XRAID Hard-Drive Speeds?

---

StephenB wrote:

Not sure why you are using oflag=dsync, so I ran the command both ways.  This is on an RN526x with 4x6TB WD60EFRX drives.

root@NAS:/data/Test# dd bs=1M count=256 if=/dev/zero of=test oflag=dsync
256+0 records in
256+0 records out
268435456 bytes (268 MB, 256 MiB) copied, 28.2462 s, 9.5 MB/s
root@NAS:/data/Test# dd bs=1M count=256 if=/dev/zero of=test
256+0 records in
256+0 records out
268435456 bytes (268 MB, 256 MiB) copied, 0.61322 s, 438 MB/s
root@NAS:/data/Test# 

I believe dsync (data only sync, no meta data) writes sequentially and ensures write cache isn't used or kept to a min (buffer flushed out evey block)...where as direct bypasses linux page cache, however writes concurrently. I'm not sure if there is an option to write sequentially and bypass linux page cahe altogether, so I used dsync for sequential writing for a more accurate write speed. Not that it really matters when comparing the same tests, but it does minimize variance in influence.

(testing with some unavoidable hard-drive use):

***@NAS:~$ dd bs=1M count=256 if=/dev/zero of=test oflag=dsync                      
256+0 records in                                                                      
256+0 records out                                                                     
268435456 bytes (268 MB, 256 MiB) copied, 22.7576 s, 11.8 MB/s                        
***@NAS:~$ dd bs=1M count=256 if=/dev/zero of=test oflag=direct                     
256+0 records in                                                                      
256+0 records out                                                                     
268435456 bytes (268 MB, 256 MiB) copied, 4.33843 s, 61.9 MB/s
***@NAS:~$ dd bs=1M count=256 if=/dev/zero of=test 
256+0 records in 
256+0 records out 
268435456 bytes (268 MB, 256 MiB) copied, 0.849287 s, 316 MB/s   

Appears the hard-drive speeds are comparable - phew :)

StephenB wrote:

Retired_Member wrote:

 

removed network bonding and appears to have speed up network transfers considerably (46Mbs)

 

You should be seeing ~100 MB/sec for large sequential file transfers over a gigabit network.  Perhaps re-run iperf and see what your raw network speeds are without bonding.

 

What bonding mode were you using on the NAS (and what mode did you have enabled on the switch)?

 

Also, what MTU are you using on the NAS?

Config: NAS MTU=1500, same as router. 

CPE: NAS 2x1gb <-> 2xcat 6 <-> 2x1gb router ports <-> cat6 <-> 1xgb pc

I was using round-ribbon bonding, but appears ok now after swapping out the ethernet cables. I appear to be getting around the 750-890Mbps range, which is around 100MB/sec.

***@PC: iperf -c 10.1.1.250 -i 5 -d
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 208 KByte (default)
------------------------------------------------------------
------------------------------------------------------------
Client connecting to 10.1.1.250, TCP port 5001
TCP window size: 208 KByte (default)
------------------------------------------------------------
[ 4] local 10.1.1.89 port 3937 connected with 10.1.1.250 port 5001
[ 5] local 10.1.1.89 port 5001 connected with 10.1.1.250 port 33836
[ ID] Interval Transfer Bandwidth
[ 4] 0.0- 5.0 sec 447 MBytes 749 Mbits/sec
[ 4] 5.0-10.0 sec 531 MBytes 891 Mbits/sec
[ 4] 0.0-10.0 sec 978 MBytes 820 Mbits/sec