RE: citrix & fat clients (was War between the states)

[WOsborne]

RE: citrix & fat clients (was War between the states)

I do have a question for you gurus. Why is a fat client slower than a Citrix
client? This seems to be a consistent fact in our case, however I don't
quite understand how it can happen.

The thing I hate about Citrix servers is the fact that package deployment is
a nightmare. Of course, that goes for fat clients also as well. But I
suppose that that is an architecture issue. On the other hand, I really like
the performance in a dial-up mode. That helps in our case with users working
from home.

 

[WalterD]

Citrix - Screen shots are sent to the client (RDP)
Fat Client - Data request are sent to the Enterprise servers for almost all tasks


Fat client performance
Processing time depends on client resources (RAM, HD, bandwidth)
OS considerations

Thin client (Citrix)
Server has alot more resources, bandwidth to Enterprise sever is larger

Walter De Melo
Mosaic Group Inc.

 

[paul_ross]

RE: citrix & fat clients (was War between the states)

I think there are possibly 3 answers to this.
1) Network segmentation. Often time a terminal server is on its own segment
and has a more direct route to the enterprise server. A fat client is
sharing his bandwidth with 12, 24, 48, etc. other users.
2) When you have a dual or quad Xeon processor machine you obviously have a
lot of horsepower. Odds are different users are doing different things at
different times. While user 1 is at lunch, talking on the phone, etc. user 2
is banging away using the processor. You also typically have faster disks in
a terminal server which means faster spec loads.
3) Some people choose to run BSFN's on a separate enterprise server. In some
instances this may speed things up. We do not do this and we still get
acceptable performance.

Paul Ross
OneWorld Hired Gun
[email protected]
Phone: (513) 984-3093 x3097

 

[got_to_love_jde]

I agree with both of you gentlemen and only want to add that specs are only loaded once for each 20-50 users running on a TSE. if they were all fat clients the deployment server can become much busier and network traffic can also increase significantly.
dave


NT 4.0 SP5, SQL 7.0, One World B7321 SP12.4

 

[altquark]

There are 2 reasons - one for WAN connected users, the other for LAN connected users.

First the obvious (LAN) connected users see faster performance because the Terminal Server they are using is usually a high-specification than most workstations. SCSI drives, more cache and dual processors often mean that the application appears quicker when the Terminal server is not heavily loaded. Of course - put large amounts of load on a Terminal Server, and it will certainly be slower than a FAT Workstation eventually.

Secondly, WAN connected users see faster performance - especially over dialup - because Citrix is a Streaming technology versus OneWorld's Client Server Technology. OneWorld is extremely chatty - making SQL statements one after the other - each dependant on the last. If it takes 200ms to traverse a WAN connection for each statement and (as an example) Sales Order Entry requires 600 application "turns" across the network - 600x200=120000ms or about 2 minutes. (600 turns is the average 2-tier Sales Order Line btw and 200ms is a poor internet connection coast-to-coast). If the connection has 10ms latency - then 600x10=6000ms or about 6 seconds. Most decent high-speed ethernet connections should have approximately 3ms or less latency between the Citrix Server and the Enterprise Server - giving less than 2 seconds response time per sales line.

Lots of whitepapers are out on the JDE Knowledge Garden that explains this in more detail - including my WAN and Distributed CNC Architecture document from 1997.

Jon Steel


Jon Steel                                       
Xe Upgrade Specialist - AppzBiz