I have to serve a calculation via algorithm, I've been advised to use a
child process per each opened socket, what I am about to do is something
like that:
var spawn = require('child_process').spawn;var child = spawn('node',
['algorithem.js']);
I know how to send argument to the algorithm pr
I think that the maximum number of opened files is the only unix's
limitation for the number of sockets, and this can be changed.
--
Att,
Alan Hoffmeister
2013/3/24 simo
> I have to serve a calculation via algorithm, I've been advised to use a
> child process per each opened socket, what I am
So, is it a recommended approach to spawn a new process for each opened
socket?
On Sun, Mar 24, 2013 at 11:24 PM, Alan Hoffmeister <
alanhoffmeis...@gmail.com> wrote:
> I think that the maximum number of opened files is the only unix's
> limitation for the number of sockets, and this can be chang
no. Definitely not.
Spawn a predefined number of workers (using the cluster api) and queue jobs
in the master process then send them to the workers when they are ready to
process.
On Monday, 25 March 2013 06:49:32 UTC+1, simo wrote:
>
> So, is it a recommended approach to spawn a new process for
a better option is spawning a separate cluster of algorthm workers as
service and using something like https://github.com/substack/dnode for
communication. spawning a new process takes tens of miliseconds, which
probably is to long for short living processes.
Am Montag, 25. März 2013 06:49:32
Thanks Floby, I appreciate your advice, do you know a good blog or source
that talks about workers and jobs queuing?
On Mon, Mar 25, 2013 at 10:52 AM, Floby wrote:
> no. Definitely not.
> Spawn a predefined number of workers (using the cluster api) and queue
> jobs in the master process then sen
You idea is great! but how can I manage load balance on the algorthm
workers? if I have 100 algorthm workers for example?
On Mon, Mar 25, 2013 at 12:35 PM, greelgorke wrote:
> a better option is spawning a separate cluster of algorthm workers as
> service and using something like https://github.
http://nodejs.org/api/cluster.html#cluster_how_it_works
Am Montag, 25. März 2013 12:08:27 UTC+1 schrieb simo:
>
> You idea is great! but how can I manage load balance on the algorthm
> workers? if I have 100 algorthm workers for example?
>
> On Mon, Mar 25, 2013 at 12:35 PM, greelgorke
> > wrote
Yes, I already found it, amazing tech
On Mon, Mar 25, 2013 at 1:22 PM, greelgorke wrote:
> http://nodejs.org/api/cluster.html#cluster_how_it_works
>
> Am Montag, 25. März 2013 12:08:27 UTC+1 schrieb simo:
>>
>> You idea is great! but how can I manage load balance on the algorthm
>> workers? if I
Hi Samir,
I solved that problem for CPU intensive jobs. I ended up with my own
control flow library, because no one has been gone in the route of
"queuing" task with a work-in-progress limit in a sane way before:
https://github.com/mcollina/kanban. There is async.queue, but it lacks some
flexibili
On Mon, Mar 25, 2013 at 11:14 AM, Matteo Collina
wrote:
> The number of worker processes you want to spawn in a given period of time
> is strictly dependent to the number of concurrent process you might want to
> run in your system. With node.js you can end up prettly quickly with load >
> 1, and
there is also threads-a-gogo lib which can be used to spawn threads. i
haven't use it yet (never really had to), but it is another option.
Am Montag, 25. März 2013 12:26:12 UTC+1 schrieb simo:
>
> Yes, I already found it, amazing tech
>
> On Mon, Mar 25, 2013 at 1:22 PM, greelgorke
> > wrote:
>
@Matteo Thanks, but as I am still at the beginning of this, I will need to
find a way to simulate lots of users, and to find a way to profile
performance, then I can decide what to do, if you can help me how to
simulate virtual users and how to profile performance of my node app in
real time, it wi
On Mon, Mar 25, 2013 at 2:17 PM, Samir Sabri wrote:
> @Mat you said ("load" should be divided by number of CPUs) isn't load
> automatically divided when I node cluster?
The load we were referring to is "load average":
http://blog.scoutapp.com/articles/2009/07/31/understanding-load-averages
--
Isn't there a cloud hosting provider who would auto scale my app? would it
be possible to stay away from load avarage details so that my cloud hosting
provider handle it? if so, who do you recommend?
On 25 Mar 2013 20:34, "Matt" wrote:
>
> On Mon, Mar 25, 2013 at 2:17 PM, Samir Sabri wrote:
>
>>
15 matches
Mail list logo
