Tag: Microsoft Azure

Last month, I spent some time at Microsoft’s Partner Business Briefing on Transitioning to the Cloud (#pbbcloud).  To be honest, the Microsoft presentations were pretty dull, the highlight being the sharp glances from Steve Ballmer as he saw me working on my iPad (which led to some interesting comments in a technical session a short while later) but there was one session that grabbed my attention – one where easyJet‘s Bert Craven, an IT architect with the airline, spoke about how cloud computing has changed easyJet’s “real world” IT strategy.

For those who are reading this from outside Europe, easyJet was one of the original UK-based budget airlines and they have grown to become a highly successful operation.  Personally, I don’t fly with them if I can help it (I often find scheduled airlines are competitive, and have higher standards of customer service), but that’s not to say I don’t admire their lean operations – especially when you learn that they run their IT on a budget that equates to 0.75% of their revenue (compared with an average of just over 4%, based on Gartner’s IT Key Metrics).

Bert Craven quipped that, with a budget airline, you might be forgiven for thinking the IT department consists of one guy with a laptop in orange shed, by an airfield, operating a shoestring budget but it’s actually 65 people in a very big orange hangar, by a big runway…

Seriously though, operating a £3billion Internet-driven business on an IT budget which is so much smaller than the norm shows that the company’s reputation for leanness is well-deserved. To deliver enterprise-scale IT with this approach requires focus – a focus on differentiation – i.e. those systems that drive competitive advantage or which define the business.  In order to achieve this, easyJet has taken commodity systems and pushed them “out of the door” – buying as-a-service products with demanding service level agreements (SLAs) from selected business partners.

easyJet has had a cloud strategy since 2005, when they started moving  commodity systems to managed services.  But, in 2009, Windows Azure caused a deviation in that approach…

Until 2007, easyJet was growing at 20% p.a. (the company is still experiencing rapid growth today, but not quite at the same level) – and that high level of growth makes it difficult to scale.  There’s also a focus on meeting the SLAs that the business demands: easyJet are immensely proud that their easyJet.com availability chart is so dull, showing a constant 100% for several years; and if their flight control systems were unavailable for more than four hours, the entire fleet would be grounded (which is why these systems are never “down”).

So easyJet classified their IT systems into three tiers:

• Commodity: operating system; security; backup; e-mail; access methods; file and print.
• Airline systems: engineering; crew rostering; finance; personnel; flight planning; schedule planning; baggage handling; payroll; slot management; payment systems.
• easyJet Systems (those differentiators that drive competitive advantage): reservation system; revenue management; departure control; crew systems; aircraft systems.

The three tiers of classification are used to drive SLAs of silver  (99.9% availability), gold (99.99% availability) and platinum (100% availability), mapped onto the IT architecture such that platinum services operate as a high availability configuration across two sites, gold services can fail over between sites if required, and silver services are provided only from easyJet’s primary site.

Alongside this, easyJet has a 5 point IT strategy that’s designed to be simple and cost-effective:

1. Use simple, standard, solutions by default.
2. Promote innovate use of mainstream technology.
3. Use Microsoft technology for the technical platform.
4. Provide scalable systems that never restrict the growth of the business.
5. Provide 100% up-time for business critical systems.

In order to take account the disruption from the adoption of cloud computing technologies, easyJet adapted their strategy:

1. Use simple, standard, solutions by default: place services in the cloud only when to do so simplifies the solution.
2. Promote innovate use of mainstream technology: continually test the market to measure capabilities and penetration of cloud technology (as it becomes more mainstream, it’s better suited to easyJet’s innovation).
3. Use Microsoft technology for the technical platform: Windows Azure will be the natural choice, but look at alternatives too.
4. Provide scalable systems that never restrict the growth of the business: look for areas where cloud will improve the scalability of systems.
5. Provide 100% up-time for business critical systems: wait for cloud computing to mature before committing to high availability usage (no platinum apps in cloud).

Naturally, easyJet started their journey into cloud computing with commodity computing systems (buying in compute and storage capabilities as a service, outsourcing email to a platform as a service offering, etc.) before they started to push up through the stack to look at airline systems. They thought that silver and gold services would be offered from the cloud within the architecture but Windows Azure turned out to be more disruptive than they anticipated (in a good way…).

With many commodity and airline systems now cloud-hosted, easyJet’s IT systems are able to cope with the company’s rapid growth.  But their departure control system (a platinum service) in 4 airports now has been running on Windows Azure since January 2010 and the easyjet.com sales channel has also been extended into the cloud, so that it may be offered more broadly in innovative ways (or as Bert put it, “when you suddenly expose your sales channel to Facebook, you need to know you can handle Facebook!”).  Now easyJet’s high-level architecture has platinum systems crossing primary and secondary sites, as well as the cloud – something that they originally said they wouldn’t do.

Bert Craven explained the crucial point that easyJet missed in their strategy was an aspect that’s often understated: integration as a service.  easyJet believe that the potential of the cloud as an integration platform is huge and they use Windows Azure AppFabric (formerly known as BizTalk Internet Services, then as Microsoft .NET Services).

With a traditional secure service in a data centre, consumers are allowed to punch through the firewall on a given port then, after successfully negotiating security, they can consume the service. AppFabric turns this model inside out, taking the security context and platform into the cloud.  With AppFabric, the service makes an outbound connection through the firewall (security departments like outbound connections) and consumers can locate services and connect in a secure manner within the cloud. AppFabric is not just for web service endpoints either: it can do anything from send a tweet to streaming live video; and endpoints are getting smarter too with rich integration functionality (message routing, store and forward, etc.)

easyJet see AppFabric as a game changer because it’s made them ask different questions:

• Instead of “is a new service we’re building a  cloud service or an on-premise service?”, the question becomes “might this service have some cloud endpoints and components – is this in fact a hybrid service?”
• “Could we migrate an existing service to the cloud?” becomes “could we extend an existing service into the cloud?”

Bert Craven believes that AppFabric is the ace in the pack of Windows Azure technologies because it’s a small step to take an existing service and expose some endpoints in the cloud (easy to swallow).  It’s also more  of an enabler than a disruptor, so AppFabric is quite rightly perceived as lower risk (and almost certainly lower cost). Extending a service is a completely different proposition to move entire chunks of compute and storage capabilities to cloud. Consequently there is a different value proposition, making use of existing assets (which means it’s easier to prove a return on investment – delivering more value with a greater return on existing investments by extending them quickly becomes very attractive) – and an IT architect’s job is to create maximum business value from existing investments.

AppFabric offers rich functionality – it’s not just a cheap shortcut to opening firewall and has a rich seam of baked-in integration functionality and, ultimately, it has accelerated easyJet’s acceptance of cloud computing.  18 months ago, Craven describes sharp intakes of breath when talking of running departure control in the cloud, but now that a few airports have been running that way for 10 months, it’s widely accepted that all departure control systems will transition to the cloud.

Bert Craven sees AppFabric as a unifying paradigm – with Windows Azure AppFabric in the cloud and Windows Server AppFabric on-premise (and it gets stronger when looking at some other Microsoft technologies, like the Project Sydney virtual private network and identity federation developments – providing a continuous and unified existence with zero friction as services move from on-premise to cloud and back again as required).

In summing up, Bert Craven described AppFabric as a gateway technology – enabling business models that were simply not possible previously, opening a range of possibilities. Now, when easyJet thinks about value propositions of cloud and cloud solutions, equal thought is given to the cloud as an integration platform, offering a huge amount of value, at relatively low cost and risk.

Today is the day that Microsoft’s cloud-based computing platform, Windows Azure, which has been running through a phased launch over recent months, becomes a chargeable service.

I don’t know if it’s expensive or not – it looks a bit steep to me – but I don’t have a reference point (other than simple web hosting) and, because of the granular charging structure, it’s difficult to get a true idea of what the charges might be for any given application.  Expensive or not – what Azure provides is flexibility (i.e to mix on-premise and cloud-based infrastructure), elasticity (e.g. to cope with bursts in computing resources requirements) and familiarity (i.e. the code deployed to the web is still Microsoft .NET code, or  indeed it could be something else - like C++, Java, Ruby or PHP… although you might need a small amount of .NET “veneer”).

I don’t pretend to understand all of the details of Windows Azure – I’m no developer – but Steve Marx has produced a great introduction to Azure video for non-techies:

If this sounds interesting to you, then read on… because, last Friday, I was at Microsoft UK’s offices, where I was fortunate to see Microsoft UK Developer Evangelist Eric Nelson (Twitter @ericnel) present his seven things that may surprise you about the Windows Azure platform. In his presentation, Eric also highlighted a couple of offers that can be used to avoid paying for Azure – for example if you have a small cloud-based app (perhaps a demonstration), or if you’re just dipping your toe into the water:

• The Windows Azure Platform introductory special offer, gives 25 hours of free compute time, together with 0.5GB of storage, and a small amount of bandwidth/storage transfers each month.
• There’s another introductory offer, for MSDN subscribers, which currently offers 750 hours of compute time a month (that’s a server running 24×7).

If you’ve already been having a look at Azure, you might want to take a look at the Windows Azure Feature Voting forum where it’s possible to vote (up to 10 times) for Windows Azure features – as the site is run by Mike Wickstrand, who is Microsoft’s Senior Director for Windows Azure Product Planning, it should be somewhere that opinions get noticed (and I’d like to see a few more Microsoft product groups take on this idea…).

If you want to know more about Azure, then the Windows Azure Team Blog would be a good place to start but, for those of us in the UK, 20-27 February 2010 is UK Azure Awareness week – watch the UK fans of the Windows Azure Platform site for more information.

At last year’s Microsoft Professional Developers’ Conference, Windows Azure was the big news. Finally, Microsoft had put its cards on the table and announced their strategy for cloud computing!

But, since Ray Ozzie’s keynote in LA last autumn, it’s seemed pretty quiet on the Windows Azure front. That’s understandable – Azure is still in development and it will be some time before we see mainstream use of this computing platform – even so, I was interested to attend the inaugural meeting of UK Azure User Group (AzureNet – not to be confused with the hosting company by the same name) at Microsoft’s London offices, including a presentation from Microsoft’s James Conard about what Windows Azure really is.

James is a Senior Director of Developer and Platform Evangelism at “corp” (i.e. he works in Redmond, not Reading!) and he looks after the Microsoft.NET Framework, Visual Studio and the Azure Services Platform – all things that I know very little about but, based on his presentation, I think I’ve got a grip on how Windows Azure hangs together.

So what is Windows Azure? James Conard described it as:

“An execution environment in the cloud for your applications.”

Which begs the question of what is the cloud? Conard’s view is that this is being made out to be more complex than it really is – as normally happens with a platform shift. Analysts/reporters/experts [bloggers!] define new terms such as: cloud computing; platform as a service; software as a service; infrastructure as a service – and some of these terms are sticking as vendors snap their marketing onto the terms.

But today’s application challenges are not specifically related to technology or to a platform – they are issues like:

• How many users will an application need to support (after 1 month? 6 months? a year?)?
• What are the bandwidth, storage, server, rackspace requirements?
• How can we handle scalability (up and down)?
• How can we provide high availability?
• How can we quickly go live?
• How can we reduce operational costs?
• How can we move to a service delivery model?
• How can we provision servers for the short term (without buying extra infrastructure)?

Windows Azure is intended to provide three core services: compute, storage and management which are:

• Scalable – with a virtualised hosting environment.
• Flexible – providing storage with blobs, tables, and queues.
• Manageable – with a model-driven service lifecycle management.
• Usable – with a rich local and offline developer experience.

Looking first at the compute service, applications are built (based on role definitions, modelled using an XML service configuration file), deployed to the web and run via a load balancing mechanism. Building the application in Visual Studio 2008 (SP1) with the SDK for Windows Azure and Visual Studio Tools for Azure ensures familiarity for Windows developers and the SDK’s Development Fabric simulates a cloud application whilst running locally for debugging purposes. The publishing process packages the application as a service package including all assemblies and configuration files ready for upload to the Azure Services Platform via the Azure Services Developer Portal. This portal allows the developer to create a hosted service and access production and staging environments. Once initialised according to the applications configuration, the appropriate number of instances (virtual machines) is started and the application deployed. Staging environments use a DNS name in the form of guid.cloudapp.net but friendly names are provided for production environments.

Most applications need to store and manage data, and Azure provides access to tables and blobs, with a queuing mechanism for communications between roles (some of which may run asynchronously, others as batch jobs, with a worker process to handle the interaction).

Management is concerned with ensuring that there are sufficient instances of a running application, monitoring when to bring more computing resources (extra instances) online, and when to take down one or more instances.

In short, Azure provides the infrastructure to run an application in the cloud including the environments upon which to run code and the underlying servers, datacentre services and connectivity. There is no need to directly manage this as the application is abstracted from the infrastructure and Microsoft highlights that, whilst some vendors focusing on the infrastructure (physical and virtual machines up to the operating system layer), over time, Azure will expose more and more capabilities of the Windows Server operating system, Microsoft .NET framework, etc. and the available services will be expanded.

Of course not all applications are as simple as the ones that Microsoft uses to demonstrate Azure. Some applications need additional capabilities in the cloud such as:

• Relational database support.
• Connectivity between on-premise applications and cloud applications (some applications cannot move to the cloud and the data needs to remain local, or the functionality does – hence hybrid applications).
• Single sign on support.
• Federation with existing identity providers.
• Orchestration of several different services.
• Access to user profile and contact data.

Windows Azure is a baseline and there are also additional services within the Azure Services Platform, implemented as building blocks which may be consumed (as required, wholesale or piecemeal) from an application running on Azure, including:

• Live Services, for building user-centric applications that require end user interaction and access to contacts, calendars, folders, etc.
• .NET Services, providing key building blocks required by many cloud-based and cloud-aware applications including access control, a service bus and workflow.
• Microsoft SQL Services, extending Microsoft SQL Server into the cloud for – cloud-based instances of SQL capabilities.

In the future, these initial services will be supplemented with:

• Microsoft SharePoint Services, for workflow, list management and document management.
• Microsoft Dynamics CRM services.

SQL Data Services link applications to SQL Server using SQL’s tabular data stream (TDS) protocol. James Conard’s presentation only mentioned it in passing, but there is a session available online from the MIX09 conference with the detail on SQL Data Services.

(At this point, I’d like to plug Jamie Thomson – SSIS Junkie – and my former rival for the top blogger spot when I worked at Conchango… Jamie was at the AzureNet meeting too – if you want to know about how SQL and .NET services fit together, he’s your man!)

A few paragraphs back, I mentioned the service bus and this could probably do with a little more explanation. Referenced using a URI with a sb:// suffix, this provides a publisher/subscriber model based on service bus queues (queuing messages until a listener is available) and service bus routers (distributing messages according to the routing policy) with a REST-based interface for managing access control (via authorisation rules).

So, what next for Windows Azure? The latest Community Technology Preview (CTP) was issued in March 2009 and includes a number of changes and improvements including:

• Full-trust support for .NET applications, allowing native code to run as part of service package.
• FastCGI support on IIS7 (for running PHP and other extensions).
• A single SDK and tools installation.

Soon, Windows Azure will be available in a second United States datacentre (and developers will be able to select which to run in via the portal) and, as Windows Azure approaches commercial availability, there will be a datacentre in Europe too.

Finally, pricing and service level agreement information is expected during Summer 2009 with commercial availability in the Autumn.

As an infrastructure guy, I might be scared by the idea of all of this infrastructure moving into the cloud, but there are a few things to remember:

• The transition won’t happen overnight.
• Many organisations will still require an extensive local infrastructure (if only for client connectivity to the cloud).
• Someone has to build and run those cloud datacentres!
• Security will be key to the success (or failure) of this brave new world.

Regardless of my future job prospects, I’m looking forward to the day when someone asks me to integrate an on-premise infrastructure with Windows Azure.