1 |
On Fri, Dec 14, 2012 at 10:44 AM, Grant <emailgrant@×××××.com> wrote: |
2 |
>> > So if I have 2 physical CPU's with 4 cores each and I enable SMP, I'm |
3 |
>> > using |
4 |
>> > 8 cores? Can NUMA be either enabled or disabled when using more than |
5 |
>> > one |
6 |
>> > physical CPU, or is it required? |
7 |
>> |
8 |
>> |
9 |
>> NUMA is a hardware architecture. It's how you access memory on a |
10 |
>> hardware level: NUMA = Non Uniform Memory Access vs a UMA architecture |
11 |
>> of typical (old/legacy) SMP systems (UMA = Uniform Memory Access). |
12 |
>> |
13 |
>> In a UMA system, all the memory belongs to all the sockets. In a NUMA |
14 |
>> system, each socket has it's "own" local memory. In modern (x86-64) |
15 |
>> processors, each socket has it's own memory controller so each socket |
16 |
>> controls its own local memory. If one socket runs out of memory it can |
17 |
>> ask another socket to lend him some memory. In a UMA system, no socket |
18 |
>> has to ask since memory is global and belongs to all sockets so if one |
19 |
>> socket uses up all the memory ... the rest "starve". In NUMA, there's |
20 |
>> more control over who uses what (be it cores or RAM). |
21 |
>> |
22 |
>> If you have a modern dual or quad (or higher #) socket system ... |
23 |
>> you've got NUMA architecture and you can't get rid of it, it's |
24 |
>> hardware, not software. |
25 |
> |
26 |
> So I must enable CONFIG_NUMA for more than one physical CPU, and disable it |
27 |
> for only one physical CPU? |
28 |
|
29 |
|
30 |
Yup. But ... Why would you want to disable a socket (CPU)? If you |
31 |
disable a socket (CPU) ... you lose the memory attached to that socket |
32 |
(CPU) not to mention you lose those cores ;) |
33 |
|
34 |
A better solution would be to use cgroups or numactl tools to pin a |
35 |
certain process to a set of cores and a memory region. |
36 |
|
37 |
If you really want to deactivate cores (but not the whole socket), you can type: |
38 |
|
39 |
echo 0 > /sys/devices/system/cpu/cpu1/online |
40 |
|
41 |
This would deactivate core #1. You can deactivate as many cores as you |
42 |
wish, except for core #0. |
43 |
|
44 |
This can be done without rebooting your server (aka during run time). |
45 |
Your memory will not be affected, but you will have less cores (and |
46 |
theoretically more memory bandwidth). I say "theoretically" because |
47 |
you always have to benchmark these things with YOUR application |
48 |
(remember logic NEVER applies to real life ;) |
49 |
|
50 |
If you want to check the # of cores you've got: |
51 |
|
52 |
cat /proc/interrupts | grep CPU |
53 |
|
54 |
Other possibilities such as cat /proc/cpuinfo or dmesg, ... can be |
55 |
useful too for this: your choice, FLOSS gives you options. |
56 |
|
57 |
If you want to activate the previously deactivated core, you can run: |
58 |
|
59 |
echo 1 > /sys/devices/system/cpu/cpu1/online |
60 |
|
61 |
Now ... be sure your core numbering is the expected core numbering. |
62 |
IOW, not all server vendors follow the same numbering scheme so core |
63 |
#1 in vendor A's server could be core #2 in vendor B's server. Never |
64 |
trust logic ;) |
65 |
|
66 |
As I mentioned previously: test/benchmark YOUR software. DON'T trust |
67 |
logic or generic benchmarks or web pages with results. Trust YOUR |
68 |
results only. |
69 |
|
70 |
HTH |
71 |
|
72 |
Rafa |