Upgrading or changing your server is a task most businesses eventually face, but it’s not a process you can undertake casually: careful consideration and planning are essential, especially as the push towards virtualisation and cloud creates more complex scenarios. Follow these real-world tips for a more straightforward upgrade path.
I’ve drawn these ideas from a presentation given by two customers at the Australian press launch for Windows Server 2012 on the Gold Coast this morning: equipment hire firm Kennards Hire and boutique advisory firm McGrathNicol. Both took part in the Windows Server 2012 Rapid Deployment Program (RDP), which meant they were reviewing the software well ahead of its official release last week. While that gives many of the tips of Windows Server 2012 focus — a reasonable emphasis given that 75 per cent of servers in Australia run Windows — many of these principles will apply whatever platform you use.
Virtualisation is great, but you get less staff
The ability to run multiple virtual machines and easily manage and switch them between hardware remains one of the most evident shifts in the server marketplace.”We simply want to virtualise everything,” Microsoft’s principal program manager Jeff Woolsey noted. That principle also influenced the inclusion of features: “We didn’t want any features that precluded live migration.”
Virtualisation mania can become extreme. “On an average day within the Windows Server team, we create around 30000 Hyper-V virtual machines,” Woolsey said.
One less-commented consequence of this switch (admittedly one which Microsoft itself won’t be worrying about) is that IT departments get smaller. Kennards has 900 employees and a 10 person tech team servicing its 125 branches. “Things are always fairly tight for us,” IT manager Richard Fox-Smith said.
McGrathNicol does slightly better, with 13 full-time staff and 5 employees, but its staff are even more mobile, often visiting remote locations to offer business advice or manage liquidations. It also has an ambitious vision for concealing the complexity of its systems from staff. “If people are on premises, they use our servers,” CIO Shiran Herath said. “If people are off premises, they use the cloud solution.”
Identify your business drivers
For Kennards Hire, the main focus is on ensuring each individual branch is able to serve customers. “The branch and the branch manager are really priority number one for us,” Fox-Smith said. Another key issue was disaster recovery: while Kennards Hire ran a large server environment based on Windows Server 2008, it was in a single location. “Disaster recovery was becoming an issue for the business. We needed a technology that would make disaster recovery easier.
Windows Server 2012 solved that particular issue through its replication feature, which allows easy duplication of virtual machines without having to stop them. “The replica feature is a simple way to get disaster recovery out of the box. “Moving those resources around is really flexible.” Whatever the platform, the more important lesson is that a need was solved with technology, rather than technology being deployed in a vacuum.
Too many vendors can be painful
A common theme in both these installations is the shift towards using a single provider. For McGrathNichol, that has been a longstanding choice, reflected not just in its software choice (Microsoft) but also its deliberate choice of other single partners (Cisco, HP and Telstra). “We want to work with the best in the industry,” Herath said.
Kennards Hire also wanted an entirely Microsoft shop, but ended up choosing Citrix as its initial hypervisor system. When it first shifted to 2008, the options in XenServer were better than Hyper-V, Fox-Smith said. With 2012, the gap was less visible.
Advance testing is critical
That step only came about as the result of long-term testing, however. “You don’t want to be on the bleeding edge too much. Being able to test across two candidates was great; our confidence level is now very high.”
Even post-deployment, McGrathNicol focuses on testing, regularly examining whether it can move workloads between its on-premises systems and its Azure services. “We are not looking at the cloud as a separate environment,” Herath said.
Don’t forget about systems management
With all this emphasis on virtual machines, it can be easy to lose focus. “People focus way too much on the hypervisor and tend to underinvest on systems management,” Woolsey said. Knowing how many virtual machines you are running is useful, but knowing whether they comply with business policies is more important.
Your users don’t care about the details
An important consequence of trying to solve business problems is that staff outside the IT department generally won’t care about what you deploy. “We want staff not to be worrying about technology at all,” Herath said. “Our objective is to make it transparent.”
“They really care about the availability of the POS system at the branch,” Fox-Smith agreed. “That’s what counts.” At board level, directors “care about the fact we have a disaster recovery solution; they understand it from that perspective.”
As an IT manager, you’ll still have to make those choices when deploying or upgrading servers. But as ever, feature lists won’t matter as much as matching those features to actual business requirements and making the process as smooth as possible.
Got your own war stories to share? Let’s hear them in the comments.
Visit Gizmodo’s TechEd 2012 Newsroom for all the news from the show.
Evolve is a weekly column at Lifehacker looking at trends and technologies IT workers need to know about to stay employed and improve their careers.
Disclosure: Angus Kidman is attending TechEd 2012 as a guest of Microsoft.