Microsoft Corp.’s researchers have lifted the lid on a new “cloud-scale” emulator they’ve built that’s designed to completely eliminate outages in its public cloud.

The emulator is called CrystalNet, and it was one of the main subjects of discussion at the ACM Annual Symposium on Operating Systems Principles 2017 event that’s taking place this week in Shanghai, China.

Victor Bahl, a distinguished scientist director with Microsoft’s Mobile & Networking Research team, said CrystalNet was the fruit of more than two years’ research and development efforts. He explained in a blog post that CrystalNet was designed in the belief that it should be possible to spot cloud network problems before they occur by taking a more proactive approach. Bahl and his colleagues quickly decided that the best way to do this was to emulate the network in order to test it.

“The idea of testing before deploying is age old, but following a two-year study by Microsoft Research looking at all documented outages across all major cloud providers, we believed that we could find most potential problems if we first validated a production network on an identical copy of the network,” Bahl said.

CrystalNet is identical because it uses the same network topology, hardware, software and configurations as Microsoft’s production network. It can be run on emulated devices inside virtual machines and can support the software images of most common networking devices.

Microsoft offered the following diagram that lays out CrystalNet’s architecture:

The whole point of having an emulated network such as CrystalNet is that Microsoft’s Azure engineers can use it to test planned changes and updates for any issues, rather than just implementing them and hoping that nothing breaks, something that happens with alarming regularity among the major public cloud providers. Bahl said that by using CrystalNet, Microsoft’s engineers can “reduce the significance of network issues” that sometimes crop up after updates have been made.

Early results look good. Bahl said that Microsoft’s Azure team has already used CrystalNet to validate a number of new network designs, architectural changes, hardware and firmware updates and network configuration changes. The company has also used CrystalNet to successfully build and test its Software for Open Networking in the Cloud, which is an operating system for network switches.

Despite the success of these tests, one expert said CrystalNet may not be able to prevent every kind of outage. Holger Mueller, vice president and principal analyst at Constellation Research Inc., said that although it’s nice to see Microsoft trying to address the network issues that have been responsible for several previous Azure cloud outages, there are many other aspects to consider when trying to prevent downtime.

“Being able to simulate and test before deploying is the right way to ensure uptime, but a test remains a test,” Mueller said. “Being able to test locally and roll back instantly are the key aspects to ensuring uptime. But rollback capabilities remain something that most infrastructure as a service providers need to work on, especially when it comes to network infrastructure updates.”

Bahl didn’t say anything about when this promising new technology might be made commercially available. But he did acknowledge that some Azure customers have already expressed an interest in using CrystalNet to prevent outages in their own networks. Most likely CrystalNet will be integrated with the Azure Cloud first, before the company starts thinking about selling it to others.

Image: noupload/pixabay