I/O Is Faster Than CPU – Let’s Partition Resources and Eliminate OS Abstractions [pdf]
1–10 of 291 posts
Re: I/O Is Faster Than CPU – Let’s Partition Resources and Eliminate OS Abstractions [pdf]
#2Very interesting shift that happened over the last 2 decades.
We likely haven't designed OSes or CPUs to match this new reality.
Re: I/O Is Faster Than CPU – Let’s Partition Resources and Eliminate OS Abstractions [pdf]
#3Does this mean we go from virtual machines (i.e. VMWare) to kubernetes & containers in data centers?
Something similar to RanchOS
Re: I/O Is Faster Than CPU – Let’s Partition Resources and Eliminate OS Abstractions [pdf]
#4Does this mean we go from virtual machines (i.e. VMWare) to kubernetes & containers in data centers? Something similar to RanchOS
Even VMWare has support for running containers on their platform now, so yeah probably.
Cisco Nexus switches can run Docker workloads also.
Re: I/O Is Faster Than CPU – Let’s Partition Resources and Eliminate OS Abstractions [pdf]
#5Is this true for most real life workloads? There's that famous rule-of-thumb indicator for latencies: https://www.prowesscorp.com/computer-latency-at-a-human-scal...
It doesn't seem to be that the orders of magnitude are so close as to require totally rethinking mainstream kernels.
Or am I looking at this the wrong way?
Re: I/O Is Faster Than CPU – Let’s Partition Resources and Eliminate OS Abstractions [pdf]
#6Is this true for most real life workloads? There's that famous rule-of-thumb indicator for latencies: https://www.prowesscorp.com/computer-latency-at-a-human-scal... It doesn't seem to be that the orders of magnitude are so close as to require totally rethinking mainstream kernels. Or am I looking at this the wrong way?
> Is this true for most real life workloads?
No. It's true if you care about NVMe drives, or high-speed networking--which is to say, it's true if you care about a few kinds of server workloads, but it's absolutely not true for most consumer hardware.
Re: I/O Is Faster Than CPU – Let’s Partition Resources and Eliminate OS Abstractions [pdf]
#7Does this mean we go from virtual machines (i.e. VMWare) to kubernetes & containers in data centers? Something similar to RanchOS
It means you go from an OS that provides complex drivers to one that barely provides drivers -- essentially muxing the raw hardware between processes.
Re: I/O Is Faster Than CPU – Let’s Partition Resources and Eliminate OS Abstractions [pdf]
#8Is this true for most real life workloads? There's that famous rule-of-thumb indicator for latencies: https://www.prowesscorp.com/computer-latency-at-a-human-scal... It doesn't seem to be that the orders of magnitude are so close as to require totally rethinking mainstream kernels. Or am I looking at this the wrong way?
> Is this true for most real life workloads? No. It's true if you care about NVMe drives, or high-speed networking--which is to say, it's true if you care about a few kinds of server workloads, but it's absolutely not true for most consumer hardware.
NVMe drives with transfer rates of several gigabytes per second are mainstream now.
Re: I/O Is Faster Than CPU – Let’s Partition Resources and Eliminate OS Abstractions [pdf]
#9Very interesting shift that happened over the last 2 decades. We likely haven't designed OSes or CPUs to match this new reality.
Largely because it's not really a new reality. IBM faced the same issues on the 360's half a decade (edit: sorry, century!) ago -- you could stream data off of stacked platters in a drive into core much faster than a CPU could manage the copy. And the solution was to invent "I/O Channels", which were early DMA controllers. And the VM layer (when it was added) was cognizant of this stuff, so applications could be written directly to the channel interface.
There's nothing new under the sun, basically. It's an Ecclesiastes design. I haven't read through the whole article, but my guess is that the "parakernel" interface the authors are positing is going to look a lot like the IBM Channel interface.
Re: I/O Is Faster Than CPU – Let’s Partition Resources and Eliminate OS Abstractions [pdf]
#10Is this true for most real life workloads? There's that famous rule-of-thumb indicator for latencies: https://www.prowesscorp.com/computer-latency-at-a-human-scal... It doesn't seem to be that the orders of magnitude are so close as to require totally rethinking mainstream kernels. Or am I looking at this the wrong way?
A quick google found the following link. It makes it look like 100gb IB can be about 1 order of magnitude faster than Ethernet - around or less than 1 us. But IB can do RDMA, so that further reduces latency.
http://sc16.supercomputing.org/sc-archive/tech_poster/poster...