1 |
On Thu, Aug 11, 2011 at 15:06, Nikos Chantziaras <realnc@×××××.de> wrote: |
2 |
> On 08/11/2011 10:35 AM, Pandu Poluan wrote: |
3 |
>> |
4 |
>> On Thu, Aug 11, 2011 at 14:27, Nikos Chantziaras<realnc@×××××.de> wrote: |
5 |
>>> |
6 |
>>> On 08/11/2011 09:56 AM, Pandu Poluan wrote: |
7 |
>>>> |
8 |
>>>> Just wondering, is it possible to specify a custom timer frequency? |
9 |
>>>> E.g., HZ=500 instead of one of the canned values (100, 250, 300, |
10 |
>>>> 1000). |
11 |
>>> |
12 |
>>> It is possible, but it's a bad idea because non-standard values can |
13 |
>>> result |
14 |
>>> in driver breakage. Some code assumes specific timer granularities |
15 |
>>> (100Hz = |
16 |
>>> 10ms, 250Hz = 4ms, etc). This usually happens with values above 1000Hz, |
17 |
>>> so |
18 |
>>> it might be possible to experiment with non-standard sub-1000Hz values. |
19 |
>>> |
20 |
>>> But why do you want a custom value anyway? |
21 |
>>> |
22 |
>> |
23 |
>> Well, for a firewall, I've calculated (gathered and extrapolated from |
24 |
>> a lot of sources), the latency per-packet is usually less than 1 ms, |
25 |
>> at worst still less than 2 ms. |
26 |
>> |
27 |
>> Thus, setting the timer freq to 100 Hz (as suggested for 'normal' |
28 |
>> server load) means the timeslice is way too long (10 ms per time |
29 |
>> slice). |
30 |
>> |
31 |
>> So, I speculate that better -- and uniform -- performance will be |
32 |
>> achieved with a timer freq of 500 Hz. |
33 |
>> |
34 |
>> Of course, this is a wild speculation/guess from me. I never quite |
35 |
>> understand netfilter/xtables' relation to the timeslices, so I may be |
36 |
>> talking nonsense :-) |
37 |
> |
38 |
> This assumes that networking is dependent on the timer interrupt, which |
39 |
> doesn't seem to be the case; going from 100Hz to 1000Hz will not result in |
40 |
> network latencies dropping by 9ms. I know because I tried it on a server. |
41 |
> |
42 |
|
43 |
That's what I was concerned of. |
44 |
|
45 |
> The only case where network latency improved with a higher HZ was with a |
46 |
> game server (Counter-Strike) and that's only because that game's server |
47 |
> component is timer dependent when calculating the game world (to do 1000 |
48 |
> updates per second it needs a HZ value of 1000). So that application has |
49 |
> some real-time demands, meaning a high HZ value helps. Otherwise, you can |
50 |
> stick to 100Hz on a server. Higher values won't change anything and will |
51 |
> only reduce throughput (though not by much anyway, which is why some people |
52 |
> set 1000Hz even on servers.) |
53 |
> |
54 |
|
55 |
Hmmm... I *do* feel better response (interaction-wise via SSH) if I |
56 |
use >100, so I think I'll settle for 250 Hz. |
57 |
|
58 |
Thanks for the explanation! |
59 |
|
60 |
Rgds, |
61 |
-- |
62 |
FdS Pandu E Poluan |
63 |
~ IT Optimizer ~ |
64 |
|
65 |
• Blog : http://pepoluan.tumblr.com |
66 |
• Linked-In : http://id.linkedin.com/in/pepoluan |