Pete Zaitcev (zaitcev) wrote,
Pete Zaitcev
zaitcev

Swift in 2021

A developer meet-up for OpenStack, known as PTG, occurred a week ago. I attended the Swift track, where somewhat to my surprise we had two new contributors show up.

I got into a habit of telling people that I did not want Swift to end like AFS: develop great software and dead, with nobody using it. Today I looked it up, and what do you know: OpenAFS made a release in June 2020 (and apparently they also screwed up and had to post an emergency release in October).

So, I was chatting with Matt O. at PTG and he said, "oh yeah, we won some contracts when I was at SuSE, Swift was beating the competition." Not entirely a surprise, but it got me thinking: is it too early to declare Swift dead, or even AFS level dead?

Since NVIDIA gobbled up Swift, I was full of concerns for the centralization. NVIDIA uses Swift as a hyperscaler, in support of their own clusters. They already started to divest themselves from Swiftstack's customer base. I envisioned a future where NVIDIA assembles all the core contributors, then fires them all and closes the project. But then I learned that Lustre went through a cycle like that, being acquired, but then sold out to a smaller, more focused company (to DDN).

To sum, I see a possibility for Swift to remain relevant through a three-step strategy, if you will. First, Swift remains open, aligned to technology, and performant. Thanks to that, it wins new deployments (in HPC and Telco in particular). And because of the field use, it will find a corporate stewardship. So, basically, suck less for success.

P.S. Also at PTG I learned that S3 Inventory existed. Seemed like implementing it in Swift could be a satisfying accomplishment for someone new.

Subscribe

  • Post a new comment

    Error

    Anonymous comments are disabled in this journal

    default userpic

    Your reply will be screened

    Your IP address will be recorded 

  • 0 comments