Sourdough in Virginia

I used the following sourdough recipe from King Arthur Flour. I struggled at first, but mainly because I wasn’t doing regular feedings every 12 hours. Once I did that I had a starter within four days. Also, I found that it was not critical to have unbleached flour. I got a vigorous starter using bleached flour.

Another note: for the first few days the starter really stunk. Then something changed and it started to smell great. I found that after I pulled a cup out for bread, I could just add another cup of flour and half cup of water (i do use filtered, not tap), then a week later I could revive the starter just fine.

Apparently if you are going to be away from your start for a long time you can just dry it out, save the chips. Unproven by me, but they say this method preserves the starter for eons.


My DIY Little Bits

I whipped up some quick LED modules and my own power module using some proto boards and components I purchased from amazon. I haven’t branched out to more complicated module types, but these ones work well. I think each module cost ten cents or so, rather than the 10 dollars little bits charges.

Note that I put a small button to enable power from the module to the daughter cards. The main reason for this is to avoid wearing down the batteries.

Also, you can see I chose a 5-pin connector. Any module can plug into another module, or directly into a power module (which has two output connectors).

When I do this over again Ill make the following adjustments:

  • Make the outside pins ground, inside pin power. This would completely avoid any problems from plugging modules in upside down
  • Provide an always-on pin – I like the idea of some modules being normally-off. But some modules, like a trinket module, should be always on, since they consume so little power as to not matter



Measuring raspberry pi performance

I initially had high hopes for the raspberry pi 3 and snapped up three of them. I hoped to see one take the place of my media center, another two for monitoring tasks. Unfortunately I became less than enthused about the pi for a media center as it has no good way to stream of youtube or vidangel. Additionally it couldn’t keep up with streaming some HD videos.

It has now been a few months since my initial disappointment and I decided to try out the new firmware and see if things got better.

The first tests i ran were using iperf. One raspberry pi, rpi1, was left behind at the old firmware. Another pi, rpi2, was upgraded to the latest.

Firmware Kernel
rpi1 1b7da52ec944a9e1691745036966b3b2a48b19e8b (Apr 7 2016) 4.1.21-v7+
rpi2 1e7b8e2c9a7319f7b22869f1334c66e2cfc99f4a (Jun 27 2016) 4.4.14-v7+

Initial iperf test (iperf client running on my macbook, server on each raspberry pi – tests were run independently, on one raspberry pi at a time):

RunRpi1 Rpi2
128 Mbits/sec 39 Mbits/sec
231 Mbits/sec 37 Mbits/sec
330 Mbits/sec 39 Mbits/sec
430 Mbits/sec 38 Mbits/sec

I also tried this with a parallel iperf test (iperf -c -P 10)

RunRpi1 Rpi2
130 Mbits/sec 34 Mbits/sec
229 Mbits/sec 38 Mbits/sec
329 Mbits/sec 35 Mbits/sec
429 Mbits/sec 35 Mbits/sec

I also wanted to test sustained (iperf -c -t <10,120>)

Duration (seconds) Rpi1 Rpi2
10 29 Mbits/sec 39 Mbits/sec
120 29 Mbits/sec 35 Mbits/sec

Even on a udp iperf test (-u on the server, -b 50m on client) rpi1 gets 32 Mbits/sec, rpi2 gets about 39 Mbits/sec. In all cases there is a 5-9 Mbits/sec average higher increase with the new firmware. Pretty significant!

Note that if I run these same tests where my macbook is the client but a virtual machine on my network is the server, I see transfer rates in the 500 Mbit/sec range (the -P 10 test gave 513 Mbits/sec and the UDP test -b 1000m gave 627 Mbit/sec). This at least demonstrates that my infrastructure is more than capable of higher transfer rates. Thank goodness for 802.11ac routers!

It is also worth noting a few differences between the pi and vm tests. One difference is that the virtual machine is on a direct-cabled server, whereas both the raspberry pi and macbook are on wireless. Thus for the VM test, only one traversal of wifi is needed, whereas macbook to pi tests involve two traversals.

Now, if I upgrade rpi1 I would expect the numbers to be equal. Right…?

Strangely, no. Even after the upgrade I see rates in the 30Mbits/sec for rpi1. Digging deeper I found one minor discrepency between the two – i set aside 256 MB for gpu ram on rpi1, but only 128 MB on rpi2. Even after switching rpi1 to 128 MB, I still see the same numbers on both. I noticed I had a wireless keyboard adapter on the slower pi – but removing it didnt affect the transfer rates.

Another difference is that there is a raspberry pi camera on the faster pi – it would be interesting to see if that affects things.

I tried testing the ethernet interfaces and found performance across the pis identical.

At the moment I’m not sure how to account for the roughly 25% higher wifi performance on one pi over the other.