[cmrdporcupine]

Cantrill is far smarter and accomplished than me, but this article feels a bit strawman and hit and run?

I think unikernels potentially have their place, but as he points, they remain mostly untried, so that's fair. We should analyze why that is.

On performance: I think the kernel makes many general assumptions that some specialized domains may want to short circuit entirely. In particular I am thinking how there's a whole body of research of database buffer pool management basically having elaborate work arounds for kernel virtual memory subsystme page management techniques, and I suspect there's wins there in unikernel world. Same likely goes for inference engines for LLMs.

The Linux kernel is a general purpose utility optimizing for the entire range of "normal things" people do with their Linux machines. It naturally has to make compromises that might impact individual domains.

That and startup times, big world of difference.

Is it going to help people better sling silly old web pages and whatever it is people do with computers conventionally? Yeah, I'd expect not.

On security, I don't think it's unreasonable or pure "security theatre" to go removing an attack surface entirely by simply not having it if you don't need it (no users, no passwords, no filesystem, whatever). I feel like he was a bit dismissive here? That is also the principle behind capability-passing security to some degree.

I would hate to see people close the door on a whole world of potentials based on this kind of summary dismissal. I think people should be encouraged to explore this domain, at least in terms of research.

[ironhaven]

If you software has no bugs then unikernels are a straight upgrade. If your software has bugs then the blast area for issues is now much larger. When was the last time you needed a kernel debugger for a misbehaving application?

[wmf]

A kernel debugger isn't magic; it's just a debugger (e.g. GDB). Debugging a VM is similar to debugging a process.

[wewtyflakes]

Aren't there more places to for things to off the rails? Vibes of https://www.usenix.org/system/files/1311_05-08_mickens.pdf

[mustache_kimono]

> On performance: ... In particular I am thinking how there's a whole body of research of database buffer pool management

Why? The solution thus far has been to turn off what the kernel does, and, do those things in userspace, not move everything in the kernel? Where are these performance gains to be had?

> The Linux kernel is a general purpose utility optimizing for the entire range of "normal things" people do with their Linux machines.

Yeah, like logging and debugging. Perhaps you say: "Oh we just add that logging and debugging to the blob we run". Well isn't that now another thing that can take down the system, when before it was a separate process?

> That and startup times, big world of difference.

Perhaps in this very narrow instance, this is useful, but what is it useful for? Can't Linux or another OS be optimized for this use case without having to throw the baby out with the bathwater? Can't one snapshot a Firecracker VM and reach even faster startup times?

> On security, I don't think it's unreasonable or pure "security theatre" to go removing an attack surface entirely

Isn't perhaps the most serious problem removing any and all protection domains? Like between apps and the kernel and between the apps themselves?

I mean -- sure maybe remove the filesystem, but isn't no memory protection what makes it a unikernel? And, even then, a filesystem is usually a useful abstraction! When have I found myself wanting less filesystem? Usually, I want more -- like ZFS.

This is all just to say -- you're right -- there may be a use case for such systems, but no one has really adequately described what that actually is, and therefore this feels like systems autoeroticism.

[cmrdporcupine]

> Why? The solution thus far has been to turn off what the kernel does, and, do those things in userspace, not move everything in the kernel? Where are these performance gains to be had?

There's all sorts of jankin' about trying to squeeze ounces of performance out of the kernel's page management, specifically for buffer pools.

e.g. https://www.cs.cit.tum.de/fileadmin/w00cfj/dis/_my_direct_up...

Page management isn't really a thing we can do well "in user space". And the kernel has strong ideas about how this stuff works, which work very well in the general case. But a DB (or other system level things like garbage collectors, etc) are special cases, often with special needs.

LeanStore, Umbra, etc. do tricks with VMM overcommit and the like to fiddle around with this, and the above paper even proposes custom kernel modules for the purpose (There's a github repo associated, I'd have to go look).

And then, further, a DB goes and basically implements its own equivalent of a filesystem, managing its own storage. Often fighting with the OS about the semantics of fsync/durability, etc.

I don't think it's an unreasonable mental leap for people to start thinking: "I'm by necessity [cuz cloud] in a VM. Now I'm inside an OS in a VM, and the OS is sometimes getting in my way, and I'm doing things to get around the OS... Why?"

[mustache_kimono]

> Page management isn't really a thing we can do well "in user space".

But it is the thing most high performance OLTP DBMSs, most of us are aware of, do? I'm also not sure your cite is relevant here. Or it is at least niche. The comparison is made to LeanStore, which is AFAICT is not feature complete, and a research prototype?

Your cite does not describe a unikernel use case, but instead in kernel helper modules. Your cite is about leveraging in kernel virtual memory for the DB buffer cache, and thus one wonders how sophisticated the VM subsystem is in most unikernels? That is -- how is this argument for unikernels? Seems as though your cite is making the opposite argument -- for a more complex relationship between the DB and the kernel, not a pared down one.

> And then, further, a DB goes and basically implements its own equivalent of a filesystem, managing its own storage. Often fighting with the OS about the semantics of fsync/durability, etc.

The fights you're describing what have thus far been the problems of ceding control of the buffer cache to the kernel, via mmap, especially re: transactional safety.

If your argument is kernels may need a bottom up redesign to make ideas like this work, I suppose that makes sense. However, again, I'm not sure that makes unikernels more of an answer here than anywhere else, though.

> I don't think it's an unreasonable mental leap for people to start thinking: "I'm by necessity [cuz cloud] in a VM. Now I'm inside an OS in a VM, and the OS is sometimes getting in my way, and I'm doing things to get around the OS... Why?"

I think that's a fair thought to have, but the problem is how it actually works in practice. As in, less code seems really enticing, the problem is what abstractions are you throwing away. If the abstraction is less memory protection, maybe this is not a good tradeoff in practice.