INFO-VAX Sun, 30 Sep 2007 Volume 2007 : Issue 534 Contents: Re: Guidance with OpenVMS IA64 8.3 with Java 1.5 Problems RE: Guidance with OpenVMS IA64 8.3 with Java 1.5 Problems Re: Guidance with OpenVMS IA64 8.3 with Java 1.5 Problems ---------------------------------------------------------------------- Date: Sun, 30 Sep 2007 06:56:55 -0700 From: "Tom Linden" Subject: Re: Guidance with OpenVMS IA64 8.3 with Java 1.5 Problems Message-ID: On Sat, 29 Sep 2007 07:20:01 -0700, Main, Kerry wrote: >> -----Original Message----- >> From: mjjerabek [mailto:mjjerabek@gmail.com] >> Sent: September 27, 2007 4:56 PM >> To: Info-VAX@Mvb.Saic.Com >> Subject: Guidance with OpenVMS IA64 8.3 with Java 1.5 Problems >> >> I have been given the task of determining whether I should port/ >> rewrite a rather large HMI program from x-windows and 'C' to java. >> >> I went through an install of java 1.5 on an rx2620 duo core with 3gb >> ram, a radeon video card, mouse, keyboard, and OpenVMS 8.3. I followed >> the patching and tuning recommendation from the java release notes and >> from the java$check_environment.com DCL file. I then started running >> the jdk demo applications to get an idea of java performance versus >> the same applications running on an 2.8ghz P4 XP PC with 256mb of ram. >> >> The performance of these java apps on the OpenVMS ia64 system was >> extremely poor as compared to a much slower PC running XP. I followed >> these tests up with a re-tune of the VMS authorize and sysgen >> parameters, taking these parameters from their recommended values to >> ridiculously large values (1gb wsquo, wsmax 1gb, everything big). The >> performance of these apps got better, but had lots of rough spots. >> >> For example, the java2d app has a tabbed dialog where each tab >> displayed different classes of graphics operations. When the app ran, >> the graphics operations ran OK, but when I tried to select a different >> dialog tab, it took 4 minutes 30 seconds (clock time) to switch dialog >> tabs and start the display of the new graphics operations. Once the >> tab was fully rendered, the new graphics operations ran at a >> reasonable speed. >> >> I tried to get help from HP by sending en email to java-bugs@hp.com >> and exchanged some email with an HP engineer. The engineer blamed the >> applications, and not java. I pointed out that these applications came >> with the JDK, but that made no difference. >> >> I also tried to use the java plugin with the CSWB/Mozilla. I tried to >> display and operator a web page with embedded java. In my case, it was >> an hp procurve managed switch management interface web page. These >> simple html/java pages hung after a few java apps were loaded and run. >> The HP engineer point to the procurve web page which was generated in >> 2000 and specified that only IE5 and the MSVM could be used to display >> the procure pages (lame excuse from HP). >> >> What I am looking for (I guess) is guidance from experienced people on >> how to get performance out of java under OpenVMS Ia64 and Alpha >> systems. I need access to better information than what HP chooses to >> provide. >> >> I could use the following: >> >> 1) VMS tuning advice for java. What are typical authorize and sysgen >> values for java usage. >> >> 2) More example applications to test with. My application will be a >> rather large app and the demo apps that come with the JDK are not. >> >> 3) Suggest some URL's, books, etc that will guide me. >> >> Also, anyone who is willing to describe their experience in using java >> with OpenVMS and have a conversation with me on or off-line, >> please ... >> >> All help will be appreciated and acknowledged. >> >> Mjjerabek > > > Just curious, but can you expand on why your company wants to re-write > something > already Working to Java? > > Is it possible to install the sample programs with any performance > monitoring tools > to determine where the issues are? That is something you would likely > want to do > in porting anything to any platform. > > Btw, I assume that you have the latest VMS V8.3 maint patches installed? > ftp://ftp.itrc.hp.com/openvms_patches/i64/V8.3/I64_V83I_MASTER_ECO_LIST.txt > > Also, check the typical OpenVMS OS perf tuning stuff e.g. no disk frag, > highwater > marking off, and you might also look at a basic difference between VMS > and Windows/UNIX > ie. write back (speed - Unix-Windows)) vs write thru (data safety - > OpenVMS). If you > would like to setup OpenVMS for same behaviour (write-back), there is a > dynamic system > parameter that will allow you to do this (I forget the actual parameter) > > There is also a cookbook for OpenVMS Java that also had a number of > suggestions. Could also be a poor implementation of Java. Isn't it an interpreter? Takes some skill (and good design) to write a good interpreter. > > > Regards > > > Kerry Main > Senior Consultant > HP Services Canada > Voice: 613-592-4660 > Fax: 613-591-4477 > kerryDOTmainAThpDOTcom > (remove the DOT's and AT) > > OpenVMS - the secure, multi-site OS that just works. > > > -- PL/I for OpenVMS www.kednos.com ------------------------------ Date: Sun, 30 Sep 2007 15:54:46 +0000 From: "Main, Kerry" Subject: RE: Guidance with OpenVMS IA64 8.3 with Java 1.5 Problems Message-ID: > -----Original Message----- > From: Tom Linden [mailto:tom@kednos.company] > Sent: September 30, 2007 9:57 AM > To: Info-VAX@Mvb.Saic.Com > Subject: Re: Guidance with OpenVMS IA64 8.3 with Java 1.5 Problems > [snip..] > > There is also a cookbook for OpenVMS Java that also had a number of > > suggestions. > > Could also be a poor implementation of Java. Isn't it an interpreter? > Takes > some skill (and good design) to write a good interpreter. > > Java takes some tuning on all platforms. Same goes for any language if you = want it to run optimally. If there is a difference in performance between one platform= and another it does not automatically mean it is an implementation issue. Even if one plat= form runs ok in single user mode, it does not mean it will scale to handle large numbers of= users. The base language implementation is only one of many factors you would typi= cally look at. Btw, I forgot about this book on OpenVMS and Java - http://www.theminimumyouneedtoknow.com/about_java_on_openvms.html http://www3.sympatico.ca/n.rieck/links/cool_openvms.html#TMYNTK-About-Java-= on-OpenVMS (Neil's review of Java OpenVMS book) http://www.theminimumyouneedtoknow.com/ (other OpenVMS App books for the re= ference shelf) Btw, here is the Java OpenVMS cookbook as well:(bit dated, but likely some = components are still applicable) http://h71000.www7.hp.com/ebusiness/optimizingsdkguide/optimizingsdkguide.p= df http://h71000.www7.hp.com/ebusiness/optimizingsdkguide/optimizingsdkguide.h= tml Regards Kerry Main Senior Consultant HP Services Canada Voice: 613-592-4660 Fax: 613-591-4477 kerryDOTmainAThpDOTcom (remove the DOT's and AT) OpenVMS - the secure, multi-site OS that just works. ------------------------------ Date: Sun, 30 Sep 2007 13:29:57 -0400 From: JF Mezei Subject: Re: Guidance with OpenVMS IA64 8.3 with Java 1.5 Problems Message-ID: <3a202$46ffdd1c$cef8887a$12224@TEKSAVVY.COM> Tom Linden wrote: > Could also be a poor implementation of Java. Isn't it an interpreter? > Takes > some skill (and good design) to write a good interpreter. Has JAVA ever been accused of having good performance ? Back then, people were complaining about ALLIN1 being a hog. Looks to me that JAVA is orders of magnitudes slower and more resource hungry. ------------------------------ End of INFO-VAX 2007.534 ************************