Re: What you get is what you want
Jean Rodrigues <[email protected]> Tue, 16 Nov 2004 10:04:14 -0300 (ART)
| Newsgroups | gmane.comp.java.seda.user |
|---|---|
| Message-ID | <[email protected]> |
--0-1844151569-1100610254=:75003 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit I got it ... thanks for the explanation. I had a problem similar to what you explained with seda but I solved it by avoiding blocking stuff. But an Asynchronous Completion Token implementation should do something similar, shouldn't it? Regards, Jean Rodrigues Quartz <[email protected]> wrote: > Why would we need a 'blocking enqueue' feature? > I just can't realize the needs of it ... > And, by the way, it should defeat the main proposal of SEDA which is to be scalable (everything > that blocks does not scale). A better example to get it clear: if an upstream stage performs 1000x faster than its downstream stage, you would need 1000x more threads in that downstream stage to be able to catchup. And that happens very often. That is called an over-producing stage and if it cannot block, if will loose the data which most obviously isn't acceptable, no more than a thousand threads per stage. So: -not everybody want a lossy systems, we need reliable message queues -practical systems don't have infinite queues -that's the only trivial backpressure mecanism. Other backpressure mecanism will always end up blocking a thread, unless you want to spin-crazy in a poll loop like a novice, wasting CPU. __________________________________ Do you Yahoo!? The all-new My Yahoo! - Get yours free! http://my.yahoo.com ------------------------------------------------------- This SF.Net email is sponsored by: InterSystems CACHE FREE OODBMS DOWNLOAD - A multidimensional database that combines robust object and relational technologies, making it a perfect match for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8 _______________________________________________ SEDA-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/seda-users __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com --0-1844151569-1100610254=:75003 Content-Type: text/html; charset=iso-8859-1 Content-Transfer-Encoding: 8bit <DIV>I got it ... thanks for the explanation.</DIV> <DIV>I had a problem similar to what you explained with seda but I solved it by avoiding blocking stuff.</DIV> <DIV>But an Asynchronous Completion Token implementation should do something similar, shouldn't it?</DIV> <DIV> </DIV> <DIV>Regards,</DIV> <DIV>Jean Rodrigues<BR><BR><B><I>Quartz <[email protected]></I></B> wrote:</DIV> <BLOCKQUOTE class=replbq style="PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #1010ff 2px solid">> Why would we need a 'blocking enqueue' feature?<BR>> I just can't realize the needs of it ...<BR>> And, by the way, it should defeat the main proposal of SEDA which is to be scalable (everything<BR>> that blocks does not scale).<BR><BR>A better example to get it clear: if an upstream stage performs 1000x faster than its downstream<BR>stage, you would need 1000x more threads in that downstream stage to be able to catchup. And that<BR>happens very often. That is called an over-producing stage and if it cannot block, if will loose<BR>the data which most obviously isn't acceptable, no more than a thousand threads per stage.<BR><BR>So:<BR>-not everybody want a lossy systems, we need reliable message queues<BR>-practical systems don't have infinite queues<BR>-that's the only trivi al backpressure mecanism.<BR><BR>Other backpressure mecanism will always end up blocking a thread, unless you want to spin-crazy in<BR>a poll loop like a novice, wasting CPU.<BR><BR><BR><BR><BR>__________________________________ <BR>Do you Yahoo!? <BR>The all-new My Yahoo! - Get yours free! <BR>http://my.yahoo.com <BR><BR><BR><BR><BR>-------------------------------------------------------<BR>This SF.Net email is sponsored by: InterSystems CACHE<BR>FREE OODBMS DOWNLOAD - A multidimensional database that combines<BR>robust object and relational technologies, making it a perfect match<BR>for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8<BR>_______________________________________________<BR>SEDA-users mailing list<BR>[email protected]<BR>https://lists.sourceforge.net/lists/listinfo/seda-users<BR></BLOCKQUOTE><p>__________________________________________________<br>Do You Yahoo!?<br>Tired of spam? Yahoo! Mail has the best spam protection around <br>http:// mail.yahoo.com --0-1844151569-1100610254=:75003-- ------------------------------------------------------- This SF.Net email is sponsored by: InterSystems CACHE FREE OODBMS DOWNLOAD - A multidimensional database that combines robust object and relational technologies, making it a perfect match for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8