What does NoOps mean for infrastructure?
Gore said the NoOps model requires an IT infrastructure designed for scale but also for failure.
Design the system the right way, and “failure is OK,” he said.
“If you get the model right, it's OK if systems fail, it's OK if my storage fails, it's OK if my network fails, because I actually designed for that. It's distributed, and runs active-active-active-active -ctive, none of this active-passive or hot-cold patterns.
"It's all active — transactions happen at multiple places at the same time. It should involve auto-scaling, auto-healing, it should be highly dynamic and elastic.”
Further, the 'where' question becomes increasingly important. Internal systems will need to be deployed closer to public clouds, to reduce latency and lower network transit costs. Running systems in active-active configurations creates more “chatter” on the network, he said, which again requires faster and cheaper network transit.
Gore also predicted a shift away from monolithic storage systems to distributed storage solutions.
“The next evolution of that is deploying servers with local attached storage and then running a global file system or a distributed file system over the top, using replication and archiving across that to manage the data load. What’s interesting about this model is you can add capacity both in terms of volume but also performance just by adding and subtracting spindles.”
The IT department will maintain relevance, he said, by focusing on systems development and the “services a public cloud provider doesn’t provide,” such as security (intrusion detection and prevention, physical firewalls, encryption devices) plus backup and recovery.
-1.jpg&w=100&c=1&s=0)
.png&w=100&c=1&s=0)
