NETGEAR is aware of a growing number of phone and online scams. To learn how to stay safe click here.
Forum Discussion
Post your performance results
I thought it might be interesting to see what kind of performance everyone is getting with IO Meter so that we can compare what we're getting. So I thought we could all post our results in this thread...
Hi, I've been doing some performance tests with 2 ReadyNAS 2100 and a ReadyNAS Pro4. I'm a bit confused because I've found that the network performance is very good, and the disks gives high performance also, but combined the results are not as good. I'm sharing here the information with the interest of clarifying this results and, if possible, improve the performance.
ReadyNAS 2100 with 4x4TB Western Digital Red Hard disks. Local disk performance test:
/c/test# time dd if=/dev/zero of=test_10G bs=1M count=10000
10000+0 records in
10000+0 records out
10485760000 bytes (10 GB) copied, 75.0244 s, 140 MB/s
real 1m15.470s
user 0m0.146s
sys 1m6.038s
Same ReadyNAS connected to a Gigabit network (Netgear 724). Network performance test 1:
On the nas: nc -v -v -l -n -p 2222 >/dev/null
On a server: cat /dev/zero |nc -v -v -n ip_address 2222 >/dev/null
Collecting stats: dstat -N eth0,eth1,bond0 -D md2,md3
----total-cpu-usage---- --dsk/md2-----dsk/md3-- --net/eth0----net/eth1---net/bond0- ---paging-- ---system--
usr sys idl wai hiq siq|_read _writ:_read _writ|_recv _send:_recv _send:_recv _send|__in_ _out_|_int_ _csw_
0 3 91 3 0 2|8092k 491k: 690k 100k| 0 0 : 0 0 : 0 0 | 0.1 5.4B|5745 3255
0 1 97 1 0 1|1132k 0 : 0 0 |2212k 312k: 140B 214B:2212k 313k| 0 0 |4108 2074
1 19 63 2 0 16| 52k 0 : 0 0 | 117M 2479k: 62k 69k: 117M 2547k| 0 0 | 13k 4134
1 17 66 0 0 16| 0 0 : 0 0 | 111M 2332k: 70B 214B: 111M 2330k| 0 0 | 12k 4245
1 18 66 1 0 16| 104k 0 : 0 0 | 117M 2858k: 258B 402B: 117M 2861k| 0 0 | 12k 4216
1 18 64 3 0 15| 92k 52k: 0 0 | 118M 2479k: 25k 27k: 118M 2505k| 0 0 | 12k 3467
1 17 67 0 0 15| 0 0 :4096B 0 | 117M 2478k: 22k 24k: 117M 2501k| 0 0 | 12k 3486
1 17 69 0 0 14|4096B 0 : 0 0 | 117M 2497k: 11k 12k: [117M 2510k| 0 0 | 12k 3159
4 17 65 0 0 14| 16k 0 : 0 0 | 115M 2303k:1744B 647B: 115M 2304k| 0 0 | 12k 2713
As you can see, the network is working well, using one gigabit interface of the bond at max speed.
I can repeat the experiment with dd specifying blocksize.
Same scenario. Network performance test2 using 1MB blocksize (as on local disk test).
On the nas: nc -l -n -p 2222 | dd of=/dev/null bs=1M
0+258465 records in
0+258465 records out
2588194952 bytes (2.6 GB) copied, 25.0691 s, 103 MB/s
On the server: time dd if=/dev/zero bs=1M |nc -n 192.168.200.101 2222 >/dev/null
2469+0 records in
2468+0 records out
2587885568 bytes (2.6 GB) copied, 23.5151 s, 110 MB/s
(the transfer time observed on the nas is higher because i'm entering the commands on different terminals, but the real time is 23.5ms -> 110MB/s)
Similar results. Command: dstat -N eth0,eth1,bond0 -D md2,md3
----total-cpu-usage---- --dsk/md2-----dsk/md3-- --net/eth0----net/eth1---net/bond0- ---paging-- ---system--
usr sys idl wai hiq siq|_read _writ:_read _writ|_recv _send:_recv _send:_recv _send|__in_ _out_|_int_ _csw_
1 28 56 0 0 15| 0 0 : 0 0 | 114M 1083k: 280B 461B: 114M 1084k| 0 0 | 15k 23k
2 27 56 0 0 15| 0 0 :4096B 0 | 114M 1067k: 70B 214B: 114M 1067k| 0 0 | 15k 24k
2 28 55 0 0 15|8192B 12k: 0 0 | 113M 1045k: 11k 2648B: 113M 1048k| 0 0 | 15k 23k
2 28 55 0 0 15| 0 1668k: 0 120k| 112M 1066k: 70B 214B: 112M 1067k| 0 0 | 15k 23k
2 26 57 0 0 15| 0 64k: 0 0 | 112M 1038k: 362B 372B: 112M 1039k| 0 0 | 15k 23k
1 28 56 0 0 15| 0 0 : 0 0 | 112M 1054k: 346B 332B: 112M 1054k| 0 0 | 15k 23k
2 29 54 0 0 15| 0 0 :4096B 0 | 113M 1185k: 70B 214B: 113M 1185k| 0 0 | 14k 24k
Summary
The local disk test show a write transfer test of 140 MB/s
The network tests shows tranfers of 110MB/s with a gigabit network interface at max speed.
With these results I was expecting to be able to do network writes on the NAS disks near 100MB/s as I understand that the bottleneck is on the network, but as you will see in the next combined test, this is not what I get.
Network writes test (combined test).
On the server: time dd if=/dev/zero bs=1M |nc -n 192.168.200.101 2222 >/dev/null
^C
real 0m26.680s
user 0m0.062s
sys 0m4.103s
On the NAS: nc -l -n -p 2222 | dd of=/c/test/test_1G bs=1M
0+97988 records in
0+97988 records out
1884771088 bytes (1.9 GB) copied, 29.1939 s, 64.6 MB/s
Poor results. Command: dstat -N eth0,eth1,bond0 -D md2,md3
----total-cpu-usage---- --dsk/md2-----dsk/md3-- --net/eth0----net/eth1---net/bond0- ---paging-- ---system--
usr sys idl wai hiq siq|_read _writ:_read _writ|_recv _send:_recv _send:_recv _send|__in_ _out_|_int_ _csw_
1 46 33 3 0 17|4096B 3924k:4096B 56M| 79M 905k: 12k 3946B: 79M 909k| 0 0 | 12k 7648
1 45 33 6 0 16| 0 1560k:4096B 28M| 79M 795k: 842B 1162B: 79M 796k| 0 0 | 13k 4730
1 33 46 11 0 9| 0 0 : 0 73M| 35M 448k: 258B 402B: 35M 449k| 0 0 |8245 4689
1 53 29 0 0 17| 0 164k: 0 95M| 71M 750k: 70B 214B: 71M 750k| 0 0 | 12k 6083
1 46 33 4 0 16| 192k 4096B: 0 31M| 86M 963k: 19k 21k: 86M 984k| 0 0 | 12k 5863
1 49 33 0 0 17| 0 308k: 0 46M| 80M 797k: 134B 214B: 80M 797k| 0 0 | 12k 5571
1 52 31 0 0 16| 0 396k:8192B 91M| 64M 678k:2084B 2268B: 64M 680k| 0 0 | 11k 5981
As you can see, the only modification is redirecting the network output to a file, using the same blocksize that on previous tests, an the use of the network goes down significantly. The CPU is still 31% idle, and the write's statistics are lower than in the first test.
With the ReadyNAS Pro4 (with Radiator 4.2.27 or ReadyNAS OS6) we get similar results. The local disks writes are a bit higher (near 200MBps) but the network writes (combined test) are about 60MBps.
Can somebody explain me these results? I'd like to improve the ReadyNAS performance if possible.
Thanks in advance,
Manuel Martínez
ReadyNAS 2100 with 4x4TB Western Digital Red Hard disks. Local disk performance test:
/c/test# time dd if=/dev/zero of=test_10G bs=1M count=10000
10000+0 records in
10000+0 records out
10485760000 bytes (10 GB) copied, 75.0244 s, 140 MB/s
real 1m15.470s
user 0m0.146s
sys 1m6.038s
Same ReadyNAS connected to a Gigabit network (Netgear 724). Network performance test 1:
On the nas: nc -v -v -l -n -p 2222 >/dev/null
On a server: cat /dev/zero |nc -v -v -n ip_address 2222 >/dev/null
Collecting stats: dstat -N eth0,eth1,bond0 -D md2,md3
----total-cpu-usage---- --dsk/md2-----dsk/md3-- --net/eth0----net/eth1---net/bond0- ---paging-- ---system--
usr sys idl wai hiq siq|_read _writ:_read _writ|_recv _send:_recv _send:_recv _send|__in_ _out_|_int_ _csw_
0 3 91 3 0 2|8092k 491k: 690k 100k| 0 0 : 0 0 : 0 0 | 0.1 5.4B|5745 3255
0 1 97 1 0 1|1132k 0 : 0 0 |2212k 312k: 140B 214B:2212k 313k| 0 0 |4108 2074
1 19 63 2 0 16| 52k 0 : 0 0 | 117M 2479k: 62k 69k: 117M 2547k| 0 0 | 13k 4134
1 17 66 0 0 16| 0 0 : 0 0 | 111M 2332k: 70B 214B: 111M 2330k| 0 0 | 12k 4245
1 18 66 1 0 16| 104k 0 : 0 0 | 117M 2858k: 258B 402B: 117M 2861k| 0 0 | 12k 4216
1 18 64 3 0 15| 92k 52k: 0 0 | 118M 2479k: 25k 27k: 118M 2505k| 0 0 | 12k 3467
1 17 67 0 0 15| 0 0 :4096B 0 | 117M 2478k: 22k 24k: 117M 2501k| 0 0 | 12k 3486
1 17 69 0 0 14|4096B 0 : 0 0 | 117M 2497k: 11k 12k: [117M 2510k| 0 0 | 12k 3159
4 17 65 0 0 14| 16k 0 : 0 0 | 115M 2303k:1744B 647B: 115M 2304k| 0 0 | 12k 2713
As you can see, the network is working well, using one gigabit interface of the bond at max speed.
I can repeat the experiment with dd specifying blocksize.
Same scenario. Network performance test2 using 1MB blocksize (as on local disk test).
On the nas: nc -l -n -p 2222 | dd of=/dev/null bs=1M
0+258465 records in
0+258465 records out
2588194952 bytes (2.6 GB) copied, 25.0691 s, 103 MB/s
On the server: time dd if=/dev/zero bs=1M |nc -n 192.168.200.101 2222 >/dev/null
2469+0 records in
2468+0 records out
2587885568 bytes (2.6 GB) copied, 23.5151 s, 110 MB/s
(the transfer time observed on the nas is higher because i'm entering the commands on different terminals, but the real time is 23.5ms -> 110MB/s)
Similar results. Command: dstat -N eth0,eth1,bond0 -D md2,md3
----total-cpu-usage---- --dsk/md2-----dsk/md3-- --net/eth0----net/eth1---net/bond0- ---paging-- ---system--
usr sys idl wai hiq siq|_read _writ:_read _writ|_recv _send:_recv _send:_recv _send|__in_ _out_|_int_ _csw_
1 28 56 0 0 15| 0 0 : 0 0 | 114M 1083k: 280B 461B: 114M 1084k| 0 0 | 15k 23k
2 27 56 0 0 15| 0 0 :4096B 0 | 114M 1067k: 70B 214B: 114M 1067k| 0 0 | 15k 24k
2 28 55 0 0 15|8192B 12k: 0 0 | 113M 1045k: 11k 2648B: 113M 1048k| 0 0 | 15k 23k
2 28 55 0 0 15| 0 1668k: 0 120k| 112M 1066k: 70B 214B: 112M 1067k| 0 0 | 15k 23k
2 26 57 0 0 15| 0 64k: 0 0 | 112M 1038k: 362B 372B: 112M 1039k| 0 0 | 15k 23k
1 28 56 0 0 15| 0 0 : 0 0 | 112M 1054k: 346B 332B: 112M 1054k| 0 0 | 15k 23k
2 29 54 0 0 15| 0 0 :4096B 0 | 113M 1185k: 70B 214B: 113M 1185k| 0 0 | 14k 24k
Summary
The local disk test show a write transfer test of 140 MB/s
The network tests shows tranfers of 110MB/s with a gigabit network interface at max speed.
With these results I was expecting to be able to do network writes on the NAS disks near 100MB/s as I understand that the bottleneck is on the network, but as you will see in the next combined test, this is not what I get.
Network writes test (combined test).
On the server: time dd if=/dev/zero bs=1M |nc -n 192.168.200.101 2222 >/dev/null
^C
real 0m26.680s
user 0m0.062s
sys 0m4.103s
On the NAS: nc -l -n -p 2222 | dd of=/c/test/test_1G bs=1M
0+97988 records in
0+97988 records out
1884771088 bytes (1.9 GB) copied, 29.1939 s, 64.6 MB/s
Poor results. Command: dstat -N eth0,eth1,bond0 -D md2,md3
----total-cpu-usage---- --dsk/md2-----dsk/md3-- --net/eth0----net/eth1---net/bond0- ---paging-- ---system--
usr sys idl wai hiq siq|_read _writ:_read _writ|_recv _send:_recv _send:_recv _send|__in_ _out_|_int_ _csw_
1 46 33 3 0 17|4096B 3924k:4096B 56M| 79M 905k: 12k 3946B: 79M 909k| 0 0 | 12k 7648
1 45 33 6 0 16| 0 1560k:4096B 28M| 79M 795k: 842B 1162B: 79M 796k| 0 0 | 13k 4730
1 33 46 11 0 9| 0 0 : 0 73M| 35M 448k: 258B 402B: 35M 449k| 0 0 |8245 4689
1 53 29 0 0 17| 0 164k: 0 95M| 71M 750k: 70B 214B: 71M 750k| 0 0 | 12k 6083
1 46 33 4 0 16| 192k 4096B: 0 31M| 86M 963k: 19k 21k: 86M 984k| 0 0 | 12k 5863
1 49 33 0 0 17| 0 308k: 0 46M| 80M 797k: 134B 214B: 80M 797k| 0 0 | 12k 5571
1 52 31 0 0 16| 0 396k:8192B 91M| 64M 678k:2084B 2268B: 64M 680k| 0 0 | 11k 5981
As you can see, the only modification is redirecting the network output to a file, using the same blocksize that on previous tests, an the use of the network goes down significantly. The CPU is still 31% idle, and the write's statistics are lower than in the first test.
With the ReadyNAS Pro4 (with Radiator 4.2.27 or ReadyNAS OS6) we get similar results. The local disks writes are a bit higher (near 200MBps) but the network writes (combined test) are about 60MBps.
Can somebody explain me these results? I'd like to improve the ReadyNAS performance if possible.
Thanks in advance,
Manuel Martínez
NETGEAR Academy
Boost your skills with the Netgear Academy - Get trained, certified and stay ahead with the latest Netgear technology!
Join Us!
