Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations gkittelson on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

Nagles Algorythm

Status
Not open for further replies.
Jun 15, 2002
31
US
We are investigating packet latency RE: Nagles algorythm.
Running HP-UX 11.31, over 50 remote locations via MPLS and supporting
RF barcode applications in these sites. Latency has been an issue especially
with the barcode guns. Running telnet (port 23) for the guns.
Hoping someone can share info with similar experience, changes made, problems found, pitfalls.
Have over 1100 other users via ssh.
Thanks
GrotonGroup :)
 
Although I am no expert on barcode scanning across WANs, I'll make an assumption that the barcode TELNET based traffic shares your MPLS links with other production traffic including SSH. Although these don't tend to be highly impacting protocols, have you had someone assess the following;

- What applications/user IPs are taking up the most bandwidth
- Are other types of traffic experiencing similar issues
- What priority is given to your Telnet traffic over other traffic
- Are there any WAN latency issues (i.e are some sites getting good response while others have slow responses)
- Locally are the barcode scanners on the same subnet as other devices, user machines, servers etc

This isn't an exhaustive list but most issues with applications over MPLS (and WAN links in general) tend to fall into the following;

- Quality of Service (prioritizing certain types of traffic)
- The protocols that are communicating. As an example SSH will only use what is needed and doesn't lend itself to network prioritization or bandwidth increases. It has to be tuned at an OS level

Hope this helps.

 
Hi,
Thanks for the response.
We have prioritized our QOS on the MPLS and for the most part, that is under control.
The wireless bar code guns appear to be the major issue with latency and one of our senior network IT specialists
discovered the Nagle's alogrythm issue.
It makes sense with how the packets are being sent with/without the _NODELAY, _DELAY options.
I'm attempting to see what ramifications others have experienced either good or bad by modifying the packet ACK/NACK with
small data byte packets.
Thanks,
GrotonGroup

GrotonGroup :)
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top