| === shokohsc51088 is now known as shokohsc5108 | ||
| === oerheks is now known as P01135809 | ||
| === P01135809 is now known as oerheks | ||
| === shokohsc51085 is now known as shokohsc5108 | ||
| faekjarz | How would i benchmark SWAP memory space? I want at least avg throughput, perhaps IOPS. I'm sure, i'm not the 1st to attempt this - there must be a ready-to-use tool/procedure, right? ;) | 15:38 |
|---|---|---|
| faekjarz | …can i mount my swap file as a block dev, and DD it? …that'd do it, i think | 15:40 |
| === alchemist__ is now known as alchemist | ||
| znf | faekjarz, why would you want to do that? | 18:35 |
| faekjarz | @znf: because i'd like to benchmark my SWAP mem | 18:59 |
| znf | It's swap, not memory | 18:59 |
| faekjarz | …in my particular case, i've got a dm-crypt layer, in between NVMe Storage, and the actual SWAP FILE | 19:01 |
| faekjarz | …i'd like to benchmar it | 19:01 |
| faekjarz | ← is perfectly aware of all the layers - the dude just wants to benchmark his SWAP | 19:03 |
| faekjarz | there must be a "swap mem" benchmark tool! Logic dictates, i'm NOT the 1st to attempt such an endeavour | 19:05 |
| znf | There is not | 19:08 |
| faekjarz | perhaps, we can StarFleet this issue ;) <3 | 19:08 |
| faekjarz | you sure about that, @znf? …i might just inquire Linus himself …nuffing rong with that … LUL | 19:10 |
| faekjarz | I'd like a SWAP benchmarking tool! -.- | 19:11 |
| znf | I'm very sure | 19:11 |
| faekjarz | k | 19:11 |
| znf | Also Linus hasn't touched those parts of the kernel in years | 19:11 |
| znf | There's other people that are much more appropriate to ask about swap performance that work on the Linux kernel these days | 19:12 |
| znf | Either way, if you just want to test encrypted VS unencrypted performance, just test that directly | 19:13 |
| znf | Also depending în what you're testing, I can tell you that a swap file is better than a dedicated swap partition these days | 19:14 |
| === Adri2000_ is now known as Adri2000 | ||
| faekjarz | thank you, @znf, technically, i just want to "benchmark" my SWAP (file) - i want throughput, perhaps IOPS. I wonder, can i mount it as a block dev, and get DD results? | 19:18 |
| faekjarz | …following the everything-is-a-file doctrine, i should be able to mount and DD-benchmark my swap thing, shouldn't i? …how would i do that? | 19:24 |
| sarnold | faekjarz: https://fio.readthedocs.io/en/latest/fio_doc.html#cmdoption-arg-filename | 19:58 |
| faekjarz | @sarnold: i'm not quite sure, how that contributes to benchmarking THE SWAP. Perhaps, if you don't mind, please elaborate, how i'd tune my Ubuntu boo params, accordingly. I'm certainly not a "please-bro", i'm happy to RTFM… why is your fedora example applicaple to my Ubuntu? | 20:19 |
| sarnold | faekjarz: fio is the most commonly used block IO benchmarking tool. it has a command line option to let you specify a file; I don't know if it can use your existing swap file, or if you'll just need to give it a path to where your swapfile is located, but that shouldn't give results too different from the existing swapfile | 20:22 |
| faekjarz | @sarnold: hmm, yeah, a file is a file - everything is a file - why wouldn't SWAP be A FILE …(loop?)mountable as a block dev? …the Kernel MUST grant me the power to write directly to SWAP, right? Do i need to run custom C-code? | 20:57 |
| sarnold | faekjarz: you could use losetup to mount it as a block device if you really must | 21:29 |
| sarnold | faekjarz: I'm hopeful that fio's --filename argument is enough to say that you don't need to :) | 21:29 |
| znf | still wasted time | 21:37 |
| znf | just get more ram | 21:37 |
| sarnold | yes | 21:43 |
| faekjarz | oh, of course - silly me - i could've just downloaded more RAM! LUL | 21:45 |
| faekjarz | MAKE MY DAY! | 21:45 |
| faekjarz | <3 | 21:45 |
| faekjarz | well, while we're all so intimate - my use case is: swap file on a dm-crypt volume. CPU has more AES 512b prowess than PCIe/NVMe Gen3 x4 bandwidth allows. | 21:48 |
| faekjarz | i still like to obtain actual, real-world, benchmarkable, performance figures. …i want to stress my SWAP …HOW TO? …anyone, bueller, bueller | 21:51 |
| faekjarz | ;) | 21:52 |
| sarnold | faekjarz: what errors did you get with fio? | 21:55 |
| faekjarz | …i'm certainly not afraid of admitting insufficient education - this is NOT the barrier we're talking about! I intend to do SCIENCE, here! …I request your assistance! | 21:56 |
| faekjarz | i did not run fio, yet, @sarnold | 21:57 |
| sarnold | ah :) you should at least try the thing I suggested you to try two hours ago. You might have already had your answer had you tried. :) | 21:57 |
| faekjarz | ok, i'm sorry - i apologize - it's the local booze day (i'm a bit tipsy) - i might not magnificise at the extremity of my potential, today! I shall increase my strive towards perfection. (i hope, you'd enjoy this overly grande ductus) | 22:09 |
| sarnold | ah, happy imbibing :) | 22:13 |
| faekjarz | imbibi-what? …my phasers are set to inebriate …yet, i still attempt to figure out complicasted stith …oof, i shouldn't be bothering with …uhm … wasting yourt 'ime | 22:18 |
| faekjarz | …sorry 'bout that | 22:19 |
| znf | You're still wasting your time. You're eventually just benchmarking the file system, at best | 22:29 |
| znf | Nobody made a swap benchmark because it's a stupid and unreliable test | 22:30 |
| znf | put it this way: if nobody *probably* smarter than us ever bothered making a benchmark... | 22:35 |
| sarnold | yeah the whole point of swap is that it should be rarely used | 22:35 |
| sarnold | stuff that's not important gets shoved there, so memory can be used for something that is used | 22:35 |
| znf | and because this is the *-server channel, heck, there can be an argument to never really having swap enabled anyway | 22:36 |
| znf | the most I do on servers lately is to have zram | 22:37 |
| znf | on the server that I do have swap, it's like 1-2 megabytes used | 22:37 |
| znf | looking at a 128GB server, with 8GB swap, I'm using 0 | 22:37 |
| znf | on a 131 day uptime | 22:37 |
| === bbezak1 is now known as bbezak | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!