Last week Google announced the general availability (GA) of its Google Compute Engine (GCE) cloud service. One of the more intriguing aspects of the GA release is described in this blurb from the Google Cloud Platform Blog:
“We’re introducing transparent maintenance that combines software and data center innovations with live migration technology to perform proactive maintenance while your virtual machines keep running.”
What makes this concept of transparent maintenance noteworthy – and unique among cloud providers – is that it will allow Google to upgrade the hardware and the software (including security-related and other OS patches) of the physical infrastructure that GCE is running on without the interruption of the virtual machines that are running on that infrastructure.
When other cloud providers need to perform maintenance on their underlying hardware, you will typically receive an email a few days or weeks in advance letting you know that you need to shut down your instance and relaunch it within the provider’s cloud to have it reassigned to a different unaffected host. This is both a hassle for you as well as for your end users who experience downtime of the service provided by that instance. With live migration, the virtual machines are moved without any downtime or noticeable service degradation.
What Our Test Revealed
As a Google Cloud Platform Partner, RightScale gets a sneak peek at some of the new features and services that Google provides, so in addition to conducting a GCE performance test earlier this year, we were also able to test the new live migration feature a short while before the GA announcement.
We set up an environment that is fairly common among our customers. Our deployment had two load balancers running HAProxy on n1-standard-1 instances; an array of 10 PHP application servers, each running on an n1-standard-2 instance; and a MySQL master/slave pair, each running on an n1-standard-4 instance type. All these servers were in a single GCE zone with the exception of the slave database, which was running in a geographically disparate zone (we wanted to see if data replication was affected during the migration).
Google Compute Engine Live Migration Test Configuration
Normally we would not run all our load balancers and application servers in the same zone (thus creating a single point of failure), but for this test we wanted as much of our infrastructure as possible to be in the zone that was going to undergo the live migration. We ran an HTTP load testing utility against our deployment on an instance running in a cloud from another infrastructure-as-service (IaaS) provider, generating traffic across the public Internet with 20 concurrent connections with a random maximum delay of two seconds.
We ran this test for four hours with every request resulting in a write to the database in a very ordered way such that we could look for gaps in the data in a post-mortem analysis.
We kicked off our test, let the Google team know that we were ready, and then we waited to hear back from them that the migration was complete. A few hours later, we were told that our instances had actually been migrated twice – once off the hardware they were running on originally and then back to the hardware where they had started.
We took a look at our log files and all the data in the database and we saw…nothing unusual. In other words, if Google hadn’t told us that our instances had been migrated, we would have never known. All our logs and data looked normal, and we saw no changes in the RightScale Cloud Management dashboard to any of our resources, including the zone, instance sizes, and IP addresses.
Repeatability and Consistency of Performance Prove Key
This test validated that GCE live migration performs exactly as expected and that we can confidently recommend it to customers. I am not surprised that we achieved seamless live migration and no downtime to our applications during our test, because all of the GCE products and services that we have had early access to have performed flawlessly and, most importantly, consistently. With the repeatability and consistency of performance we have seen from Google, I believe the GCE platform to be an extremely reliable choice for enterprises on which to run all cloud-based workloads. With the recent announcement of Google Compute Engine GA and this new live migration capability, it’s now time to check out GCE.
To speak with a RightScale expert and learn how your enterprise can benefit from Google cloud management, request a complimentary consultation.
