Comments on: Scalability != concurrency http://www.sauria.com/blog/2007/08/25/scalability-concurrency/ Open Source, Modern Programming Languages, OS X, Photography, and ... Sun, 27 Jul 2008 01:03:05 +0000 http://wordpress.org/?v=2.5.1 By: Jasen Halmes http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6414 Jasen Halmes Wed, 29 Aug 2007 21:12:06 +0000 http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6414 I think that your core argument is that you can't achieve effective concurrency with today's JVM architecture? That may be true, but there are alternate ways to achieve concurrency without depending on the JVM design. When you use a multiple commodity box solution you get the scale and the potential for concurrency. With an application fabric, like the <a href="www.appistry.com" rel="nofollow">Appistry EAF</a> you can abstract the hardware from the Java application. You therefore have the ability to take advantage of massive concurrency (and for a much cheaper price tag than buying bigger boxes). For example, Google uses the map/reduce pattern to scale out their problems concurrently and recombine the answers. Appistry provides a productized platform to get that kind of massive scale for Java (and C/C++/.NET). -jasen I think that your core argument is that you can’t achieve effective concurrency with today’s JVM architecture? That may be true, but there are alternate ways to achieve concurrency without depending on the JVM design. When you use a multiple commodity box solution you get the scale and the potential for concurrency. With an application fabric, like the Appistry EAF you can abstract the hardware from the Java application. You therefore have the ability to take advantage of massive concurrency (and for a much cheaper price tag than buying bigger boxes).

For example, Google uses the map/reduce pattern to scale out their problems concurrently and recombine the answers. Appistry provides a productized platform to get that kind of massive scale for Java (and C/C++/.NET).

-jasen

]]> By: Isaac Gouy http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6409 Isaac Gouy Wed, 29 Aug 2007 03:30:16 +0000 http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6409 <b>IF</b> I'll just say that I think you've overstated your position. IF

I’ll just say that I think you’ve overstated your position.

]]> By: Ted Leung http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6407 Ted Leung Wed, 29 Aug 2007 02:18:43 +0000 http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6407 If the machinery for that gets in the way of making the VM efficient for actors, then I consider that a fatal flaw. We know that any computational model can emulate any other. But efficiency matters. If the machinery for that gets in the way of making the VM efficient for actors, then I consider that a fatal flaw. We know that any computational model can emulate any other. But efficiency matters.

]]> By: Isaac Gouy http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6404 Isaac Gouy Tue, 28 Aug 2007 22:40:32 +0000 http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6404 Does needing the machinery for volatile or for enforcing the memory model make the VM fatally flawed for actor style concurrency - or just a pain? It surprises me that someone managed to implement Occam style concurrency for JVM - but they have <a href="http://www.booksonline.iospress.com/Content/View.aspx?piid=5982" rel="nofollow">Integrating and Extending JCSP</a>. "Fatally flawed" sounds so very definite, so very conclusive. Does needing the machinery for volatile or for enforcing the memory model make the VM fatally flawed for actor style concurrency - or just a pain?

It surprises me that someone managed to implement Occam style concurrency for JVM - but they have Integrating and Extending JCSP.

“Fatally flawed” sounds so very definite, so very conclusive.

]]> By: Ted Leung http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6398 Ted Leung Tue, 28 Aug 2007 06:45:03 +0000 http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6398 Isaac, Well you could start with the fact that you don't need any of the machinery for volatile, or for enforcing the Java Memory Model. Immutability and shared nothing get you a lot. Isaac,

Well you could start with the fact that you don’t need any of the machinery for volatile, or for enforcing the Java Memory Model. Immutability and shared nothing get you a lot.

]]> By: Reid http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6387 Reid Mon, 27 Aug 2007 03:48:18 +0000 http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6387 <em>(commenting here since you didn't supply a URL for the original article)</em> Just a nit. Azureus is the most popular BitTorrent client, and it's a Java app. I couldn't find a lot of stats to back this up; these two pages are the best I can find: <a href="http://torrentfreak.com/limewire-most-installed-p2p-application-bittorrent-clients-runner-up" rel="nofollow">torrentfreak.com "Most installed P2P apps"</a>, and <a href="http://www.associatedcontent.com/article/256053/azureus_the_best_bittorrent_client.html" rel="nofollow">an assertion</a> by Associated Content. (commenting here since you didn’t supply a URL for the original article)

Just a nit. Azureus is the most popular BitTorrent client, and it’s a Java app. I couldn’t find a lot of stats to back this up; these two pages are the best I can find: torrentfreak.com “Most installed P2P apps”, and an assertion by Associated Content.

]]> By: Isaac Gouy http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6384 Isaac Gouy Sun, 26 Aug 2007 16:11:47 +0000 http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6384 <em>"...the VM is fatally flawed when it comes to actor style concurrency"</em> It would be nice to see some explanation of that assertion. “…the VM is fatally flawed when it comes to actor style concurrency”

It would be nice to see some explanation of that assertion.

]]> By: J Aaron Frr http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6383 J Aaron Frr Sun, 26 Aug 2007 13:47:35 +0000 http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6383 "Hmm, Erlang lab, anyone?" You propose it, I'll +1 it! :-) “Hmm, Erlang lab, anyone?”

You propose it, I’ll +1 it! :-)

]]> By: Matt Croydon http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6382 Matt Croydon Sun, 26 Aug 2007 13:04:19 +0000 http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6382 I've posted <a href="http://www.postneo.com/2007/08/26/there-is-an-erlang-community-its-just-smaller-than-youre-used-to" title="There is an Erlang community, it’s just smaller than you’re used to" rel="nofollow">my thoughts on the Erlang community</a> that you might want to check out. I’ve posted my thoughts on the Erlang community that you might want to check out.

]]> By: Sam Ruby http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6381 Sam Ruby Sun, 26 Aug 2007 11:22:48 +0000 http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6381 <blockquote>This is actually 2 problems. There’s the issue with the libraries, and there’s the issue with the community that did/didn’t produce the libraries.</blockquote> Either that or Sam's an idiot. <code>apt-get install <a href="http://packages.ubuntu.com/cgi-bin/search_contents.pl?searchmode=filelist&word=erlang-examples&version=feisty&arch=all" rel="nofollow">erlang-examples</a> <a href="http://packages.ubuntu.com/cgi-bin/search_contents.pl?searchmode=filelist&word=erlang-doc-html&version=feisty&arch=all" rel="nofollow">erlang-doc-html</a> <a href="http://packages.ubuntu.com/cgi-bin/search_contents.pl?searchmode=filelist&word=erlang-manpages&version=feisty&arch=all" rel="nofollow">erlang-manpages</a></code>

This is actually 2 problems. There’s the issue with the libraries, and there’s the issue with the community that did/didn’t produce the libraries.

Either that or Sam’s an idiot.

apt-get install erlang-examples erlang-doc-html erlang-manpages

]]> By: ingo http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6380 ingo Sun, 26 Aug 2007 11:06:51 +0000 http://www.sauria.com/blog/2007/08/25/scalability-concurrency/#comment-6380 What does "designed from the ground-up assuming that objects typically are immutable and serializable" mean for the VM? What does “designed from the ground-up assuming that objects typically are immutable and serializable” mean for the VM?

]]>