Replication performance between: E5620 @ 2.40 GHz V.S. X5677 @ 3.46 GHz

[helloha]

I'm on the older X8 platform but not really planning to upgrade to anything newer.

I was wondering however if I could get better performance with data replication going from my primary server to my backup over 10GbE.

On freenas 9 (have not tested this on 11.2) the data replication was bottlenecked because of SMB or SSH was single threaded. When running TOP, one of these processes would be maxed out. During backup, I would get 200 MB/s, about 1/3 of the pool performance.

So would going from an E5620 to the X5677 make a lot of difference?

Or has data replication changed since freenas 11?

Thanks!
Karel.

 

[dlavigne]

Replication is changing for 11.3 (no significant changes before then, though 9 is quite old).

 

[helloha]

Thanks, I'll hold off for now then.

 

[helloha]

Reviving this old thread.

I upgraded to 11.3 but so far speeds seem about the same. 150 MB/s

When running top I can see SSH is pinned to the max.

Single threaded SSH seems still to be the case?

Regards,
K.

Code:

last pid:  5473;  load averages:  1.61,  1.18,  0.61
66 processes:  3 running, 63 sleeping
CPU:  5.1% user,  0.0% nice,  2.9% system,  0.5% interrupt, 91.5% idle
Mem: 41M Active, 291M Inact, 1115M Laundry, 59G Wired, 1477M Free
ARC: 56G Total, 34G MFU, 22G MRU, 20M Anon, 132M Header, 28M Other
     54G Compressed, 54G Uncompressed, 1.01:1 Ratio
Swap: 10G Total, 10G Free

  PID USERNAME       THR PRI NICE   SIZE    RES STATE   C   TIME    WCPU COMMAND
 5254 root             1 101    0 12428K  8076K CPU6    6   6:26  95.44% ssh
 5255 root             2  22    0 10036K  4232K CPU15  15   0:43   9.01% zfs
   78 root            40  20    0   332M   250M kqread  3   1:34   2.34% python3.7
 1759 root            15  22    0   175M   138M umtxn  13   0:18   1.33% uwsgi-3.7
 5473 root             1  21    0  7888K  3896K CPU3    3   0:00   0.14% top
 3547 root            12  20    0 92588K 64436K usem    7   0:03   0.07% python3.7
 1503 www              1  20    0 30916K  9056K kqread  4   0:01   0.03% nginx

 

[Johnny Fartpants]

I can achieve approx 3Gb/s over 10Gb network when replicating on 11.2 but I had to tweak my replication settings to use no compression during replication and a 'Fast' Cipher.

 

[helloha]

Nevermind, seemed to have tweaked it a bit and now it is running fine. LZ4 compression was also enabled on destination which might also have been an issue.

Peaking at 700-800 MB/s !

SSH is not pinned anymore. Seems to be handled by python:

Code:

  PID USERNAME       THR PRI NICE   SIZE    RES STATE   C   TIME    WCPU winbindd
 3612 root             1  31    0 21736K 17756K sbwait  8   0:39  43.43% python3.7
   78 root            38  25    0   340M   256M kqread  3   1:04   0.58% python3.7
  516 root            10  20    0 90660K 64440K usem    8   0:04   0.13% python3.7
 3277 root             1  20    0  7888K  4436K CPU15  15   0:01   0.10% top
 1522 www              1  20    0 30916K  9292K kqread 14   0:00   0.01% nginx
 2446 root             1  20    0 12920K  8020K select 13   0:00   0.01% sshd
 1507 root             8  20    0 31452K 11300K select  2   0:04   0.01% rrdcached
 1592 root             1  20    0   127M   107M kqread  3   0:02   0.01% uwsgi-3.7
 1453 root             1  20    0 12464K 12560K select 12   0:01   0.00% ntpd
 2113 root            11  20    0 94556K 64044K nanslp 10   0:35   0.00% collectd
 1804 root            15  21    0   178M   137M umtxn  15   0:10   0.00% uwsgi-3.7
  139 root             3  20    0   167M   140M usem   14   0:05   0.00% python3.7
  136 root             3  20    0   169M   139M piperd  4   0:05   0.00% python3.7
  138 root             3  20    0   176M   141M usem    9   0:05   0.00% python3.7