Six rules for migrating your data center successfully
by Francis Miers, Director, Automation Consultants
There are various reasons why companies might need to migrate their data centre. They are spurred into action by regulatory requirements, mergers and acquisitions, space limitations or the simple need to reduce overheads. Moving from one data centre to another, or from a data centre to the cloud, is a common event – however, it can be quite challenging to get right.
The process involves a number of complexities, such as relocating application software, infrastructure software and sometimes hardware, and all their interdependent elements. Throughout the entire migration, there are two cardinal imperatives: no data can be lost and no unplanned downtime can occur.
Fulfilling these imperatives is easier said than done. Some applications, for example, may communicate over specific connections that are not so easy to replicate when a new location is a long distance away. In some instances, security and regulatory concerns prohibit the movement of applications. In others, the information on existing IT assets may not be up-to-date, making it impossible to predict the impact of decommissioning certain elements of hardware or infrastructure.
However, the benefits of a successful migration are well worth the effort. A company can gain significantly from improved performance, administration and more efficient administration and operation of the migrated systems. Here are six rules to help guide a successful data centre migration.
- Assess the existing setup
A company’s knowledge of its IT assets is often incomplete. As the assets get older and become more and more outdated, this knowledge declines even further. Over the years, process reports, training manuals and internal documents often go missing, and those that remain are rarely kept current.
What’s more, employees who are familiar with the older systems may have moved on and taken their knowledge with them. To ensure as smooth a migration as possible, you need to review your existing data centre in its entirety. If you start the migration with any information gaps, you’ll encounter problems.
Bear in mind that each application will very likely be involved in multiple relationships and dependencies with other applications and infrastructure elements. Network tracing tools can help untangle this web and (re)discover which components are communicating with others.
However, there is no point using these tools a couple of months before your planned migration as some systems only communicate with each other every few months or even once a year. Start as soon as possible and build a comprehensive picture of how your data centre works well in advance of a migration.
- Choose the best migration method for each application
There are several different methods of migration available for each application – it all depends on what kind of technology the application is running on. The ‘lift and shift’ method refers to physically moving the hardware from one location to another and is best-suited for applications that use older technologies.
Other options that are more relevant for applications running on newer technologies include restoring the machine images, copying virtual machines to the new data centre, reinstalling the application and migrating the data. Once you have chosen the most relevant migration method, you need to work out how much testing the process will require.
It’s worthwhile to conduct a trial migration and rigorous testing on particularly important or ‘mission-critical’ applications.
- Give the high-priority applications a trial migration
When moving ‘mission-critical’ applications, a trial migration followed by intensive testing is a worthwhile process. If an application is less crucial, such as one which tracks employee birthdays, then a trial and testing phase is not particularly necessary. The decision is ultimately based on how important the system is to the business and what technology is in play.
A high priority application typically contains sensitive data. To avoid any downtime or data loss during the migration, you need to schedule regular backups and have a solid pre-migration roll-back plan in place – just in case. A trial migration tests the chosen method and will identify any problems before they can upset the live migration.
- Consider your legacy systems
An IT system may be old, but that doesn’t mean it’s redundant. Many IT systems still provide companies with good service decades after their installation. The challenge is that migrating a legacy system to the cloud, or to a new data centre, is not a simple procedure.
For example, if one of your applications runs on a VAX minicomputer from the 1990s and is moving to another traditional data centre, then the ‘lift and shift’ method is a feasible solution. However, if the application is being transferred to the cloud, it would need to be ported to a VAX emulator before migration. Failing that, you’ll have to decide whether to rewrite, replace or remove the application completely.
- Make space for applications in the new network
Before embarking on a migration, you need to allocate space for each application in the new network. Once again, this is not a simple process. You need to consider all the infrastructural elements such as local network design, external connectivity, monitoring, operating systems and databases, and servers.
Some of the older applications may not be compatible with the new data centre’s security. In such cases, you may need to relax network security measures to accommodate them, have aspects of the application rewritten to fit the new data centre’s requirements, or replace them with something more modern.
- Prepare for latency
Older legacy systems are more susceptible to latency issues during a migration than modern, web-based applications.
Temporary latency can be mitigated by scheduling the migration of interdependent applications as closely together as possible. In other words, systems that communicate to each other via a local area network should all be moved within a tightly defined time period – especially if the new server is many miles away.
If latency is predicted to affect the performance of an application in its end-state, then an application delivery system such as Citrix XenApp can help ensure that users remain unaffected. Or, the application can be reconfigured, rewritten, abandoned or replaced.
It’s important to consider all the relevant preventative measures upfront as network latency, in any form, affects performance and needs to be managed well in advance.