-
Live migration avoids downtime but using EVC can result in less efficient performance
-
Cold migration results in minimal downtime — typically less than 30 seconds
-
VM migration is necessary to keep systems operating efficiently
Virtual machine (VM) migration is a necessary evil in cloud computing and data center management. The task involves moving VMs from one host server to another to keep systems running smoothly, and it's often used when servers need to be upgraded or consolidated because of limitations in resources such as CPU, memory or storage.
While VM migration isn’t a new phenomenon, AMD says that the process is a crucial component for enterprises that are transitioning their IT infrastructure to a cloud environment because it can help them benefit from emerging technologies such as artificial intelligence (AI) and machine learning by freeing up space and making more power available for AI.
Specifically, Robert Hormuth, corporate VP for architecture and strategy in AMD's data center solutions group, said that VM migration is become more prevalent because of AI and the need to create more operational efficiency.
As hardware evolves, companies are looking to capitalize on the latest features, such as improved performance, security enhancements and energy efficiency as they migrate VMs to newer hosts, he said.
According to Red Hat, virtual machine migration can also be useful for load balancing, saving energy or moving geographic locations.
Myth: VM migration equals downtime
Many IT departments avoid VM migration whenever possible, primarily because they don’t want to deal with the inconvenience and they also fear it will result in network downtime. Instead, they just provision more resources onto the existing VMs or use software patches to extend the performance.
“IT shops don’t want to deal with VM downtime,” said Chuck Gilbert, senior director of cluster platform engineering, data center cluster engineering at AMD.
Experts, however, warn that this type of piecemeal solution can result in poor performance and poor power usage. Plus, it often means that companies are not able to take advantage of new features and capabilities.
Myth: Live migration is better
Some IT departments might opt for a live migration to avoid any downtime. A live migration is possible when moving to newer servers with the same vendor. When doing a live migration most companies will use VMware’s Enhanced vMotion Compatibility (EVC) to migrate to newer servers.
According to Hormuth, the downside of doing a live migration and using EVC is that VMs are forced to emulate the older hardware on the new servers.
“In order to support live migration, you are lying to the virtual machine about the version of hardware,” he said.
When VMs run on new hardware as if it is old hardware, that means that the enterprise won’t be able to take advantage of the latest security features or other improvements, which in turn can impact performance.
According to AMD this is particularly critical if the applications running involve AI, media or analytics.
In addition, using the EVC for a live migration also impacts performance because the servers will run less efficiently. “The guest OS believes it is running on older hardware so you are impacting the instances and the VMs that you can run on the machine,” Gilbert said.
Making servers run more efficiently is one of the advantages of VM migration, particularly as enterprises try to optimize their cloud assets and make way for AI.
“Historically, live VM migration was touted as a way of consolidating workloads live to enable powering down of unused capacity (sustainability),” said Roy Chua, analyst at AvidThink. “In reality, most IT teams were unwilling to shut machines down entirely in the fear that it would impact uptime if they didn’t come back up cleanly.”
However, Chua added that current CPUs that have lower-powered idle states, and the EVC technology could result in less energy use without a full shutdown. Because of this, EVC is often used for uptime management, such as migrating workloads at risk on servers with pending hardware failures.
Another advantage to live migration is that it helps with the portability of applications because it allows applications to continue to run while the IT administrator manages the maintenance in the background.
Myth: Cold migrations are unreliable
Live migration isn’t the only option. When an IT shop needs to do a VM migration and is using servers from either the same vendor or different vendors, they can do a cold (a.k.a. non-live or offline) migration. Cold migration is when the VM is powered off then moved to the new server and then powered up.
Red Hat noted in its virtualization documentation that cold migration options are "generally more reliable than live migration" and are recommended for VMs that have heavy memory loads.
IT managers often disregard cold migration as an option because of the potential downtime. However, Gilbert said that the downtime created during a cold migration is only about 30 seconds or less.
Plus, this type of reboot is typically used by IT departments when upgrading an operating system or deploying a security patch so it’s not an unfamiliar process. “Downtime is not that big of an issue,” he concluded.
While server upgrade and refresh cycles are an essential part of executing exceptional workload performance and IT services, Chua, ultimately argued that VM migration is less of an issue today because of the increasing number of enterprise and clouds workloads using microservice architectures and deployed on containers that are orchestrated by Kubernetes.