| |
| Publicado el 11,enero 2016 - 16:59 |
Hi,
I have to rewrite my old WD14 program.
Now all data are transferred via httprequest to the client.
Today I have make some speed tests:
With WD14 (native HFSQL - 20.000 records need 17 seconds)
With WD20 (HTTPrequest - the same 20.000 records need 30 seconds)
I use deserialize /serialize with the "psdBinary" option.
I know thats a lot of overhead because of the http-server in the middle but I am wondering if ther is a smarter (and faster) possibility to trransfer the data ? |
| |
| |
| | | |
|
| | |
| |
| Publicado el 13,enero 2016 - 07:37 |
Hi,
Perhaps you can tell us more about the server
1. type of webservice (soap / rest ) , (apache /IIS) , (.net/java/nodejs)
2. server os/ spec ( cpu / ram). is this dedicate server just to host webservice
3. the server database type
4. size of data per send . (number of records cannot tell) |
| |
| |
| | | |
|
| | |
| |
| Publicado el 13,enero 2016 - 07:58 |
| the faster to transfer large data is write to raw file in fixed length (don't use xml) , this way no need deserialize /serialize when read/write . then compress the file before send. |
| |
| |
| | | |
|
| | |
| |
| Publicado el 13,enero 2016 - 08:57 |
Hi CCC,
its completly WX technologoy:
at the client WD20 with structures and httprequest and at the serverside WB20 with AWP websites which receives the request and make somthing with the data like:
hexecutequery(myquery)
fiteoarray(myquery,myarray_of_structures)
serialize(mybuffer,myarray_of_structures)
result mybuffer
How could this done with raw files ? How can I write a array in a file and it is really smaller ? |
| |
| |
| | | |
|
| | |
| |
| Publicado el 15,enero 2016 - 20:10 |
Hi Michael,
I think that 17sec v 30sec is actually VERY GOOD !!!!!
I don't think you will get anything better with httprequest.
I am using WEB Services (developed with WD) for this kind of things and I am not getting better results.
Steven Sitas |
| |
| |
| | | |
|
| | |
| |
| Publicado el 15,enero 2016 - 21:59 |
Hi Steven
unfortunately you're right.
I try to use paralletask how much as I can to keep the good user
experience from my old wd14 program. |
| |
| |
| | | |
|
| | |
| |
| Publicado el 16,enero 2016 - 05:26 |
Hi Michael,
you need to let us know the size per record / per batch (20.000 records) .
if the size per batch is 1K and took 17 seconds , that is not very good |
| |
| |
| | | |
|
| | |
| |
| Publicado el 16,enero 2016 - 09:12 |
Hi CCC,
my example should show the ratio between the nativ access and the access over httprequest.
The absolut size is not so relevant.
If there is no fundamental other method to transfer a query result over html then I must live with that.
My application has go through all firewalls without openening any specific port, thats my dilemma.
If webdev has all the controls and possibilities of windev I would use that ... |
| |
| |
| | | |
|
| | |
| |
| Publicado el 16,enero 2016 - 09:30 |
Hi,
I Would check out ZEROMQ.
Somebody has already take the trouble to get it to work with WinDev WebDev .
Cheers
Tor-Bjarne |
| |
| |
| | | |
|
| | |