Cloud Migrations: Solid Foundation or Shifting Quicksand? Part 1

Challenges of Cloud Migrations

Cloud MigrationsMigration to the cloud is supposed to deliver the business benefit of agility, specifically the ability to rapidly deliver services and applications at scale over top of a flexible pool of resources. Instead, network operators often discover too late in the game that their Cloud infrastructure is not so much a solid foundation, but rather resembles shifting quicksand.

IT organizations responsible for Cloud operations are often split into two teams, with one team responsible for Cloud infrastructure, and another for delivering services over that infrastructure. The Cloud infrastructure team handles all the hardware, including servers, switches, NICs, and basically everything up to and including the hypervisor. The services team is responsible for deploying and managing virtualized applications which utilize that infrastructure.

For better or worse, Cloud infrastructures are constantly shifting, as various new elements are rolled out, such as new versions of hypervisors, or new drivers NICs, both physical and virtual. Such changes can be happening on a weekly or even more frequent basis. While such changes can be benign, they can also affect application performance. 

A recent example which caused worldwide pain of this sort was when patches were rolled out for Meltdown and Spectre, which caused a performance overhead hit of as much as 800% in certain circumstances. Networks to which this patch was applied suddenly needed to have additional resources (in this case, as much 9 times the number of vCPUs) allocated to each application and service, just to keep up with pre-patch levels of performance. Even the CEOs of the largest chip vendors are publicly stating that flaws like Spectre and Meltdown should be expected again.

The Blame Game: How to Manage Change with Cloud Testing Services

Who is to blame in this scenario? The infrastructure team, for the sudden, un-planned shift in the foundation of what their resource pool could provide? Or is it the services team for not anticipating such possible changes themselves?

Other than saying that such events will continue to happen, nobody (other than perhaps the hackers themselves) can say with certainty what and when will happen. However, IT organizations can mitigate the risks of such performance-reducing patches by using the appropriate tools to manage change. In my opinion, it’s not a question of blame or even one team keeping another one honest. It’s more a matter of having visibility into the implications such infrastructure changes may cause to an organization’s particular workload, and about managing changes to an updated infrastructure.

To provide users with visibility into infrastructure changes is precisely why Spirent CloudScore application was developed. It allows IT organizations to measure Cloud performance with a multi-dimensional set of workload tests which stresses Clouds in various ways, then summarizes all the performance data collected into a single, aggregated score. Want to find out if your Cloud is better, worse or the same after an upgrade? Spirent CloudScore can show you. Want to know if your Cloud performance was better on Monday, Tuesday, or Wednesday, after a series of changes were rolled out? Spirent CloudScore can help.

CloudScore Topology

However, knowing the “score” is only half of the story. The second challenge is managing changes to an updated infrastructure. Each organization is unique, and moreover, the workload that each organization’s Cloud must bear, is also unique. The performance implications for a particular workload may be very different from an extreme case. In the case of the patches for Meltdown and Spectre, heavy Cloud user Netflix had to deal with a performance impact of only 0.1%. 

Read our new white paper on the implications of the Spectre and Meltdown vulnerabilities.

We explore this second challenge in greater depth in the sequel to this blog. Contact us to learn more about our cloud testing services.

comments powered by Disqus
× Spirent.com uses cookies to enhance and streamline your experience. By continuing to browse our site, you are agreeing to the use of cookies.