vSphere vMotion keeps getting better and better. We made huge improvements in vSphere 7 allowing vMotion to take advantage of high-speed bandwidth NICs, improving large memory VM page copies, and improvements in the fast suspend and resume process for Storage vMotion just to name a few. For everything vSphere vMotion related, check out core.vmware.com/vmotion.
In this blog, we'll introduce vMotion notifications to improve vMotion compatibility with sensitive applications and the Unified Data Transport protocol to dramatically speed up the data transfer during disk migrations introduced in vSphere 8.
vSphere vMotion is easily one of the best features in vSphere and virtualization in-general, and often taken for granted. It allows for almost all virtual machines to be seamlessly migrated and moved without any impact whatsoever to the virtual machine and applications running within. What about those very few cases where using vSphere vMotion may impact an application? Applications that are very sensitive to latency, applications that are clustered and span multiple virtual machines and distributed databases, just to name a few examples.
With vMotion notifications, these applications can have improved interoperability with vSphere vMotion.
Reduce Outages by Preparing Applications for Migration
As mentioned, certain applications cannot tolerate the stuns associated with vSphere vMotion. These applications can be written to be migration aware to improve their interoperability with vSphere vMotion. Applications can prepare for a migration event. This could be gracefully stopping services, quiescing the application, performing a failover in the case of a clustered application or other operations. The application can delay the start of the migration up until the configured timeout but cannot decline or prevent the migration from occurring.
Go deeper with vMotion notifications:
Unified Data Transport
vSphere vMotion is blazing fast and can migrate the running state of virtual machines from one ESXi host to another in seconds. vSphere Storage vMotion of powered-on virtual machines is also extremely fast because it uses the same optimized vMotion protocol.
But what about migrating powered off virtual machines? If you've ever migrated a powered off virtual machine with a large disk size from one datastore to another, you probably have noticed it taking a significant time. This is because when a virtual machine is powered off, the Network File Copy (NFC) protocol is used. The same is true for disks with snapshots when a VM is powered on. The non-active disks will be transferred using NFC. NFC is magnitudes slower than vSphere vMotion.
Introducing Unified Data Transport
To solve this problem, we introduce a new protocol called Unified Data Transport (UDT). In a nutshell, UDT combines the best of the NFC and vSphere vMotion protocols. Unified Data Transport (UDT) uses NFC as a control channel but offloads the data transfer to the vSphere vMotion protocol to benefit from the substantially greater performance and throughput.
Go deeper with Unified Data Transport:
With vSphere 8, vMotion notifications improve vMotion compatibility with sensitive applications by allowing applications to be aware of migration events and react accordingly. The Unified Data Transport protocol dramatically speeds up the data transfer during disk migrations by leverage the vSphere vMotion protocol to perform cold data migrations.
For more on vSphere vMotion, check out these other articles and resources.