Support for Windows Server 2003 will end next year, but careful planning can ease operating system migration.
With Windows Server 2003 support ending next year, there is a pressing need for organisations to move their applications and servers to a newer version of Windows Server.
This is something a lot of companies will have to deal with over the next 12 months. Many organisations have already migrated, since the end of support for Windows Server 2003 was first announced in September 2009.
But according to figures from IT professional network Spiceworks, 67% of UK firms still have Windows Server 2003 running somewhere in their organisation.
Within those companies, on average nearly one-third (30.2%) of their servers are running that venerable operating system (OS).
According to a survey by managed service provider Avanade, four out of five businesses that are still using Windows Server 2003 plan to move from the platform before the end-of-life deadline in July 2015. Some 40% will migrate this year and 41% will do so between January and July 2015.
Once support is no longer available, servers running Windows Server 2003 will be vulnerable with no security updates – unless firms pay eye-wateringly expensive fees to Microsoft, at about $200,000 each.
With that in mind, what is the best course of action for companies needing to move to a newer, supported platform?
Lee Norvall, CTO of Fusion Media Networks, says one strategy is to continue to run some applications on the outdated 2003 and move important data and systems management to Windows 2012 R2, protecting both with a managed doubled-tiered firewall system.
“There are some cases where it is not possible to migrate applications,” he says. “It might also be the case that some files on the system are no longer live or required for ongoing business processes. These applications and files can be left as they are, moved to another server and put behind a firewall for safe storage, rather like locking them in a cupboard where you can see them through the keyhole and access them for reference when required.”
Ed Shepley, solutions architect at Camwood, says that for some legacy applications, making the move to a newer OS, such as Windows Server 2012 R2, may put older applications at the greatest risk of failure.
“They may well have been ported to 2003 from 2000,” he says. “Also complex apps that integrate with a lot of systems, as not only may their dependencies be incompatible, but their communication transports may have been deprecated on the new platform. Knowing upfront provides the information an enterprise needs to plan effectively around this risk.”
Quocirca analyst Clive Longbottom says applications could fail if they have been written with particular hardware in mind. “Some systems would access a SCSI device, such as a scanner or specific disk array,” he says. “As SCSI has been deprecated, it may be necessary to rewrite these areas of code to be more abstracted so as to enable USB, iSCSI, Sata or whatever device is to be embraced.”
Longbottom adds that anything based on completely physical constructs, such as the storage being in a particular location or the network card configuration, will struggle to move to a new server with a new OS. “The world has moved on so much since then,” he says.
Some applications had specific installation routines that were OS-specific, Longbottom points out. “As 2012 will report itself in a different way to the routine than 2003, you could end up with the dreaded ‘This application is not suitable for your OS’, or just ‘Incorrect OS found’, message. There are tools around that will run through the installation routine of an application and correct it wherever possible to make it more likely to install.”
Shepley says organisations should consider what business function these applications provide and break them down into component functions. “You are likely to be surprised how quick and easy it is to replicate, not the whole application, but just the parts the business needs, as a web app,” he says.
“To achieve this, the application needs reverse-engineering, its technologies identified, their communication transports determined, the data quantified and the use cases mapped. This information should be passed to the chosen development team to implement,” says Shepley.
“Old-school self-contained software development could be leveraged to answer this requirement, but consideration should be given to leveraging whatever technologies are supplied as services from cloud providers, that meet the needs identified in the reverse engineering.”
He adds that by hosting an application in the cloud, significant cost savings can be realised, not only in operational expenditure, but in the development cycle too.
There is no denying that the shift away from Windows Server 2003 could entail a shift away from the hardware that ran it.
Dan Simmons, technical architect at Trustmarque, says: “Windows Server 2012 R2 is a 64-bit only operating system and this should be taken into consideration if an organisation is currently using a 32-bit OS. It could throw up compatibility issues with older hardware or applications.”
Of course, Windows Server 2003 was also available in a 64-bit version, so this could give organisations a few more options for upgrading their datacentre with less risk. Longbottom points out it is likely that 2012 will have less of a resource requirement than 2003, due to improvements in coding.
“This is not to say that I recommend trying to shoehorn 2012 onto a Pentium Pro server, however,” he adds.
An upgrade to 2012 brings the opportunity to look to more modern and efficient hardware as well, with less power and cooling required, says Longbottom, and better in-built system management and automation, where patches and updates to outdated hardware items are available and maintained. “It also means a more virtualised approach can be brought in, giving better performance and higher levels of system availability,” he says.
Another aspect to migration is what tools an organisation will use to shift infrastructure onto a new OS. AppDNA and ChangeBASE can help move applications from 2003 to 2012 R2, and Shepley says organisations should be aware how these apps would be supported if they are manipulated to work where they are not designed to do so.
“It’s fine as a stop-gap, but nothing is as reliable as an application running on the platform it was designed for,” he says. “Lift and shift if you have to, but plan to upgrade to the supported version as soon as logistically possible.”
Tools such as AppZero will help in the migration process and BDNA could assist in application assessment, identifying the best migration approach.
Of course, the next version of Windows Server may well have a different name, but the same issues will probably still apply in dealing with compatibility and legacy of applications. Shepley says more forward-thinking enterprises will already have sorted out their migration plans and are looking at their next move as the stepping stone to the one after.