Well stated, Fred.  I want to make one extra point about the case where a
database call is involved.  Even if a single call is only .001 times faster,
I do not feel that the performance difference is always negligible if you
are developing a multi-user system.  All extra CPU cycles used to make the
EJB call rather than a local call are CPU cycles that cannot be used for
something else.  For example, the CPU cycles cannot be used by another
application sharing the CPU, a database server sharing the same CPU, or
another thread within the same application sharing the same CPU.  So while
the response time of the specific call may not be noticeably effected, the
total throughput of the system may be noticeably effected.

Mike Bresnahan

P.S. I am CCing you because I have yet to see one of my posts make it
through the list.  Should I be seeing my own posts?  Do you get bombarded
with out-of-office emails after a post?

> -----Original Message-----
> From: A mailing list for Enterprise JavaBeans development
> [mailto:[EMAIL PROTECTED]]On Behalf Of Fred Loney
> Sent: Sunday, April 07, 2002 3:49 PM
> To: [EMAIL PROTECTED]
> Subject: Re: SSB vs. Singleton
>
>
> There was a related thread on this last week; cf.
> http://swjscmail1.java.sun.com/cgi-bin/wa?A2=ind0204&L=ejb-interest&D=0&;
> P=1833 and ensuing discussion.
>
> Regarding speed, it is important to consider the execution context.
> Presumably the example in the message cited below is a simple in-memory
> operation. On the other hand, if the SSB call hits the database, as is
> often the case, then the time is dominated by IO rather than the method
> call differential. So, to extend the example, if the actual duration of
> the call under load is as follows:
>
> 1.0000001 sec with a Java class
>
> 1.001 sec with a SSB
>
> then using the Java class is less than .001 times faster than a SSB, not
> 10,000 times faster. There is no pat rule for which is better, but in an
> EJB project it is preferable to look at code integrity and quality of
> service first, since it is usually easier to analyze than performance.
>
> Fred Loney
> Spirited Software, Inc.
> www.spiritedsw.com

===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff EJB-INTEREST".  For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".

Reply via email to