log in | register | forums


User accounts
Register new account
Forgot password
Forum stats
List of members
Search the forums

Advanced search
Recent discussions
- Archive Edition 27:2 reviewed (News:)
- RISC OS 5.30 arrives (News:12)
- Rougol Talk May 2024 - Andy Vawer (News:)
- Rougol May 2024 meeting on monday with Andy Vawer (News:1)
- WROCC May 2024 meeting - Gerph talks games (News:)
- Drag'n'Drop 13i3 edition reviewed (News:1)
- Wakefield Show 2024 in Pictures (News:5)
- April 2024 News Summary (News:2)
- Upgrading your RISC OS system to 5.30 (News:2)
- WROCC May 2024 meeting on wednesday - Gerph talks games (News:)
Related articles
- ABug provide more interesting retro talks to pass the time this Christmas
- Software Preservation at the ROUGOL Show
- Arculator updated to add A4 emulation and more podule support
- Acorn Preservation Team want to help archive your old data before it's too late
- Pass the time this Christmas with a selection of RISC OS and BBC Micro talks
- Cloverleaf drops promise of including retro games with Kickstarter rewards (Updated)
- Organizer 2.29 Reviewed
- Episode VI - Return of the JFPatch
- Wakefield show postponed due to COVID-19
- RISC OS London Show 2019
Latest postings RSS Feeds
RSS 2.0 | 1.0 | 0.9
Atom 0.3
Misc RDF | CDF
View on Mastodon
Site Search
Article archives
The Icon Bar: News and features: Victims of our own success...

Victims of our own success...

Posted by Richard Goodwin on 15:51, 29/11/2000 |
Some of you may have noticed that the site has been down for the past couple of nights. This appears to be because we are now running too many web server processes, so I've had to shut down some test servers and basically put our machine on a diet.

Extra memory has been ordered and should be installed over the weekend or early next week. Until then things should keep running, but there's a slight possibility that at peak times you might get errors, or the server may go down again :(

  Victims of our own success...
  (15:52 29/11/2000)
  Rob Kendrick (05:08 30/11/2000)
    Richard Goodwin (09:37 30/11/2000)
      Rob Kendrick (11:26 1/12/2000)
        Richard Goodwin (11:35 2/12/2000)
          Rob Kendrick (21:53 2/12/2000)
            Rob Kendrick (21:54 2/12/2000)
              Richard Goodwin (10:18 4/12/2000)
                Rob Kendrick (16:19 4/12/2000)
                  Matthew (13:21 8/12/2000)
                    Rob Kendrick (23:43 8/12/2000)
Richard Goodwin Message #88184, posted at 15:52, 29/11/2000
Unregistered user It's intereting to note that although the server only has a 400Mhz processor - which was fairly fast when it was bought, but about a quarter of the top spec these days - the server's only using 5% of it's processor power and 95% of it's memory when not heavily loaded. Just goes to show it's memory and bandwidth that count when you set up a web server, not how fast the processor is.
  ^[ Log in to reply ]
Rob Kendrick Message #88185, posted at 05:08, 30/11/2000, in reply to message #88184
Unregistered user This is because most webservers are IO-Bound, rather than CPU-Bound. Also, Apache is multi-process, which is crap. You'll get much better performance if you use an internally multiplexed webserver, such as thttpd or Zeus. (This is why both of these webservers are faster than Apache; rather than having a process per connection, which is slow because most Unices have a nasty penalty for starting new threads. thttpd and Zeus have one process per CPU)

As an example once, I showed how a AMD K6-II/350MHz box running Zeus outperformed a tuned Apache running on a dual 550MHz PentiumII for a laugh once (both Debian GNU/Linux)
  ^[ Log in to reply ]
Richard Goodwin Message #88186, posted at 09:37, 30/11/2000, in reply to message #88185
Unregistered user Zeus is faster than Apache for flat-file stuff, but our sites use server includes where Apache pulls back a lot of ground speed-wise.

Zeus is commercial, and can afford to be more optimised. Apache's price is right, easily configured and has been running fine until I went overboard with new domains.

There's got to be some reason why it's still at 60% (http://www.netcraft.com/survey/) and the mighty Microsoft only 20%. Zeus isn't even at 3%.
  ^[ Log in to reply ]
Rob Kendrick Message #88187, posted at 11:26, 1/12/2000, in reply to message #88186
Unregistered user I think you'll find Zeus is faster at everything, not just static data. Apache is slow. If you don't want to pay any money (Zeus saves money, even though it isn't free; total cost of ownership etc) then use thttpd which is as fast as Zeus, and free (if not as flexible).

If you are refering to the Netcraft survey, the reason Apache is at the top is because it's free, not because it's good. Zeus have a problem with Netcraft, every time they submit the new sites, (some 2 million sites every 3 months) it crashes Netcraft, so they've been banned from submitting new entries.

Server side includes are evil and badly thought out anyway.
  ^[ Log in to reply ]
Richard Goodwin Message #88188, posted at 11:35, 2/12/2000, in reply to message #88187
Unregistered user That's a very sad way of thinking, but that's your right I guess.

Apache isn't top just because it's free - you've already pointed to another free product, which is nowhere near as popular. Apache is at the top because of it's flexibility and ease of use; it's also stable and fairly secure (compared with some of the alternatives anyway). Even in environments where IIS has already been bought and in use, Apache is often brought in to replace it. I'm working for one such company in the heart of London's banking district right now, who writes software and websites for various other companies. They've considered the alternatives, and Apache's the server *of* *choice*.

I did say that Apache pulls back a lot of ground speed wise, not that it was faster. Apache is a general purpose web server hacked into an existing piece of software (hence the name - from "a patchy server"), Zeus was written from scratch for speed. Apache was designed - modified - for flexibility. Alas speed is not the only criteria required for web serving. If Apache is so slow, why do people continue to use it? Can that be just because it's free? Well we've already covered that one. Apache's speed, even on an underspecced box (compared with my desktop NT4 workstation at work, Neo's a tiddler), is more than adequate.

And as for Zeus crashing the Netcraft survey, that sounds like bull I'm afraid. Netcraft basically does a HTTP fetch and examines the results; if that crashes the survey tool, then it'd likely crash most web browsers too, which would make Zeus a lousy web server if it were true. I've submitted various RISC OS web servers to Netcraft and, apart from mistakenly reporting RISC OS as a BSD varient (most likely because of the firewall), was quite able to tell me it was Netplex or whatever.

Who said anything about server side includes anyway? The IconBar is run using PHP and Perl; we just configured Apache to run .html files as PHP for certain domains.

You've got a real downer on Apache, but maybe that's a RISC OS disease - it's popular so let's knock it. I've found it a joy to work with, and will continue to run it with no qualms.
  ^[ Log in to reply ]
Rob Kendrick Message #88189, posted at 21:53, 2/12/2000, in reply to message #88188
Unregistered user Perhaps you should download a demo of Zeus, and then try and say that Apache is easy to use and flexible. thttpd /isn't/ as flexible as Apache or Zeus, but it is as quick. And where it's needed (such as Demon's homepage service) it is used (although www.demon.net runs on Zeus, and Demon will be changing their homepages to Zeus when RIPE take away half their IP allocation). Zeus is growing at a faster rate than iPlanet and IIS are falling. Apache is falling, too. (Although you need to adjust the Netcraft survey results somewhat to get something near the truth due to a mathamatical error).

Zeus wasn't actually written to be fast. Zeus 1.0 wasn't as fast as the original httpd. It was however, written by two students who didn't have much memory in their computers, and didn't want to run Apache. You also seem to think that Apache is more flexible than Apache. This is an error.

Why do people still use Apache? Yes, because its free. Its actually far more capible than iPlanet, but not many people know this because the licence agreement for using iPlanet forbids you from publishing benchmarks etc.

If you give Apache bandwidth, it will use it. That's not the only issue when it comes to speed. The point is that a 386 can outperform a Pentium.

You further your catalogue of misunderstandings with your idea of how Netcraft works. It sends an HTTP request to the server; yes. However, this isn't what I said crashed Netcraft. What crashed Netcraft was Zeus submitting large numbers of websites to include in the survey. (Netcraft doesn't have a crawler; it only includes sites people have done a manual search on).

The reason Netcraft detects RISC OS as a BSD varient is by its stack. (RISC OS has a BSD style stack, rather than a Linux or a Windows one, for example).

Who said anything about server side includes? You did.

I don't have a real downer on Apache at all. 3 of my boxes run it. What I am saying, however, is that the reason your webserver has been having problems is that it runs Apache; both points here are undisputable. I mearly mentioned two other products that don't suffer this problem. You seem to be the person with the RISC OS disease; blindly believing that what you have is the best.
  ^[ Log in to reply ]
Rob Kendrick Message #88190, posted at 21:54, 2/12/2000, in reply to message #88189
Unregistered user s/Apache. This is an/Zeus. This is an/
  ^[ Log in to reply ]
Richard Goodwin Message #88191, posted at 10:18, 4/12/2000, in reply to message #88190
Unregistered user So let me get this straight - you say that SSI is badly thought out and attribute it to me, you say I'm blind because I like what I have when I've obviously checked out the competition and taken into account the opinions of others whose opinion I respect, and somehow I'm the blinkered one?

Apache suits my needs admirably. The problem with the box is that I got an all-in-one motherboard that takes half the 128MB SIMM for the shared video memory.
  ^[ Log in to reply ]
Rob Kendrick Message #88192, posted at 16:19, 4/12/2000, in reply to message #88191
Unregistered user I don't recall saying that SSI is badly thought out, neither did I attribute it to you. You may have read about the competition, but you obviously havn't taken it in. You are misinformed in numerous areas. I wouldn't say you're blinkered at all.

The second paragraph mearly shows how little you've read and understood.

I give up.
  ^[ Log in to reply ]
Matthew Message #88193, posted at 13:21, 8/12/2000, in reply to message #88192
Unregistered user Richard, you <em>did</em> mention server includes first - in your second comment;
Rob did say they were badly thought out in his reply...

Just because Apache is the most widely used doesn't make it the best (witness Windows, ObWebsite :) ).

I use Apache, I like it, it works, but I haven't looked at the alternatives (except laughing at IIS), and wouldn't like to make
any claims for or against it. And I don't think I need to worry about my server crashing due to too many processes... :-)
  ^[ Log in to reply ]
Rob Kendrick Message #88194, posted at 23:43, 8/12/2000, in reply to message #88193
Unregistered user I didn't say SSI was badly thought out. I said SSI was evil. There is a world of difference. The use of SSI /can/ be forgivable, but rarely is with the other techniques that are available today. Just consider what dammage SSI does to caching, for example...
  ^[ Log in to reply ]

The Icon Bar: News and features: Victims of our own success...