Axboe Achieves 8M IOPS Per-Core With Newest Linux Optimization Patches

[phoronix]

Phoronix: Axboe Achieves 8M IOPS Per-Core With Newest Linux Optimization Patches

It was just last week that Linux optimizations were leading to possible 6M IOPS per core and then at the start of this week new patches pushed Linux past 7M IOPS per-core with an ideal hardware configuration as well. In ending out the week, 8M IOPS has been reached!..

https://www.phoronix.com/scan.php?page=news_item&px=8M-IOPS-Per-Core-Linux

[perpetually high]

Awesome work, Jens.

[Danny3]

That's incredible, congratulations !
Hopefully some of these optimizations will be helpful for normal consumer storage too.

[pal666]

Hopefully some of these optimizations will be helpful for normal consumer storage too.

they'll reduce cpu usage

[blackshard]

I'm still puzzled of how can this matter can be a news...
I mean, measuring throughput of an algorithm/api/whatever is sensible only when all the variables around stay the same.

Here I see: Guy 1 achieves xxxx IOPS, Guy 2 achieves yyyy IOPS, Guy 3 achieves zzzz IOPS... ok but who cares?
Those guys are all on different hardware and there is no comparison, so that's just pure numbers that actually tell nothing about the optimizations they do to the api...

Paradoxically some kind of optimization for a setup may result in performance loss for another, where are the comparisons?

[Erwan]

I'm still puzzled of how can this matter can be a news...
I mean, measuring throughput of an algorithm/api/whatever is sensible only when all the variables around stay the same.

Here I see: Guy 1 achieves xxxx IOPS, Guy 2 achieves yyyy IOPS, Guy 3 achieves zzzz IOPS... ok but who cares?
Those guys are all on different hardware and there is no comparison, so that's just pure numbers that actually tell nothing about the optimizations they do to the api...

Paradoxically some kind of optimization for a setup may result in performance loss for another, where are the comparisons?

I think you perfectly miss the point here.

What Jens is doing here, is for a defined setup, how much IOPS we can expect from a single core.
Not to mimic what an application does but to find what are the bottlenecks of the current implementation.

To clarify, IOPS are sensitive to the CPU clock speed, the device itself and the source code between the two.
Once you get a faster processor, some high-end NVME drives, you discover bottlenecks you never saw before.
Jens did an amazing job inspecting all the IO code path to ensure it is optimized at most.

What an end-user can expect from this :

• running the same hardware with updated software to increase its performance level or run the same perf with less CPU overhead.
• upgrading their platform and reach such a high number of IOPS/core, that wasn't possible before

So turning from 6M to 8M while running a single core in a month, yes that's a news.

[darkbasic]

It's always the same guy.

[George99]

Not only the same guy but also same hardware. Otherwise it would be pointless.

[MastaG]

But could this also have a positive effect for a regular desktop user running a webbrowser and playing some games on Steam for example?

[RedEyed]

But could this also have a positive effect for a regular desktop user running a webbrowser and playing some games on Steam for example?

Potentially, any kind of optimisation is good for regular user.

Even if it is not faster, it will use less power.