In my regular job I'm responsible for a website that's accessed by people all over the world (and yes..I realize that applies to OpenNTF too!). We did some analysis to measure response time and availability of the site around the world. Our servers are in US and we've seen consistenty that performance in Asia Pacific region is almost 3 to 5 times slower than US. One of the solutions recommended was to compress some of the content using gzip compression. For those that are not familiar with it, you can learn more about it on http://www.http-compression.com but in short, here's the definition from their website:

"According to RFC 2616, Internet HTTP Compression is a method to send, from the Web server, an HTTP response message in compressed format to a requesting Web browser. This technology assumes that the Web server is capable of encoding the outbound content and the Web browser is capable of (automatically) decoding the received content. HTTP Compression saves transfer data volume and speeds ups Web page load time. "

As some of you may know, Domino 7 has this capability, but its only exposed in Domino Web Access There have been a few posts/complaints about the limited capability and there is an undocumented way to enable gzip for all NSFs posted by Manfred Dillman which you can use at your own risk. There are of course numerous products available that act as proxies in front of your web server and do the compression and on OpenNTF we did try Web Booster from Puakma. Compression can make a significant difference in performance of the website but it seems there are some side effects which may be more trouble than its worth. I'm wondering if others have experienced similar problems with performance in different countries and also how the performance of OpenNTF is in other countries in Asia/Pacific.

1. andy b10/06/2006 02:32:42 PM
Homepage: http://andy.the-broyles.com

I have attempted to use the 'hidden' Domino 7 Zlib with horrible results, even with the default settings.

First was the instablity of the site. The site I tested on is a fairly heavy use site, ~30,000 hits/day, and the HTTP service continually crashed with no warning or recourse (full domino service restart required.)

Next was an odd and inconsistent delay in preparing the pages...i figured that this was just the natural cost of compression...the site is heavily text with only three small images. The compression was good, 64K down to ~9K and the performance improvement, when it was working, was great. Some times the page would make it to the browser without delay, other times it was upwards of 30-45 seconds before it was delivered. I used MS Fiddler to watch what was going on and there never seemed to be any rhyme or reason to the delays.

Finally, I had intermittent problems with the delivery of Javascript files. Every so often, during testing, I would get an error in either FireFox or IE (6 or 7) that looked like the server/browser disagreed on how GZip worked (decompressed files were different from the pre-compressed ones.)

Can't say that using the undocumented 'feature' is a good thing from my perspective.

---

2. Stephan H. Wissel10/08/2006 08:21:26 PM
Homepage: http://www.wissel.net

Could you elaborate on the side effects?
stw

---

3. Brendon Upson10/23/2006 05:15:24 AM
Homepage: http://www.puakma.net/blog

The "side effects" as I understand it, were that some databases had their own built-in IP address logging. Because Web Booster accepts all http hits and forwards them on to Domino, Domino sees Web Booster as the only client. The work around is not to use the IP address of the connecting host, but the X-Forwarded-For http header which contains the client's IP address.

For our testing, gzip compression makes a HUGE difference to a site's performance, especially iNotes (Domino Web Access mail) and Quickplace. A large html page will compress up to 90% and compression time is so close to zero it's not worth factoring in (eg 6ms). You can try your own site here to see how it performs: http://www.puakma.net/puakma/boostertest.pma/URLReport?OpenPage&URL=http%3A//blog.openntf.org

---

4. Kamal Rij03/11/2007 04:46:22 PM

For people looking for the article on Manfred Dillman site.
From Web Archive
http://web.archive.org/web/20060428024952/http://www.madicon
.de/content/view/616/57/

---

5. Kamal Rij03/11/2007 04:57:22 PM

Correct URL
http://web.archive.org/web/20060428024952/http://www.madicon.de/content/view/616/57/

---

6. Starrow Pan04/08/2007 03:17:46 AM
Homepage: http://saintstarrow.blogspot.com

I'm from china.
The response time is definitely slower than that of local websites,but is still accectable,about 2 seconds i think

---

7. odysseas04/30/2007 01:13:39 PM

:- :- :- :-

---

8. Daniele06/08/2007 12:44:51 PM
Homepage: http://When HTTP Compression are stable ?

I've tested on ND7.02 FIXPACK2 and Domino 8.0 Beta 3...<br><br>But I have serius problem on rendering output but don't have any log in a domino console.<br><br>Tnx for all

---

9. a08/09/2007 08:48:10 AM
Homepage: http://aa

---