iSCSI Storage Performance & Benchmarking with IOMeter



This is a storage performance report that I have done on 3 different storage types. I was curious to see the performance difference between different types of iSCSI storage units. I wanted to share the results with others because when I wanted to compare my results with others, I was surprised to see that there are a lot of people discussing about storage technologies and all but no one is actually sharing their benchmarks openly.
Because, a storage benchmark can vary a lot depending on your environment, I am going to list my setup so you can compare these results to your own environment as close as possible.

Storage Units

Dell Equallogic PS4000 – Model # 70-0120
16 X 500GB 7200 RPM Enterprise Disks (RAID50 Policy) – 512 bytes sector sizes
2 X 1Gbps NIC (Jumbo Frames enabled mtu9000) – Multipathing

Dell Equallogic PS6100 – Model # 70-0400
24 X 1TB 7200 RPM Enterprise Disks (RAID50 Policy) – 512 bytes sector sizes
4 X 1Gbps NIC (Jumbo Frames enabled mtu 9000) – Multipathing

Freenas 9.3
Dell PowerEdge 2950 with 5 GB RAMand Intel Xeon CPU 3.00 GHz
6 X 1TB 7200 RPM Desktop Grade* disks (RAID 10) with an old Dell RAID controller
2 X 1Gbps NIC (Jumbo Frames not enabled!) – Multipathing

Host
Dell PowerEdge R610 with 2 X Intel Xeon E5520
VMware ESXi 5.5.0

VM
Windows Server 2012 R2 – 2vcpu + 8GB RAM
OS partition is on the local disk of the host
The IOMETER test partition is a 30GB (unformatted) partition.

I’ve used the pre-configured IOMETER configuration file for a vmware environment that I found on the internet. The only thing I’ve changed is the sector size of this config file. It was testing for 8GB only so for some tests I bumped that to 20GB to make sure that I wasn’t using any cache while running these tests. Please read carefully at the beginning of each result how the test is done.
IOMETER is saving the results into a csv file that you can format later. The problem with IOMETERs csv file that it needs formatting… And lots of it J. This process can take a lot of time but… Thanks to this blog http://vmktree.org/iometer/ , you can now transform these unreadable csv files into a well presented simple Excel kinda tables.


 
 THESE TESTS ARE DONE WITH 20GB sectore size 41943040 blocks.
 
IOMETER RESULTS
VMWARE PS6100-03_20GB
Test name
Latency Avg iops Avg MBps cpu load
4K; 100% Read; 0% random
1.34 23820 93 31%
4K; 50% Read; 0% random
3.83 8344 32 12%
4K; 0% Read; 0% random
2.27 14106 55 25%
4K; 100% Read; 0% random
3.77 8487 132 8%
4K; 50% Read; 0% random
5.11 6258 97 7%
4K; 0% Read; 0% random
4.39 7286 113 16%
16K; 100% Read; 0% random
10.94 2925 91 0%
16K; 50% Read; 0% random
9.11 3510 109 1%
16K; 0% Read; 0% random
5.78 5531 172 8%
32K; 100% Read; 0% random
20.46 1563 6 0%
32K; 50% Read; 0% random
14.33 2233 8 0%
32K; 0% Read; 0% random
6.59 4853 18 3%
4K; 100% Read; 100% random (1)
25.93 1233 19 0%
4K; 50% Read; 100% random (1)
18.58 1721 26 0%
4K; 0% Read; 100% random (1)
12.48 2563 40 0%
16K; 100% Read; 100% random (1)
28.31 1129 35 0%
16K; 50% Read; 100% random (1)
22.21 1439 44 0%
16K; 0% Read; 100% random (1)
18.06 1771 55 0%
 
IOMETER RESULTS
VMWARE PS4000-02_20GB
Test name
Latency Avg iops Avg MBps cpu load
4K; 100% Read; 0% random
0.82 38845 151 45%
4K; 50% Read; 0% random
3.8 8415 32 9%
4K; 0% Read; 0% random
2.55 12551 49 21%
4K; 100% Read; 0% random
2.08 15403 240 27%
4K; 50% Read; 0% random
5.67 5644 88 2%
4K; 0% Read; 0% random
5.21 6144 96 7%
16K; 100% Read; 0% random
6.32 5059 158 3%
16K; 50% Read; 0% random
7.92 4039 126 1%
16K; 0% Read; 0% random
7.04 4541 141 4%
32K; 100% Read; 0% random
14.04 2278 8 0%
32K; 50% Read; 0% random
17.34 1844 7 0%
32K; 0% Read; 0% random
17.64 1813 7 0%
4K; 100% Read; 100% random (1)
19.44 1645 25 0%
4K; 50% Read; 100% random (1)
20.94 1527 23 0%
4K; 0% Read; 100% random (1)
27.51 1162 18 0%
16K; 100% Read; 100% random (1)
21.2 1509 47 0%
16K; 50% Read; 100% random (1)
26.17 1222 38 0%
16K; 0% Read; 100% random (1)
36.08 886 27 0%
 
IOMETER RESULTS   VMWARE FREENAS-02_ 20GB
Test name Latency Avg iops Avg MBps cpu load
4K; 100% Read; 0% random 0.86 37400 146 37%
4K; 50% Read; 0% random 2.3 13897 54 19%
4K; 0% Read; 0% random 2.23 14317 55 25%
4K; 100% Read; 0% random 3.22 9926 155 15%
4K; 50% Read; 0% random 4.65 6874 107 7%
4K; 0% Read; 0% random 6.07 5272 82 2%
16K; 100% Read; 0% random 8.64 3704 115 0%
16K; 50% Read; 0% random 7.8 4104 128 1%
16K; 0% Read; 0% random 11.24 2845 88 0%
32K; 100% Read; 0% random 1.36 23472 91 28%
32K; 50% Read; 0% random 4.24 7537 29 6%
32K; 0% Read; 0% random 11.78 2716 10 1%
4K; 100% Read; 100% random (1) 4.6 6950 108 10%
4K; 50% Read; 100% random (1) 8.86 3610 56 1%
4K; 0% Read; 100% random (1) 17.25 1855 28 0%
16K; 100% Read; 100% random (1) 9.07 3525 110 0%
16K; 50% Read; 100% random (1) 9.86 3246 101 0%
16K; 0% Read; 100% random (1) 18.76 1705 53 0%





















































































































 THESE TESTS ARE DONE WITH default sector size 80960 blocks.
 
IOMETER RESULTS
VMWARE PS6100-03
Test name
Latency Avg iops Avg MBps cpu load
4K; 100% Read; 0% random
1.03 30910 120 40%
4K; 50% Read; 0% random
2.69 11907 46 20%
4K; 0% Read; 0% random
2.34 13682 53 28%
4K; 100% Read; 0% random
3.25 9854 153 21%
4K; 50% Read; 0% random
3.43 9332 145 25%
4K; 0% Read; 0% random
3.71 8632 134 19%
16K; 100% Read; 0% random
5.61 5704 178 13%
16K; 50% Read; 0% random
5.27 6069 189 14%
16K; 0% Read; 0% random
5.37 5962 186 17%
32K; 100% Read; 0% random
1.69 18918 73 31%
32K; 50% Read; 0% random
4.65 6877 26 14%
32K; 0% Read; 0% random
6.45 4960 19 1%
4K; 100% Read; 100% random (1)
3.48 9184 143 21%
4K; 50% Read; 100% random (1)
4.6 6958 108 15%
4K; 0% Read; 100% random (1)
6.42 4980 77 7%
16K; 100% Read; 100% random (1)
5.85 5469 170 12%
16K; 50% Read; 100% random (1)
5.41 5912 184 15%
16K; 0% Read; 100% random (1)
6.8 4707 147 7%

 
IOMETER RESULTS
VMWARE PS4000-02
Test name
Latency Avg iops Avg MBps cpu load
4K; 100% Read; 0% random
0.82 39108 152 46%
4K; 50% Read; 0% random
3.2 10001 39 13%
4K; 0% Read; 0% random
2.39 13404 52 22%
4K; 100% Read; 0% random
3.15 10159 158 22%
4K; 50% Read; 0% random
4.51 7092 110 7%
4K; 0% Read; 0% random
5.46 5857 91 4%
16K; 100% Read; 0% random
5.72 5592 174 10%
16K; 50% Read; 0% random
6.24 5130 160 3%
16K; 0% Read; 0% random
6.72 4763 148 1%
32K; 100% Read; 0% random
1.34 23865 93 44%
32K; 50% Read; 0% random
4.96 6456 25 4%
32K; 0% Read; 0% random
7.38 4338 16 0%
4K; 100% Read; 100% random (1)
3.53 9072 141 17%
4K; 50% Read; 100% random (1)
6.01 5325 83 2%
4K; 0% Read; 100% random (1)
9 3556 55 0%
16K; 100% Read; 100% random (1)
6.19 5169 161 5%
16K; 50% Read; 100% random (1)
7.64 4185 130 0%
16K; 0% Read; 100% random (1)
10.52 3042 95 0%


 
IOMETER RESULTS
VMWARE Freenas-02
Test name
Latency Avg iops Avg MBps cpu load
4K; 100% Read; 0% random
0.07 40790 159 11%
4K; 50% Read; 0% random
0.06 0 0 23%
4K; 0% Read; 0% random
0 0 0 23%
4K; 100% Read; 0% random
0.82 39141 152 19%
4K; 50% Read; 0% random
1.78 17990 70 23%
4K; 0% Read; 0% random
2.2 14567 56 24%
16K; 100% Read; 0% random
4.53 7055 110 6%
16K; 50% Read; 0% random
3.98 8035 125 11%
16K; 0% Read; 0% random
5.62 5694 88 2%
32K; 100% Read; 0% random
9.02 3545 110 0%
32K; 50% Read; 0% random
7.23 4424 138 0%
32K; 0% Read; 0% random
9.52 3360 105 0%
4K; 100% Read; 100% random (1)
1.25 25491 99 29%
4K; 50% Read; 100% random (1)
1.73 18541 72 18%
4K; 0% Read; 100% random (1)
2.22 14409 56 21%
16K; 100% Read; 100% random (1)
4.56 7015 109 4%
16K; 50% Read; 100% random (1)
4.02 7958 124 11%
16K; 0% Read; 100% random (1)
5.76 5557 86 2%

Comments

Post a Comment

Popular posts from this blog

System.Messaging.MessageQueueException (0x80004005): A workgroup installation computer does not support the operation (Public Queue create issue)

This is a weird issue that I don't fully understand why it would happen. But the symptoms are very persistent as we're able to reproduce these behavior easily. Issue: The MSMQ (Microsoft Message Queuing) fails to create Public Queues and throws  "the System.Messaging.MessageQueueException (0x80004005): A workgroup installation computer does not support the operation" exception.
Read more

Veeam Backup Error : Failed to prepare guest for hot backup. Error: VSSControl

Image
Problem Description; I had this issue on one of my backup jobs recently and I was able to fix the issue with the solution explained below. After a quick google search, I realized that I'm not the only one and neither you are :) Full Error Description; Failed to prepare guest for hot backup. Error: VSSControl: Failed to prepare guest for freeze, wait tim eout 9 00 sec Error: VSSControl: Failed to prepare guest for freeze, wait timeout 900 sec
Read more

warning: Win32API is deprecated after Ruby 1.9.1; use fiddle directly instead - Chef Development Kit Update

Today, I needed to do some tests with a newer version of the Chef Client and test my cookbooks. To version 14.12.9-1 to be precise. I also upgraded my ChefDK on my workstation to the 4.2.0 version. I believe that upgrades Ruby version as well and now I have ruby version 2.6.3p62 Right after the upgrade, I tried with a basic knife node list command to see if I needed to reboot or anything was broke... Everything seems to be working fine, except I get this annoying "warning" message. As it's just a warning, I don't really care about it as it just seems to annonce something that is deprecated... Problem; You receive "warning: Win32API is deprecated after Ruby 1.9.1; use fiddle directly instead" warning when using ruby or chef knife commands. Solution; You need to comment out that warning message in Ruby lib scripts.
Read more