Messages
Reply
Edit
Delete07:36 am
3RY0GAZLS
| From: | Thierry Nivelet (FoxInCloud) |
| To: | ChrisR |
Here are some average metrics with FoxInCloud:
- Server time = application time + 200 to 400ms depending on form's complexity (mostly size and # of grids and cursors)
- Internet both ways = # 100 ms with a professional connection - FoxInCloud requests and response are very light: 9-char cookie and avg response around 30kb (similar to an image)
- Browser <= 200 ms
CPU power (# and performance) is the most critical on server side. Memory and bandwidth requirements are moderate.
With a very fast application you can easily get a global response time around .5 s.
More details here: http://foxincloud.com/facts-low-fat.php
Hello Chris,
+1 with Rick..
In my own application most of the requets are under 1 second. In that case, they are small request needing a small job.
But in my application, i do very longs jobs (between 3s and 20s). Of course the user who request this kind of job have to wait since the job is running. But, with the others wc.dll server (between 1 to 32), the others users are not disturbed with thoses longs jobs.
I think, in my application with 32 wc.dll server's and a 4 cpu with 16go ram, can have 500 concurrent users.
Now with a 2cpu and 2GO ram with 3 wc.dll server's, my application have 24 concurrent users and have 93 regulary users.
regards
Gilles
Chris,
There are never any simple metrics to dish out for x number of clients. It depends on what those clients are doing and how long individual requests take. We have a few customers that are receiving close to 2 million VFP hits against Web Connection servers a day at peak times of the year, but that doesn't meant that every app can handle that. It depends entirely one how long requests take and how much CPU they use, how the app is configured etc.
For most business scenarios Web Connection running on reasonably powerful machine can service a lot of requests. The key is to not have long requests that take more than a second to run. Once you start having long running requests scalability drops rapidly. there are ways to scale out from there to multiple machines to make that scenario work but it gets more complex quickly at that point.
+++ Rick ---
Hi Rick
Many thanks for your prompt response. In computing terms, that is ancient history!
But my question is current. Do you know whether there is any experience of running an application over a WAN (not on the Net) on Foxpro providing service to over 800 doctors' offices, or the equivalent in any other industry or domain? or any source of metrics on large scale Foxpro applications?
Many thanks
ChrisR
Not really. That article is extremely dated. There's not really that much useful data in it anyway as that was done as a case study for Microsoft and they stripped most of the technical content and the fact that this actually was a Web Connection (ie. third party not pure MS tech) application in the day.
It was updated in 1998, but the actual article is actually older than that I believe :-)
+++ Rick ---
-- thn (FoxInCloud)