Using a Raspberry Pi seems to make alot of sense if you're in the market for a small NAS server, it's low cost, low profile, low energy etc...

Is It Up To It?

The only possible problem is the storage the only way to attach any real storage to a Pi is via USB. USB 2.0 has a maximum theoretical throughput of 480Mbits/s with a but I dont think you'll ever get anywhere near that speed.

We only really need to acheieve around 10MB/s as then the ethernet connection becomes the bottleneck


I wanted to see what kind of throughput I could actually get with USB devices connected to a Pi so I wrote a device benchamarker tool that is able to measure both the read and write speeds of one or more block devices.

The Setup

I setup my Raspberry Pi B in the following way:

  • Clean install of Raspbian 2015-02-16
  • Ran apt-get upgrade on 00:00 10/04/15
  • 2x Verbatim 4GB memory sticks ID 18a5:0302 Verbatim, Ltd Flash Drive

The HDDs

Ok memory sticks aren't HDDs but they are what I had to hand at the time.

I formatted both the memory sticks to ext4 using gparted and ran my device benchmarker on each of them individually then both in parallel.

The Results

I ran both dd-read and dd-write tests 10 times and then took an average:

Device Reading (MB/s) Writing (MB/s)
/dev/sda1 14.98 5.23
/dev/sdb1 14.04 5.25
/dev/sd* 12.86 5.05

The results for the combined /dev/sd* are the speeds of the induvidual devices when tested at the same time. Full details of all the tests that were run, what exactly was run, how long the tests ran for etc can be found [here](/documents/2015/04/13/USB Speed Results.ods)

The write speeds seemed a little slow to me, I knew they were going to be slower than the read speed but 3x slower? After a quick Google I stumbled upon http://usbspeed.nirsoft.net, it's basically a crowdsourced database of USB memory stick read write speeds. I looked up the entires they had for my USB drive and took an average, they pretty much matched up so it seems that's what they capable of.

What Next?

Interestingly when both devices were written to/read from in parallel the speed seemed to only drop a little meaning the USB bus wasn't being saturated, the drives are the bottleneck. Could we boost the read/write speeds if we were able to treat both devices as a single device, in RAID0 for example?

I realise going after a higher read speed is a little pointless as at 14MB/s we would be saturating the ethernet connection which is only capable of an sbsolute maximum of 10MB/s but there is room for a 2x increase in the write speeds.


I'm going to use MDADM to setup a RAID0 array and then run the benchmarks again to see if we can achieve a write speed of >10MB/s

MDADM Array Setup

Installing MDADM was as simple and painless as:

sudo apt-get install mdadm

Once MDADM was installed I assembled a new RAID0 array from both my drives and formatted the array to ext4, the same filesytem as before:

# Create the array
sudo mdadm --create --verbose /dev/md0 --level=stripe --raid-devices=2 /dev/sda1 /dev/sdb1

# Format the new array as ext4
sudo mkfs -t ext4 /dev/md0


I re-ran exactly the same benchmarks as before and got the following results:

Device Reading (MB/s) Writing (MB/s)
/dev/md0 27.92 8.92

Again full details of all the tests that were run can be found [here](/documents/2015/04/13/USB Speed Results.ods)

Thats ~192% the read speed and ~170% the read speed. More importantly the write speed is nearing the 10MB/s maximum the Pis ethernet controller can handle!


The memory sticks I was using definitly aren't the fastest out there and given the realistic through put of USB 2.0, which according to wiki is around 35MB/s, there's definitly still alot of room for improvement.

But what about that 10MB/s limit of ethernet? Gigabit ofcourse! I'm not the first to try this, see this article for example, I have a Raspberry Pi 2 so maybe i'll try hooking to a USB Gigabit adapter and add 2 or 3 faster USB drives and see if I can max out the USB bus.