Open Sourcing and Community Involvement: While the current focus is on product development, Antithesis is interested in open-sourcing parts of their work to contribute back to the community and foster collaboration.
That is very interesting. I wonder what parts of their tech stack could feasibly be open-sourced without undermining their business model. Of course they could do the standard FOSS thing of transitioning to a consulting shop around this product, a sort of industrialized aphyr/jepson combo, but I doubt that would be more lucrative than a SaaS model for a technology that is currently without any real competitors.
Browsing their docs, it looks like Antithesis actually tries to do quite a bit more than “just” providing a deterministic hypervisor. E.g. it includes features such as network/clock fault injection, a time-travel debugger, it’s got some instrumentation features… Deterministic hypervised execution, at this scale and with this sort of integration (I mean, it literally runs container images) is the neat foundation that understandably gets more attention in a FreeBSD foundation case study, but I think the tech they build on top of it might be even more interesting.
The docs sound somewhat optimistic at times, and they’re a little CTA-ish at times, so I should point out that I’m not affiliated with them in any way. I literally just read about this and have no idea how much of it works and how well.
I love that this is a real product!
Sure, I’d be happier if it were in the public domain and I could run it on my own hardware, but just that it exists is freaking awesome. Ever since I first saw rr I fell in love with the idea of time-travel debugging, but rr starts running into problems once you add I/O and multi-threading into the mix. I came to the conclusion that both of those limitations could be solved with a deterministic VM, but it felt like a daunting task to implement, so I’m really excited now that someone put the work in to make it a real product. Imagine easily observable and replicable race conditions!
That is very interesting. I wonder what parts of their tech stack could feasibly be open-sourced without undermining their business model. Of course they could do the standard FOSS thing of transitioning to a consulting shop around this product, a sort of industrialized aphyr/jepson combo, but I doubt that would be more lucrative than a SaaS model for a technology that is currently without any real competitors.
Browsing their docs, it looks like Antithesis actually tries to do quite a bit more than “just” providing a deterministic hypervisor. E.g. it includes features such as network/clock fault injection, a time-travel debugger, it’s got some instrumentation features… Deterministic hypervised execution, at this scale and with this sort of integration (I mean, it literally runs container images) is the neat foundation that understandably gets more attention in a FreeBSD foundation case study, but I think the tech they build on top of it might be even more interesting.
The docs sound somewhat optimistic at times, and they’re a little CTA-ish at times, so I should point out that I’m not affiliated with them in any way. I literally just read about this and have no idea how much of it works and how well.
I love that this is a real product! Sure, I’d be happier if it were in the public domain and I could run it on my own hardware, but just that it exists is freaking awesome. Ever since I first saw
rrI fell in love with the idea of time-travel debugging, butrrstarts running into problems once you add I/O and multi-threading into the mix. I came to the conclusion that both of those limitations could be solved with a deterministic VM, but it felt like a daunting task to implement, so I’m really excited now that someone put the work in to make it a real product. Imagine easily observable and replicable race conditions!