Re: A4D scalability

Guess I will have to type a bit :-)

In our setup we were using 4D 2003 with the ITK Webserver (modified to  
handle partial requests, continuus requests, keep-alive, etc.).

Although we never used the full setup as we did not need it, here is  
the basic setup we had:


                                THE INTERNETS
----------------------------------------------------------------------------
|
                                                                           |
|                                  
Gateway                                  |
|                       Linux System with Apache  
2.2                       |
----------------------------------------------------------------------------
|                        |                        
|                        |
|  Public HTTP Server 1  |  Public HTTP Server n  |   Private HTTP  
Server  |
|  4D 2003 / Apache 2.2  |  4D 2003 / Apache 2.2  |  4D 2003 / Apache  
2.2  |
----------------------------------------------------------------------------
|
                                                                           |
|                             4D Server  
2003                               |
|                            on Windows 2000  
WS                            |
----------------------------------------------------------------------------

Details

Gateway:

This machine ran Apache 2.2 and some Linux (RedHat I think it was).

Apache had rewrite rules to do a round robin Load Sharing of as many  
public
  HTTP Cients as you wanted, hiding them from the user, so no w1, ww2  
etc.

The round robin list ws automatically handled by apach, if on http  
server sent an error (Bad Gateway or other) it would automatically be  
removed from the list. The Script would then poll the client every 60  
Seconds to see if he was online again to then put him back in the list.

On the Client a script would execute killing hte 4D Client process and  
restarting it.

The system was designed to run on a DELL blade System with up to 6  
blades. The private HTTP Server was for Backoffice and background task.

The biggest load we ever had was some 15.000 hits per hour, and using  
this setup we had an uptime service level of 99.74 % (externally  
Tracked with RedAlert) once the teething problems were solved (2-3  
months), the remaining .26 were scheduled downtimes. The individual  
clients only got to 96% but due to the automatic recovery system that  
did not matter much.

Sadly we never got to gear it up to full glory, as the company we did  
this for went bankrupt and their owner ran with whatver money was left  
to oen a cafe on Crete...

The whole setup was scrapped and sold in the bankruptcy.

A simple, 1-machne version fo the whole system is still operational.

I am currently working on findng a new customer for the system in  
order to bring it up to current versions.

In the 18 Months we had the system running it never failed us and only  
ran at a fraction of its potential.

Basically there were only 2 points of failiure left:

The Linux Gateway which, when we swithced it off, had an uptime count  
of 690 days!

The 4D Server, this one had NO stored procedures nd ran nothing but 4D  
Server, we had one crash of it early on in the project, we never found  
out why ...

The server scripting was all done by Michael Bond.

Cheers,

Alex


_______________________________________________
Active4D-dev mailing list
Active4D-dev@...
http://mailman.aparajitaworld.com/mailman/listinfo/active4d-dev
Archives: http://mailman.aparajitaworld.com/archive/active4d-dev/