Thursday, June 14, 2012

Hardware's hard facts

We just bought some toys to do a Tamias test deployment. We started small with just 15 nodes made of :
  • A Barebone chassis that has motherboard, CPU, chipset and loads of integrated stuff we won't use : the Shuttle XS36V
  • A 1TB 2.5" hard disk from Samsung
  • A 2GB RAM stick from Elixir (unrelated to the card game, for you french readers)
Now, the law of large numbers predicts that for millions of hardware pieces, a few of them will be faulty. I was really interested to see how lucky we would be with the Tamias hardware, and here is the result for the case of not so large numbers :

PartQty boughtQty faultyRemark
Shuttle XS36V152Won't boot from USB key, but installs from USB CD
Samsung 1TB HDD150so far so good
Elixir 2GB RAM1521 won't P.O.S.T., 1 fails during Debian install
So that's it, 13% failure for RAM. I was expecting 1 failure and hoping for zero... The motherboard could be rated at 0% failure because it actually works, but we had to boot from a USB CD. I would have to check if the BIOS release is the same as the other 13 pieces, but since we got it to work, we decided not to RMA them. Memory was RMAd and replaced swiftly, Memtest86 (using debian memtest86+ packages) has passed so we expect no problems here either...

Wednesday, May 2, 2012

FTDI USB driver woes on Debian LiveCD

Just a quick info that might be useful for some of you Xbee users. I wrote this little software to visualize and plot Xbee accelerometer data a while back.

Now, my brother tried to use it, and since he is not into Linux at the moment, I prepared a LiveCD for him using the Debian Live project. The reason why I did that is because the python program needs wxpython which is not included in the default Debian LiveCD.

The problem is that the kernel currently used (at the time of build: started Mon, 05 Mar 2012 20:21:01 -0600 and ended Mon, 05 Mar 2012 20:52:50 -0600) includes a buggy driver for the FTDI USB kernel driver. Details about the bug can be found here :

and changes needed to any python software (including my own xviz) that wants to work around this bug, so that you don't need to replace the kernel or rebuild the liveCD, are to look for a line like this:
 self._port = serial.Serial(self._comport, 9600)
and add a rtscts= parameter like this:
 self._port = serial.Serial(self._comport, 9600, rtscts=1)

This tells the driver to open the serial port using RTS/CTS flow control, which actually disables it, because the driver inverts both states, duh.

Tuesday, April 17, 2012

From information overflow, to a spotless mind

From Charles-Philippe Larivière...
It seems that I have been suffering an acute case of writer's block for the first few months of year 2012.

And what better way to start the year, than writing a non-technical vaguely metaphysical post. It is fun to see how I got to posting about this. While I write these first few lines, I still haven't chosen a title. Like B. Pascal put it a long time ago, the last thing I'll do is figure what will go first.

So, a piece of news not so recently suggested that the last DRAM chip maker in Japan, called Elpida, would go bankrupt. And then it did. You've never heard of it before, have you ? Well that's how I started to think again about memory.

As self-centred egotistical human beings, we basically build our memories with our own experience. Unfortunately, some things go well beyond our understanding and need taking a few thousands steps back to behold the big picture. That's where History comes into play. You could see it as a purveyor of fabricated memories, or as a conservation mechanism for the collective memory of mankind. Notice how History is both important and out of control at the same time. This is the setting that gave so much strength to the 1984 book in my own self-centered opinion.

The folk memory of fishermen in the Fukushima prefecture says that when a quake hits, a tsunami will follow. It also says that you had better go out to sea before the wave becomes to steep to climb. This is how a few boats were saved.

Our History is full of important lessons that should be remembered, and yet, we are now bathing in useless information, both figuratively and literally speaking. From the electromagnetic waves of wifi, cellphones, satellites, land TV and radio, to the clusterfuck of social media, RSS aggregation and whatever will come next - we are even drowning in information. And although the face value of it depends on how much each piece of information relates to you, chances are that not a billionth of it is worth your attention.

This is like dilution and filtering. The important information is diluted, your interest might not match those of dedicated mainstream portals, and obtaining what you want is frustrating. Meanwhile, the human brain received billions of sensory reports from all over the body, and yet, you could read these words without having to deal with it. Only valuable information made it to conscious level. Why can't our computers do this work of filtering for us, skipping over all that is not useful, skimming the things that we must read ?

Granted, this is hard from a computer software point of view. This is why regular users don't have it. Big companies do, however. That's how your webmail is free of monetary cost. Instead, computer programs are laughing at your information and selling select pieces to each other. Sandow Birk
But eventually, you would expect people governing us to employ such filters, be they humans or computers, to isolate what is important, focus on it, and rule accordingly. So, how come a megaquake more than 1000 years ago went forgotten ? How come a ridiculously low $5 millions cut can be decided on a tsunami advisory program ? When will people stop doing exactly the same wrong thing every once in a long while, that is, long enough to not remember a bit...

NOTE: the two paintings came to my attention through the "google filter" and this page.