1 Benchmarks:WD Elements
2 ======================
3 :author: Aaron Ball
4 :email: nullspoon@iohq.net
5
6 == Benchmarks:WD Elements
7
8 For my work computer, I installed https://archlinux.org[Arch Linux] on an
9 external USB hard drive. My idea behind that is that if I'm ever working from
10 home and forget to bring my charger with me, I can just plug the drive into
11 another computer and I'm back up and running. So far it's worked great. A few
12 months back though, I dropped the drive while it was running and while it was
13 surprisingly okay (despite being read-only until reboot), it was a bit slower.
14 I would assume a head crash, but thus far I have noticed no data corruption.
15
16 All that said, I want to get another drive that I can mirror with (software
17 raid 1 anybody?), just in case something happens. I've been hunting around
18 online for the last few days and have found it to be impressively difficult to
19 find real specs on external USB hard drives. Sure, you can get that it's USB3
20 and maybe even its rpm, but you're almost guaranteed not to find cache size or
21 even what drive is inside the enclosure, metrics I consider to be very
22 important. That's why I've decided to post the IO metrics for this drive.
23 Hopefully someone will find these metrics useful.
24
25 image:files/WD_Elements.jpg[height=300]
26
27 * *Manufacturer*: http://www.wdc.com/en/[Western Digital]
28 * *Name*: Elements
29 * *Made in*: Malaysia
30 * *Size*: 1TB
31 * *Interface*: USB 3.0/2.0
32 * *Average Write Speed*: 104 MB/s
33 * *Average Read Speed*: 107 MB/s
34
35
36 [[benchmarks]]
37 === Benchmarks
38
39 [[usb3-devzero-write]]
40 ==== USB3 /dev/zero Write
41
42 The fastest place I can think of to get data from and avoid any bottlenecks
43 outside of the drive is to write from /dev/zero. The amount of processing power
44 that goes into writing all zeros __is insignificant next to the power of the
45 force__...er...reading data from another drive, potentially introducing more
46 bottlenecks and not getting good measurements. Let us begin...
47
48 ----
49 dd if=/dev/zero of=/dev/sdc bs=1M count=8192
50 8191+1 records in
51 8191+1 records out
52 8589131776 bytes (8.6 GB) copied, 82.9999 s, 103 MB/s
53 ----
54
55 Double the amount of data being written...
56
57 ----
58 dd if=/dev/zero of=/dev/sdc bs=1M count=16384
59 16384+0 records in
60 16384+0 records out
61 17179869184 bytes (17 GB) copied, 161.13 s, 107 MB/s
62 ----
63
64 Looks like overall this drive consistently averages just over 100 MB/s plugged
65 in to USB3.
66
67
68 [[usb3-read-to-devnull]]
69 ==== USB3 Read to /dev/null
70
71 Here we're basically doing the same as writing from /dev/zero, but instead
72 we're reading verbatim the first _x_ consecutive number of bytes and sending
73 them to a device that literally can't be a bottleneck: /dev/null. It's like
74 sending dead satellites floating into space
75 (https://www.youtube.com/watch?v=rWVGupqvCL8[spaaaaacce]) instead of spending
76 the time to land them (if they don't burn up) and disassembling. If I had to
77 pick somewhere to send something fast where there wouldn't be any bottlenecks,
78 the vast void of space is where I'd send it - that is equivelant to /dev/null.
79 Not a great analogy, I know, but honestly, I just wanted to reference
80 https://www.youtube.com/watch?v=rWVGupqvCL8[that Portal] video.
81
82 ----
83 dd if=/dev/sdc of=/dev/null bs=1M count=8192
84 8192+0 records in
85 8192+0 records out
86 8589934592 bytes (8.6 GB) copied, 80.5907 s, 107 MB/s
87 ----
88
89
90 [[conclusion]]
91 === Conclusion
92
93 * **Average write speed**: 104 MBps (832 Mbps = .832 Gbps)
94 * **Average read speed**: 107 MBps (856 Mbps = .856 Gbps)
95
96 Overall I'd say this drive is okay. As mentioned, the maximum speed of the
97 https://en.wikipedia.org/wiki/USB_3.0[USB3 spec] is 5 Gbps and this is getting
98 just shy of 1/5 that. I won't balk at that because a 100 megabytes per second
99 transfer rate is still pretty impressive for an external hard drive (that's
100 838,860,800 bits per second!).
101
102 One final thing to note, I ran these benchmarks on two systems, my laptop and
103 my server, to make sure the USB3 port, processor, bus, etc. weren't themselves
104 bottlenecks. The transfer rates were nearly identical (insignificantly
105 different).
106
107
108 Category:Western_Digital
109
110 Category:Hard_Drives
111
112 Category:Benchmarks
113
114
115 // vim: set syntax=asciidoc:
|
