The ability to migrate workloads seamlessly from one physical server to another -- often without significant downtime...
or noticeable disruption -- is one of the features of server virtualization that has driven its adoption. The capability directly supports Enterprise Management Associates' top five goals for virtualization deployments: increasing server utilization, reducing system downtime, reducing hardware costs, enabling disaster recovery and business continuity, and increasing flexibility and agility.
When considering server virtualization performance in your environment, the amount of time it takes to migrate a virtual machine (VM) from one physical server to another is an important factor. In this article on virtual machine migration, we examine this key performance indicator as well as virtualization best practices, provisioning software and other considerations for virtual machine migration -- particularly for heterogeneous virtualization environments.
One of the big myths of virtualization is that in isolation it enables seamless and instant VM migration. But according to the EMA report "Best Practices in Virtual Systems Management (VSM)," many of the 153 respondents have hardly achieved immediate migration. While the best performers conduct VM migrations in less than 30 minutes -- and many could probably perform migrations even faster -- the median migration time was between two and five hours. And the most common VM migration time was between one and two hours. Somewhat surprisingly, an enterprise that can migrate virtual machines in less than five hours can consider itself no worse than the average. In fact, below-average performers took five hours or more to migrate virtual machines, with a surprising 44% of enterprises falling into this group.
Virtualization management best practices
Live migration is one best practice -- but it is not a silver bullet, especially for non-x86 environments.
Provisioning is especially useful for migrating Unix, Linux, i/5 and Type 2 virtualization deployments.
Integration with ITIL, ITSM and BSM processes is another important way to speed VM migration.
VM migration is clearly not always as easy as it sounds. Even for enterprises with automated tools such as live migration and resource scheduling -- and not all organizations use these tools -- other factors creep in. The security and compliance postures of physical servers must be analyzed to ensure that sensitive workloads are not colocated in ways that create security or compliance risks. For example, locating a VM with customer credit card data on the same physical machine as a publicly accessible Web server would violate Payment Card Industry (PCI) regulations. Even running two VMs on a single machine can violate PCI . VM migration may also need to go through a rigorous IT Infrastructure Library (ITIL) process requiring approval by a change-approval board (CAB) and configuration recording in a configuration management database (CMDB) or configuration management system (CMS). In this case, instant and ungoverned live migration would violate corporate policies. Unplanned migration of virtual workloads can even cause resource issues outside the server environment, as colocation can introduce database contention, overload a network device or transport, or cause unexpected delays in storage I/O.
In EMA's research into VSM best practices, a few disciplines correlate strongly with the best outcomes. Best performers use the following practices more often than average or below-average performers.
- Virtual machine management software. Of course, migration tools such as VMware VMotion, Citrix XenMotion or Microsoft Virtual Machine Manager are likely to improve migration speed for VMware ESX, Citrix XenServer and Microsoft Hyper-V environments, respectively. But live or quick migration is not a silver bullet. Apart from policy requirements, which may make live migration unacceptable, most enterprises have some Unix, z/VM or i/5 (AS/400) virtual machines, and many use operating system virtualization (such as Microsoft Virtual Server, Parallels Virtuozzo, Solaris Containers and Linux Jails). None of these virtualization environments have live migration capabilities.
- Provisioning software. Perhaps counterintuitively, EMA's research shows that average and best-practice outcomes for new VM deployment are actually the same as for VM migration. This may be because provisioning of OSes and applications from scratch with tools such as Symantec Altiris Server Management Suite or Avocent LANDesk Server Manager is still a surprisingly popular method of managing VM migration. Indeed, these tools are especially important for migration of virtual Unix or Linux environments, which require up-front provisioning.
- Integrated process management tools The best performers in VM migration also are more likely to use software that integrates management of virtual systems with ITIL, IT Service Management (ITSM) or business service management (BSM) processes. Tools such as CA Spectrum Automation Manager or Opalis Integration Server can migrate virtual workloads automatically using policy-based workflows that accommodate not only low-level VM migration but also processes for storage and network migration, security and compliance assurance, and integration with change recording, change-approval and configuration management systems. This allows for faster VM migration, while still following complex processes and policies.
Certainly, for x86-based server virtualization, live migration is an excellent starting point. But it is not the last word in best practices for VM migration. For non-x86 environments especially, provisioning is one of the main methods to enable migration of virtual workloads. Because almost all environments are somewhat heterogeneous, provisioning is a critical VSM discipline for VM migration. Even in pure x86 environments, integrated tool sets that encompass an entire workload's migration within approved policy can deliver better overall results than can live migration alone. The best approaches to accelerating VM migration time with virtual systems management will accommodate all these disciplines.
About the author:
Andi Mann is a research vice president at IT analysis firm Enterprise Management Associates (EMA). Mann has over 20 years of IT experience in both technical and management roles, working with enterprise systems and software on mainframes, midrange servers and desktops. Mann leads the EMA Systems Management research practice, with a personal focus on data center automation and virtualization. For more information, visit EMA's website