1 |
On Wednesday, August 10, 2016 10:13:29 AM james wrote: |
2 |
> On 08/10/2016 07:45 AM, Michael Mol wrote: |
3 |
> > On Tuesday, August 09, 2016 05:22:22 PM james wrote: |
4 |
|
5 |
> >> |
6 |
> >> I did a quick test with games-arcade/xgalaga. It's an old, quirky game |
7 |
> >> with sporadic lag variations. On a workstation with 32G ram and (8) 4GHz |
8 |
> >> 64bit cores, very lightly loaded, there is no reason for in game lag. |
9 |
> >> Your previous settings made it much better and quicker the vast majority |
10 |
> >> of the time; but not optimal (always responsive). Experiences tell me if |
11 |
> >> I can tweak a system so that that game stays responsive whilst the |
12 |
> >> application(s) mix is concurrently running then the quick |
13 |
> >> test+parameter settings is reasonably well behaved. So thats becomes a |
14 |
> >> baseline for further automated tests and fine tuning for a system under |
15 |
> >> study. |
16 |
> > |
17 |
> > What kind of storage are you running on? What filesystem? If you're still |
18 |
> > hitting swap, are you using a swap file or a swap partition? |
19 |
> |
20 |
> The system I mostly referenced, rarely hits swap in days of uptime. It's |
21 |
> the keyboard latency, while playing the game, that I try to tune away, |
22 |
> while other codes are running. I try very hard to keep codes from |
23 |
> swapping out, cause ultimately I'm most interested in clusters that keep |
24 |
> everything running (in memory). AkA ultimate utilization of Apache-Spark |
25 |
> and other "in-memory" techniques. |
26 |
|
27 |
Gotcha. dirty_bytes and dirty_background_bytes won't apply to anything that |
28 |
doesn't call mmap() with a file backing or perform some other file I/O. If |
29 |
you're not doing those things, they should have little to no impact. |
30 |
|
31 |
Ideal values for dirty_bytes and dirty_background_bytes will depend heavily on |
32 |
the nature of your underlying storage. Dozens of other things might be tweaked |
33 |
depending on what filesystem you're using. Which is why I was asking about |
34 |
those things. |
35 |
|
36 |
> |
37 |
> |
38 |
> Combined codes running simultaneously never hits the HD (no swappiness) |
39 |
> but still there is keyboard lag. |
40 |
|
41 |
Where are you measuring this lag? How much lag are we talking about? |
42 |
|
43 |
> Not that it is actually affecting the |
44 |
> running codes to any appreciable degree, but it is a test I run so that |
45 |
> the cluster nodes will benefit from still being (low latency) quickly |
46 |
> attentive to interactions with the cluster master processes, regardless |
47 |
> of workloads on the nodes. Sure its not totally accurate, but so far |
48 |
> this semantical approach, is pretty darn close. It's not part of this |
49 |
> conversation (on VM etc) but ultimately getting this right solves one of |
50 |
> the biggest problems for building any cluster; that is workload |
51 |
> invocation, shedding and management to optimize resource utilization, |
52 |
> regardless of the orchestration(s) used to manage the nodes. Swapping to |
53 |
> disc is verbotim, in my (ultimate) goals and target scenarios. |
54 |
> |
55 |
> No worries, you have given me enough info and ideas to move forward with |
56 |
> testing and tuning. I'm going to evolve these into more precisely |
57 |
> controlled and monitored experiments, noting exact hardware differences; |
58 |
> that should complete the tuning of the Memory Management tasks, within |
59 |
> acceptable confine . Then automate it for later checking on cluster |
60 |
> test runs with various hardware setups. Eventually these test will be |
61 |
> extended to a variety of memory and storage hardware, once the |
62 |
> techniques are automated. No worries, I now have enough ideas and |
63 |
> details (thanks to you) to move forward. |
64 |
|
65 |
You've got me curious, now you're going to go run off and play with your |
66 |
thought problems and not share! Tease! |
67 |
|
68 |
> |
69 |
> >> Perhaps Zabbix +TSdB can get me further down the pathway. Time |
70 |
> >> sequenced and analyzed data is over kill for this (xgalaga) test, but |
71 |
> >> those coalesced test-vectors will be most useful for me as I seek a |
72 |
> >> gentoo centric pathway for low latency clusters (on bare metal). |
73 |
> > |
74 |
> > If you're looking to avoid Zabbix interfering with your performance, |
75 |
> > you'll |
76 |
> > want the Zabbix server and web interface on a machine separate from the |
77 |
> > machines you're trying to optimize. |
78 |
> |
79 |
> agreed. |
80 |
> |
81 |
> Thanks Mike, |
82 |
> James |
83 |
|
84 |
np |
85 |
-- |
86 |
:wq |