Sounds neat, but no demo, no video, no evidence at all of it achieving the stated goals. Also they're doing a survey to see how much people would pay per month for what sounds like a single download app (not a service).
Tech-wise, would be interested to hear what this is using. I assume it's using the macOS Hypervisor framework?
They may mean host->container which can have performance limits, at least when running via slirp4netns in user mode due to all the frame copying (who knows if they meet both those requirements on macOS). Newer things like pasta avoid some overheads.
Macs also don't typically have disks that fast, so unless it's transferring out of memory, which won't saturate 30Gbps for long, that also seems to not be that useful.
Latency feels like it would be a better trade-off here, I wonder what they're doing to reduce host to container latency.
> I wonder what they're doing to reduce host to container latency
Considering that any solution they came up with is almost assured to be using Apple’s virtualization frameworks, I’d venture the answer is near zero (at least for work they specifically did).
I feel like the latest Docker Desktop on Ventura works quite well in both amd64 and arm64 containers. This might be with the experimental VM settings on as well as the Rosetta x64 in containers support. As much as I don't love Docker Desktop with the forced login and what not, I can't tell what this is offering over it? Maybe VMs?
Looks interesting. The biggest problem I have with Docker Desktop on MacOS though is the constant cpu draw of 3-5% (or more, I can't tell if "kernel_task" seeming more busy when Docker is running is real). Wondering if we get any improvement there?
I'll vouch for colima. Been using it since I got my M1 Air at launch and it's been stellar.
That said, it still uses ~2.5-3% CPU when idle. This is probably background processes inside the VM and it's not a lot, but it's there. I have had instances where I had completely forgotten I had the VM up and didn't notice any impact on battery life though, so I don't really think it's a big deal. It's also dead easy to just stop it when not in use.
To the GP, increased kernel_task usage (if any) might not be having any impact on battery at all. The main utility for this process, from my experience, is to idle the CPU when it gets hot. This is how the thermal throttle works on macOS.
Tech-wise, would be interested to hear what this is using. I assume it's using the macOS Hypervisor framework?