Scale Up, Scale Out
Over the past week, we’ve spoken to a number of professionals in the cloud native computing space, as well as those in the software development, about what changes they are seeing in network and application usage, as COVID-19 alters the way we use these systems.
Apurva Joshi, vice president of products at DigitalOcean, noted that the cloud service has seen a considerable increase in usage, especially from customers offering video streaming and gaming services for folks stuck at home waiting out the pandemic. These users are spinning up more compute instances to meet this demand. “We usually plan our capacity six months in advance, but the current, unexpected situation has forced us to fast-forward our usual process. For example, we’ve added more capacity to our data centers than usual to make sure our customers are not facing interruptions,” he said
We are learning that scale-out technologies, such as Kubernetes, can help an organization more easily meet this increased demand. Video teleconferencing service 8x8, for instance, saw a 50-fold increase in its traffic as the pandemic stretched around the globe. Most organizations can squeeze two to three times the capacity from their current systems, but 8×8 had to go to 350 times the norm — to serve about 150,000 new users a day. Kubernetes ensured that there was no single point of failure and that bottlenecks could be eliminated through scaling out. “If properly architecturally implemented, you can scale horizontally. That allows you to handle the load and have high uptime and availability,” 8×8 Chief Product Officer Dejan Deklich told The New Stack.
Michael Ferranti, vice president of marketing at container-native storage company Portworx, also saw a number of Portworx’s customers scaling up with Kubernetes to meet demand, especially in the retail and gaming industries. “The nice thing about Kubernetes is that it’s basically a scale-out model, so they don’t have to re-architect the application in order to hit that demand, they just need to add more servers,” he said.
Another helping factor is automation. Fortunately, 8x8 “invested tremendously in automation,” Deklich said, which helped when the company saw its explosion in demand. It follows the design rule that if you do it once, you do it by hand; but anything more than once needs automation. IT automation software provider SaltStack agreed with this assessment as well. Software reliability engineers have been able to more-or-less smoothly scale-up servers and applications through automation, noted Thomas S. Hatch, founder and chief technology officer of SaltStack.
But not every layer of the stack is so easily scalable. One relatively new challenge for SaltStack engineers has been with network automation, Hatch pointed out. “One of the bottlenecks that they’ve been running into is network automation. There’s a lot of work that has to be done around those network switches. So that’s one of the areas [where] a lot of our network management components have been really useful for people,” he said.
What are you, or your customers, seeing on the front lines of service fulfillment in the time of COVID-19? Drop us a line at firstname.lastname@example.org.