Jason Cunliffe wrote:
I am wondering how fast zope responds for others, and what tips you have to improve regular performance (hardware, software, config).
When you start with a fresh Zope install, how many seconds (from mouse click) does it take:
1. To boot Zope 2. To enter the manage interface 2. To open index_html 3. To save the changes 4. To View this 5. Ditto for image uploads (say with 8K, 32K, and 64k Jpegs)
What's your best?
I haven't tried a clean install yet, but using ramdisk (as noted in another thread) I have acheived a 6 second load time with a 4MB Data.fs file. IMO, I don't think you can get better than this, due to python overhead in loading (though I will try a clean Data.fs just to see :-).
What's your worst? ~45 second load time on the 133 with 3 others running, and a Data.fs of around 3MB. a 45MB Data.fs loading in ~12 seconds on the dual 450s.
For making changes as above, it is pretty much as fast as the bandwidth allows IME. Busier sites will, of course, increase them somewhat, though for simple changes as such, I would not expect much increase (zope.org tends to get a bit slow ... but their server is reportedly pretty hammered).
Can you suggest an approach to build in this kind of local-action performance checking via a DTML method which uses timestamps comparison. For example a method which I can toggle on and off at different levels of the hierarchy to create a history file of certain actions. Does this already exist?
I've am testing Zope on two 'minimal platform PCs to check performance I can expect for clients in an upcoming project who I know will _not_ have the latest and best systems. My two current 'test' machines are:
1. P5-150Mhz,64Mb, Win95b 2. P233mhzmmx, 32mb, Win95b -laptop
I've been experimenting with these in 3 ways: with Zope local to each machine, over a small ethernet LAN and remote over 56k modem to hosted Zope site (codeit).
In a word: RAM. Sort of increasing processor speed, more memory is your better option. I have seen Zope use upwards of around 30-40 MB+ at times. For Database loading speed, it is kind of a trick. You only use it when loading and packing the database (AFAIK). The larger the site, and hence Data.fs, the more important it is. Then again, the more important memory becomes. Not trying to start an OS war, but: In the above machine examples, you would getter better performance by running Linux on the server, as it would allow you to avoid the overhead of a graphics system, use less memory, and more efficiently manage the memory you have. But this is generally true of servers, graphic systems by nature increase the memory needed. When you have a quarter gig of memory ona screaming machine, it is less of an issue, but in the examples you cited, a graphical interface taking up all or half of your ram is definitely an issue. On my p133, performance is definitely respectable, even though I am running 4 Zservers and squid over a DSL connection. As the sites get larger (or I can afford to) I will be replacing it with a much faster dual cpu system. I currently run two on dual pIIs w/256M and peformance is outstanding. Thus, I would say that it is hardware limitations in my case, not Zope itsself. I think many people tend to go overboard in their expectations. Zope is a dynamic appserver, not a static file server over http. Many people try to compare the two, and I do not consider it a valid comparison. It is like saying that a stripped-down-for-racing 'stock car' can be copared to the same 'model' of car on the street for speed. FWIW, I tried the LocalFS product, and it served up local files at a significant speed improvement, thus illustrating my point. In fact, I am contemplating the use of the LocalFS product for the basically static pages that I have on some sites to take advantage of this. That will have to wait for some testing though. :-) Hope that helps, Bill -- "They laughed at Columbus, they laughed at Fulton, they laughed at the Wright brothers. But they also laughed at Bozo the Clown." -- Carl Sagan