what network device, feature, configuration may cause delay

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
05-06-2009 08:33 PM - edited 03-04-2019 04:40 AM
Without knowing what network devices, features, and configuration at the remote network. What may cause too much delay of loading jsp/php scripts, and images to the user web browser at the remote network?
If the same is not experience by users in other remote networks and eliminating web server, user pc/browser, DNS, virus, trojan, proxy, congestion, and latency as the culprit. What could cause too much delay that can range from 7 seconds to 2 minutes where it only takes 2 seconds to 4 seconds in other remote networks to load the whole page.
I use HTTPwatch http://www.httpwatch.com and StopWatch http://www.numion.com to observe the loading at all remote sites, this is what I observed at the remote site in question.
- the TTFB (Time To First Byte) of scripts and images takes a lot of time.
- the TTLB (Time To Last Byte) of scripts and images took a lot of time also.
This doesn't happen in other remote sites, the TTFB there execute immediately once the user browse the site, and the TTLB finish quickly.
This is a general opinion discussion, not show me your config discussion. Reason is that all remote networks (including the remote network in question) is not manage by me.
Any comments/opinion are welcome.
TIA
- Labels:
-
Other Routing
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
05-07-2009 03:24 AM
If performance is really, really bad for a TCP app, such as HTTP, many lost packets is often the cause. Have you checked for TCP retransmits?
L2 issues, duplex being one, can greatly impact performance.

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
05-07-2009 03:46 AM
Hi Joseph,
Thank you for your reply.
Nope, none of those is a problem. The remote network in question does not experience those problem with us. We also eliminate them together with previously excluded service/config/features. We have other application (e.g. web and ftp) that don't experience this with the same connection to them.
Actually, the performance changes dramatically to 2 seconds web page loading speed if we do the following;
1. Ask user to connect using Remote Access VPN. VPN Server is nearer the web server in our network.
or
2. We re-route the return traffic to other remote networks (which have direct connection to them thru different entry point).
From what I know, item 2 bypasses QoS of the remote network in question, because they QoS our connection to them. I did ask whether they use CAR/rate-limit or SHAPING, they replied that they are using SHAPING. I asked for config to confirm and review, so far the first reply is NO.
I remember that I read in this forum long ago that CAR/rate-limit 'may' cause problem to SCRIPTS, the poster even post the config and someone suggested an enhancement. I search high and low but I can't find that discussion anymore.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
05-07-2009 04:23 AM
Doesn't make sense why a CAR, alone, would impact SCRIPTS. What CAR often does is aggressively drop packets. This would slow any app depending on network performance.
A shaper too can very much impact performance since it will mimic a slower interface. A special problem is one that doesn't provide enough shape queuing, which too results in a high rate of discarded packets.
Both a CAR or shaper might only reveal a problem when stressed.
