Disc IO 512mb vs The World

general

forum:fiat 13 years, 6 months ago

Ran Unixbench on a 512mb Linode and then again on a 1,2,4Gb Linode. All the boxes ran identical processors, and scored identical results except for Disc IO, where the 512 kicked arse by a large margin..

512mb Linode

File Copy 1024 bufsize 2000 maxblocks 323047.5 KBps (30.0 s, 2 samples)

File Copy 256 bufsize 500 maxblocks 84232.5 KBps (30.0 s, 2 samples)

File Copy 4096 bufsize 8000 maxblocks 864676.5 KBps (30.0 s, 2 samples)

Pipe Throughput 449392.8 lps (10.0 s, 7 samples)

1,2,4Gb Linode

File Copy 1024 bufsize 2000 maxblocks 76893.8 KBps (30.0 s, 2 samples)

File Copy 256 bufsize 500 maxblocks 19415.0 KBps (30.0 s, 2 samples)

File Copy 4096 bufsize 8000 maxblocks 276594.9 KBps (30.0 s, 2 samples)

Pipe Throughput 86488.9 lps (10.0 s, 7 samples)

At first I thought thought this was a mistake so i ran this on another 512 and I got very similar results. I'm wondering what the hardware difference exactly is between 512mb linodes and their bigger cousins(?SSD in 512?).. Obviously more nodes on a box puts a lot of strain on IO, but I was really surprised by the margin.

Am I missing something here?

8 Replies

forum:AceStar 13 years, 6 months ago

I've always had a bit of a suspicion - that there will be the least contention for resources (ie IO) on the smallest plan, because the majority of people on this plan will have next to no traffic, since this is the 'catch all' for anyone for whom IO and CPU etc are no problem (dev boxes, small hobby boxes etc). Whereas anybody on any other plan, including the second smallest plan, is likely to be making good use of their resources.

forum:obs 13 years, 6 months ago

You also have to consider that your 512 might be on a relatively empty host. Try knocking up another 512 in a different DC and see if you get consistent results.

forum:Mr Nod 13 years, 6 months ago

~~@AceStar:~~

I've always had a bit of a suspicion - that there will be the least contention for resources (ie IO) on the smallest plan, because the majority of people on this plan will have next to no traffic, since this is the 'catch all' for anyone for whom IO and CPU etc are no problem (dev boxes, small hobby boxes etc). Whereas anybody on any other plan, including the second smallest plan, is likely to be making good use of their resources.

I have to agree with this. Last night I migrated from a 512 to a 768 plan and noticed that the CPU usage on the 768 plan was higher than the 512 one.

Makes sense really when you explain it that way AceStar.

forum:Guspaz 13 years, 6 months ago

The hardware is identical, and you have access to the same theoretical maximum CPU/bandwidth/disk IO. The only difference between plans is how many nodes they stick on a machine. There's a fixed amount of RAM on the hosts, and linode plans are not mixed and matched (one type of linode per host). This means that buying a bigger linode gets you a bigger share (and there are less other linodes) of the host.

On a box with low contention, this probably means no difference except in RAM. On a box with high contention, this means you'll get a bigger guaranteed share of CPU time and higher IO priority.

Personally, I've always felt that it's probably more productive to scale horizontally than vertically at Linode, as long as your bottleneck isn't RAM or bandwidth. Jed disagrees with me, though ;)