SMP test of Erlang in multi-core environment under Linux

Goal

(1) Understand the effect of Erlang concurrent process scheduling on CPU core load in multi-core CPU environment;

(2) The memory increment mechanism of erlang virtual machine;

(3) The operation of Erlang process;

(4) Monitoring CPU usage under Linux

Experimental environment

Lenovo Minicomputer: Operating system: RedHat Enterprise linuxserver release6.4 (Santiago)

Kernel version: Linux Server1 2.6.32-358.el6.x86_64#1 SMP

CPU Model: Intel (R) Xeon (r) CPU e7-4820 @2.00ghz;

4 CPUs with physical cores per CPU of 8, total physical cores of 32, logical cores of 64

Memory: 125G

Disk: 289G

Erlang OTP:ERLANG/OTP 17[erts-6.0][source][64bit][smp:64:64][async-thread:10]

The test process Erlang concurrency process

Test code

-module (Test).

-export ([START_PROC/1]).

Start_proc (Num)

Case Num =:= 0 of

True-OK;

False--spawn (fun (), loop () end), Start_proc (Num-1)

End.

Loop (), loop ().

CPU Health

Test:start_proc (1000000) After starting 1 million processes, execute Mpstat–pall 2 10

Concurrent IO Test Code

-module (Testio).

-export ([START_PROC/1]).

Name (),

{a,b,c} =erlang:now (),

Integer_to_list (A) ++integer_to_list (B) ++integer_to_list (C).

Io ()

{OK,FD} = File:open ("./data/" ++name (), [Write,raw,binary,append]),

File:write (Fd, "Hi Zcc,nice to Meet You"),

File:read (fd,23),

File:close (Fd).

Start_proc (Num)

io = Fun ()-io () end,

Case Num =:= 0of

True->ok;

False-> spawn (IO), Timer:sleep (1), Start_proc (Num-1)

End.

View kernel-related information under Linux

Vmstat–n 3 Refresh every 3s

Analysis of CPU utilization using SAR Sar–u 2 10

Profiling run Process Queue Length Sar–q 2 10

CPU%usr%nice%sys%iowait%irq%soft%steal%guest%idle

Mpstat–p 1: View information for CPU number 1th

%steal:% of Virtual CPU unconscious wait time

%nice:cpu time percentage in user mode with nice value

Press the F key after Top–m and press J, enter to see which core the process is running on

Conclusion

(1) in the default Erlang SMP scheduling process Schedualer is the CPU's core corresponding to the CPU, not the number of corresponding, as in this environment there are 4 cpu,32 physical cores, open hyper-thread 64 logical core, the number of scheduling process is 64;

(2) When Erlang concurrent multiple processes, through the load of the CPU can be seen, each process will be evenly distributed across the core to run, and not a nuclear load is too large, a nuclear load is too small to happen, this thing is the operating system to do, the programmer does not care;

(3) The memory of the Erlang virtual machine will automatically request an increase from the host memory as the number of processes increases, unlike the Java Virtual machine, jvm1.6 memory does not automatically increase memory, can only be manually configured to increase the JVM memory, from 1.7 to share the physical host memory; Erlang virtual The machine does not suppress the memory growth mechanism. The virtual machine constantly allocates memory, forcing the system to use swap zones until the virtual machine runs out of memory and becomes very dull. The ' private heap ' and queue-based programming models of Erlang virtual machines are both an advantage and a disadvantage when running Erlang VMs in a production environment, with a system-level detection to kill processes when Erlang memory usage soars.

(4) The Erlang Scheduler runs on an OS thread, and the OS determines whether it executes on a core, and generally the OS guarantees that the thread will run on a core during execution;

(5) Estimating the maximum number of processes in a system = Total memory/erlang:process_info (self (), memories), Wordsize=erlang:system_info (wordsize) of Erlang. A process in Erlang accounts for approximately 2667 bytes of memory.

(6) test results

boot process number	virtual machine memory	start time
100,000	299m	12.3s
1 million	2.7g	6.8s
10 million	27.2g	57.6s
55164851	132g	400s