|
midrange.com
I guess it depends on your application, Brad. I just ran an informal test
on my machine. I loaded my server from scratch, which took about 45 seconds
or so of CPU time - it takes a while to load all that good stuff. Then I
made 200 or 300 hits to my server, the total was about another 30 seconds or
so. Let's see my results from here:
QBASE SBS 2 0 2.2 0 .0
LINLST BCH 2 50 3.6 0 3.6
QPADEV0002 INT 3 20 .7 1 .0 0 .0
QSYSSCD BCH 3 20 .3 0 .0
QEJBSBS SBS 2 0 .0 0 .0
DEFAULT_SE BCI 2 25+ 24.2 0 10.4
QEJBADMIN BCI 2 25+ 58.1 0 .1
QEJBMNTR ASJ 2 25 .0 0 .0
Note that LINLST is the RPG program actually servicing the result. He does
all the actual database I/O. So let's see. I'm going to do another 200
hits.
QBASE SBS 2 0 2.2 0 .0
LINLST BCH 2 50 10.6 0 2.2
QPADEV0002 INT 3 20 .7 2 .0 0 .0
QSYSSCD BCH 3 20 .3 0 .0
QEJBSBS SBS 2 0 .0 0 .0
DEFAULT_SE BCI 2 25+ 44.5 0 6.6
QEJBADMIN BCI 2 25+ 58.5 0 .1
QEJBMNTR ASJ 2 25 .0 0 .0
Okay, 200 hits (actually, I lost count so I did an extra 40 hits, so this
could be anywhere between 200 and 240 hits, but let's say 200 for argument).
I was putting out the following screen, all data being read from files on
the AS/400:
Kenosha Weekday Inbound
Train #300
Station Time
Kenosha ----
Winthrop Harbor ----
Zion ----
Waukegan 4:58
North Chicago 5:01
Great Lakes 5:05
Lake Bluff 5:10
Lake Forest 5:13
Fort Sheridan 5:16
Highwood 5:19
Highland Park 5:22
Ravinia 5:25
Ravinia Park ----
Braeside 5:27
Glencoe 5:30
Hubbard Woods 5:33
Winnetka 5:36
Indian Hill 5:38
Kenilworth 5:40
Wilmette 5:42
Central Street 5:45
Evanston 5:49
Main St. 5:51
Rogers Park 5:54
Ravenswood 5:59
Clybourn 6:06
Chicago 6:15
There are also a few buttons. What I did was page up and down through ten
different schedules, and I did that 10 times, for a total of 200 pages.
So, we can see that I consumed 7 seconds RPG time, 20 seconds server time
and pretty much zero administration time. Extrapolated to your benchmark of
7000 page hits, my application would consume about 700 seconds of HTTP
server time, and maybe 250 seconds of RPG server time. These numbers are
less than 2/3 those of your CGI application. They're more than 100 times
better in performance than your Java version, which doesn'teven do any
database I/O.
Now, things that can greatly increase those numbers include starting and
stopping server instances and creating new Java Server Pages. The latter in
particular can cause lots of work, because they have to be unloaded,
compiled, and reloaded. Have you been making lots of changes to your JSPs?
That number shows up in the QEJBADMIN number, I'm pretty sure.
The other load might be from how you're "Java-izing". As you can see, my
programs don't have nearly as much load per page - I'm talking overhead of
less than 100 msec a page, while you're seeing closer to a 30 seconds. I
don't know why, but you might want to start putting some trace statements in
your code to determine where your bottlenecks are.
Good luck!
Joe
BTW, for anybody who might wonder how to architect an RPG program that
performs so well as a web application, it's actually a revitalized green
screen program. That means it runs in green screen using a display file AND
as the server for the web, WITHOUT changes. Here's the same program run
interactively:
STNSTP-01 Stop Master List
3/01/01
SECJDP For Line: Union Pacific North Line
17:46:24
For Schedule: Kenosha Weekday Inbound
For Train: 00300
Station Time
Kenosha ----
Winthrop Harbor ----
Zion ----
Waukegan 4:58
North Chicago 5:01
Great Lakes 5:05
Lake Bluff 5:10
Lake Forest 5:13
Fort Sheridan 5:16
Highwood 5:19
Highland Park 5:22
Ravinia 5:25
Ravinia Park ----
Braeside 5:27
Glencoe 5:30
More...
F3=Exit F19=Previous F20=Next
Pretty cool, eh? This is what I teach...
<grin>
> -----Original Message-----
> From: owner-java400-l@midrange.com
> [mailto:owner-java400-l@midrange.com]On Behalf Of Stone, Brad V (TC)
> Sent: Thursday, March 01, 2001 2:35 PM
> To: 'JAVA400-L@midrange.com'; ign_list@ignite400.org
> Subject: Java and CPU
>
>
> This I found interesting.
>
> We have an AS/400 that is a web server only. We're running CGI to do
> customer service online for multiple customer/catalogs. Right now it's
> about 30 sites (all using the same 3 or 4 programs, webfacing controled by
> config files, SSIs, etc..).
>
> Some CPU stats for the normal HTTP Server: (jobs entered system 2/14/2001)
> Subsystem/Job Type Pool Pty CPU
> CSO BCH 2 25 1516.4
> CSO BCI 2 25 .1
> CSO BCI 2 25 .3
> CSO BCI 2 25 69.0
> CSO BCI 2 25 26.1
> CSO BCI 2 25 727.4
>
> Now, I have also have websphere running. I've been trying to
> duplicate this
> CGI system using servlets and and making progress (no wonder I'm asking so
> many questions!).
>
> Anyhow Here's the CPU stats for the default WAS server that is processing
> the servlets: (entered system 2/19/2001)
>
> Subsystem/Job Type Pool Pty CPU
> DEFAULT_SE BCI 2 25+ 2847.7
> QEJBADMIN BCI 2 25+ 6203.9
> QEJBMNTR ASJ 2 25 .1
>
> The HTTP server has received about 7000 hits since it started on
> 2/14/2001.
> This server serves up those 30 sites and is available 24/7.
>
> The WAS server hasn't even broken a couple hundred hits since it's only me
> testing when I have time (started on 2/19/2001). It's not
> available to the
> interent and I am the only developer on the box.
>
> The Java version of the application isn't querying customer service data
> yet, I'm simply building the web pages. The CGI programs run dynamic SQLs
> to select and display data depending on customer signon information.
>
> The numbers compared to just the default WAS server may be close, but
> remember, it's doing about 1/100th of the work right now.
>
> Any thoughts, ideas, or otherwise? I'm really wondering why so much more
> CPU is used. If I were to take my app and java-ize it, it looks like it
> would bury the current machine we are using.
>
> Last question, does WAS start more instances when it needs to
> them to handle
> more requests like the HTTP server does? What about the new
> Apache server,
> does it start more jobs to handle more requests?
>
> Thanks!
>
> Brad
>
> +---
> | This is the JAVA/400 Mailing List!
> | To submit a new message, send your mail to JAVA400-L@midrange.com.
> | To subscribe to this list send email to JAVA400-L-SUB@midrange.com.
> | To unsubscribe from this list send email to
> JAVA400-L-UNSUB@midrange.com.
> | Questions should be directed to the list owner: joe@zappie.net
> +---
>
+---
| This is the JAVA/400 Mailing List!
| To submit a new message, send your mail to JAVA400-L@midrange.com.
| To subscribe to this list send email to JAVA400-L-SUB@midrange.com.
| To unsubscribe from this list send email to JAVA400-L-UNSUB@midrange.com.
| Questions should be directed to the list owner: joe@zappie.net
+---
As an Amazon Associate we earn from qualifying purchases.
This mailing list archive is Copyright 1997-2024 by midrange.com and David Gibbs as a compilation work. Use of the archive is restricted to research of a business or technical nature. Any other uses are prohibited. Full details are available on our policy page. If you have questions about this, please contact [javascript protected email address].
Operating expenses for this site are earned using the Amazon Associate program and Google Adsense.
