There are a number of choices for working in the cloud, aka, renting resources, but there are three big choices: Azure, AWS, and GCP. These three dominate the cloud services, though arguably AWS and Azure are way ahead. All three have lots of services, and all three can likely meet your cloud or hybrid needs, even if you work in a regulated industry or government.
This week there was a piece on the work Microsoft is doing to increase their capacity. Between the 0365 Office services, Teams for collaboration, and their regular PaaS/IaaS workloads, they’ve grown quite a bit this year with the COVID-19 pandemic. As they try to increase their resources, they’ve had people working round-the-clock to add new servers. I’m sure they are still using containers, but likely they are struggling to fill these with new hardware.
They’ve also been adding bandwidth, which is a bit need between the data centers. I had no idea they have their own undersea cables, but apparently they do. I suspect Amazon and Google do as well. There is a note that they’re also coding, trying to improve efficiency. After all, while you can throw hardware at a problem, many of us know that bad code can overwhelm those efforts. We see it in our jobs every day. I would hope that many of our organizations that might be depending on software more realize that better code is an advantage for us and our customers.
I know Microsoft runs a lot of their own stuff on Azure, including development work, and they’ve had to reschedule their internal workloads as well as move other work outside of regions. I think that’s fine in many cases, but with different laws in different countries, I hope they’re extremely careful about what moves. Data sovereignty can matter.
If you read through the article, it seems that Microsoft has learned a lot about adapting their work methods, both with people and systems, to meet the increased workloads. This is one area that I hope we see some detailed technical blogs, which might help others learn how to better manage their own resource challenges, or even learn to code better. While some of this might be proprietary, with the move to having many things open source, why not share ways to improve application resource usage.
Steve Jones
