Being a basket of concurrency-related morsels too short to stand alone and to long to tweet.
[This is part of the Concur.next series.]
Node.js · This had escaped my attention somehow, but Avi Flax hailed me via twitter, asking me what I thought of Ryan Dahl’s presentation on node.js. Well, I watched this substantial chunk of video end-to-end without even twitching. Ryan’s presentation style could use a little polishing, but his node.js pushed a bunch of my software buttons.
Then, also on Twitter, Terry Jones pointed me at Simon Willison’s Node.js is genuinely exciting and yep, I think it is. I’m going to see how it reacts to 45G of data on a slow many-core machine.
(That Twitter thing, it’s gonna catch on, you just watch).
Optimism · I found myself smiling over and over as I read Greg Pfister’s Oh, for the Good Old Days to Come, which rambles through Moore’s Law and Second Life and concurrency in general. I think a lot of others will smile too.
On Not Stopping · Back around Hallowe’en, InfoQ ran another major Joe Armstrong presentation: Systems that Never Stop (and Erlang). If you’re thinking about concurrency and you’re not paying attention to what Joe Armstrong is saying, you’re Doing It Wrong.
Haff on Reinders on Parallelism · As my audience is sure to remind me, parallelism and concurrency aren’t the same things. That acknowledged, Gordon Haff over at CNET has a two-parter with a gentleman I’ve never met named named James Rainders, who is identified as “Intel’s leading spokesman on tools for parallelism”.
And indeed, the title is On Parallelism, even though the conversation spends much of its time in concurrency territory. Here are Part 1 and Part 2.
Methinks I should have a conversation with Mr. Reinders sometime.
Contributions
Comment feed for ongoing:
From: Sam Ruby (Nov 30 2009, at 12:49)
"... on a slow many-core machine"
Am I missing something? node.js is single core at the moment, with plans to support the Web Workers API in the future.
[link]
From: Tim (Nov 30 2009, at 12:57)
Hmm... In Ryan's speech, I seem to remember him explicitly claiming there's a thread pool, and I don't see how you could do that stat(2) trick he showed without one.
Will look more closely and report back.
[link]
From: Ben Hutchings (Nov 30 2009, at 18:41)
The way I understood it, there is a thread pool to run blocking operations but the Javascript code all runs in a single thread. So the average concurrency is likely to be slightly greater than 1.
[link]
From: Avi Flax (Dec 01 2009, at 16:05)
@Tim, I thought you'd like that.
I can confirm that there's a threadpool of some kind; a wrote a simple script that saturates both of my cores. I checked out Activity Monitor, and there are only two threads. Seems inline with Ryan's philosophy.
Script is here:
http://groups.google.com/group/nodejs/browse_thread/thread/e1d0bf03ee18543d
[link]