the hardware support issue by enabling retirement of the poorly utilised servers. The real challenge is in how to manage multiple Virtual images, and how to manage the guest OSs and applications. SUN partnering with VMWare, and focus on XEN open source virtualisation is understandable given their competing position with Microsoft, but it leaves them with no story around how to manage the full software stack of Hardware/Virtualisation layer/OS/Application.

When a customer chooses to virtualise production services, the imperatives are the overall availability and performance of the service. Focussing exclusively on the OS virtualisation layer (e.g. VMWare's focus on expedient deployment of an OS image) does nothing to manage the availability of the Exchange, SQL or Citrix application service. Thus it is important to deploy virtualisation in the context of the toolset you will use to manage the overall application. Virtualisation alone makes life more complicated, not simpler!

Consider the Exchange example - IT professionals need management tools which give a single picture of the performance and health of all the Exchange services, any failures in the underlying hardware, the charachteristics of the Virtualisation layer, and what the multiple virtualised OS images are up to.

As Virtualisation becomes a commodity (moving into the Linux and Windows OSs) customers choice of virtualsation layer will be more about the strength of management tools moving forward. Only Microsoft have a complete story with System Center Virtual Machine Manager and Operations Manager....maybe an opportunity for a closer partnership with SUN J