Transparent cache absolutely dismal

[jem16]

I'm on the Renfrew UBR and the transparent cache is absolutely dismal. Changed over to Edinburgh, on port 8080 and it's flying. In fact, I think anything that doesn't use port 80 and goes through the transparent cache would be flying. It's been like this for the last few days in particular. Any ideas what's going on?



Speed Test Report

Broadband Speed Test - for ntl:World - 1Mb service
By Daniel Elwell - dannyelwell2001@msn.com
Get this test from http://www.vantage.pwp.blueyonder.co.uk

*** THIS TEST IS NOT OWNED OR PAID FOR BY BLUEYONDER ***
If you like this Speed Test, please make a donation to help me
meet the development costs. No ISPs pay me to maintain this test.
More information: http://www.vantage.pwp.blueyonder.co.uk/donate.htm

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
These results are only a snapshot of this particular moment. They are dependant
on any other Internet use on this computer and any delays
on the Internet. To improve accuracy, run the test several times.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Section 1: General Ping Results
A Ping test is the time taken for a server to reply to a request

ntl:World Server:
21ms - Normal - it's about right!

UK Server 1 (www.demon.net):
0ms - *** PING timed out ***
Tracert to specified location:
Tracert was cancelled

UK Server 2 (www.bbc.co.uk):
23ms - Normal - it's about right!

UK Server 3 (www.ic24.net):
33ms - Normal - it's about right!

USA East Coast Server (www.yahoo.com):
0ms - *** PING timed out ***
Tracert to specified location:
Tracert was cancelled

USA West Coast Server (www.lycos.com):
96ms - Fast - very fast response

European Server (www.marcopoly.com):
24ms - Excellent - better than expected

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Section 2: Custom Server Ping Results
These results are to four user-definable servers.

Blueyonder Counter-Strike 1:
0ms - *** PING timed out ***
Tracert to specified location:
Tracert was cancelled

Blueyonder Team-Fortress Leage 1:
0ms - *** PING timed out ***
Tracert to specified location:
Tracert was cancelled

BarrysWorld Master Server (all games):
0ms - *** PING timed out ***
Tracert to specified location:
Tracert was cancelled

Barrysworld Quake III Arena:
22ms -

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Section 3: Download Speed Results
Three downloads are tested; one from one of your ISPs servers, one through your Transparent Cache and one through an alterntive cache. This will give you an idea of how capable your connection is. Note: You should take the highest download speed into account.

ntl:World - 1Mb service Server Download:
114KB/s (912Kb/s) - Average, though much higher should be achieved

Download through Transparent Cache:
14KB/s (112Kb/s) - Dismal - this is extremely poor

Download through alternative Cache (webcache.blueyonder.co.uk):
0KB/s (0Kb/s) - **** SERVER ERROR: Unexpected error from Server****

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Section 4: Packet Loss
50 Ping Requests are sent to your server. The packet loss is measured by how many of the packets don't bounce back to the client. High packet loss may slow down your connection.

50 packets were sent, 50 bounced back; 0% loss (estimated)
Fantastic - no packets dropped!
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Section 5: Transparent Cache Page Fetch Time
This test times how long your Transparent Cache takes to fetch a page from the Internet. Please note that this test is subject to many different parameters and sometimes may not be 100% accurate. This test can be used to show delays when trying to view Internet sites. The timing takes place from the moment the request for the file is sent to the moment the proxy server responds.

Average time taken to retrieve a page: 40ms - Fine, nothing to worry about


Test Statistics:
Information on the test

Time and Date: 01/01/2004 22:34:02
Test duration: 145 secondsIP address:
Username: [Not disclosed]
Test Version: 2.81

Broadband Speed Test - www.vantage.pwp.blueyonder.co.uk

Sadly, transparent caches are the worst thing ever invented and NTL's are particularly prone to slowing down.
The only answer is what you have done (switch to another) and every now and then switch back ans see if they have fixed your local one.

I hope that was the bad report, proxy performance is still a bit iffy, disregard the blueyonder results, they recently blocked such stuff.

Gaz

[jem16]

Originally Posted by pem

Sadly, transparent caches are the worst thing ever invented and NTL's are particularly prone to slowing down.
The only answer is what you have done (switch to another) and every now and then switch back ans see if they have fixed your local one.

My own proxy using port 80 (the transparent cache) can be fine off-peak but during peak times is absolutely terrible.

My own proxy using port 8080 ( hence avoiding the transparent cache ) is fine. I've pasted the relevant section from the speed test below;

Speed Test Report

Broadband Speed Test - for ntl:World - 1Mb service
By Daniel Elwell - dannyelwell2001@msn.com
Get this test from http://www.vantage.pwp.blueyonder.co.uk

*** THIS TEST IS NOT OWNED OR PAID FOR BY BLUEYONDER ***
If you like this Speed Test, please make a donation to help me
meet the development costs. No ISPs pay me to maintain this test.
More information: http://www.vantage.pwp.blueyonder.co.uk/donate.htm



Section 3: Download Speed Results
Three downloads are tested; one from one of your ISPs servers, one through your Transparent Cache and one through an alterntive cache. This will give you an idea of how capable your connection is. Note: You should take the highest download speed into account.

ntl:World - 1Mb service Server Download:
117KB/s (936Kb/s) - Perfect - running at full speed!

Download through Transparent Cache:
37KB/s (296Kb/s) - Poor - this is too low

Download through alternative Cache (inktomi1-ren.server.ntl.com):
105KB/s (840Kb/s) - Average, though much higher should be achieved

* It appears that the alternative cache is working better than the ntl:World - 1Mb service cache. To improve Web Browsing, set the Proxy Server in your Web Browser to inktomi1-ren.server.ntl.com on port 8080*


Broadband Speed Test - www.vantage.pwp.blueyonder.co.uk


Surely NTL can do better than this with the transparent cache or take them offline?

[ntl customer]

Do Telewest still force transparent caching on their broadband subscribers?

I've heard rumours that they have removed it - is this true??!

[jem16]

Originally Posted by homealone

I hope that was the bad report, proxy performance is still a bit iffy, disregard the blueyonder results, they recently blocked such stuff.

Gaz

It was the bad report.

I've now changed the blueyonder cache to my own local cache but on port 8080. What a difference!

Originally Posted by jem16

It was the bad report.

I've now changed the blueyonder cache to my own local cache but on port 8080. What a difference!

I'm glad about that - the report in post #4 looks a lot better.

It's pants you should have to mess about like that to get your connection working as it should, but at least you have the knowledge - I just feel sorry for the people who havn't, or aren't members of forums like this so they can aquire it.

Gaz

[jem16]

Originally Posted by homealone

I'm glad about that - the report in post #4 looks a lot better.

It's pants you should have to mess about like that to get your connection working as it should, but at least you have the knowledge - I just feel sorry for the people who havn't, or aren't members of forums like this so they can aquire it.

Gaz

I agree. I'm paying for a 1MB service and getting 48kb/s. At 8.26pm this is what I'm getting;


Speed Test Report

Broadband Speed Test - for ntl:World - 1Mb service
By Daniel Elwell - dannyelwell2001@msn.com
Get this test from http://www.vantage.pwp.blueyonder.co.uk

*** THIS TEST IS NOT OWNED OR PAID FOR BY BLUEYONDER ***
If you like this Speed Test, please make a donation to help me
meet the development costs. No ISPs pay me to maintain this test.
More information: http://www.vantage.pwp.blueyonder.co.uk/donate.htm

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
These results are only a snapshot of this particular moment. They are dependant
on any other Internet use on this computer and any delays
on the Internet. To improve accuracy, run the test several times.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Section 1: General Ping Results
A Ping test is the time taken for a server to reply to a request

ntl:World Server:
21ms - Normal - it's about right!

UK Server 1 (www.demon.net):
0ms - *** PING timed out ***
Tracert to specified location:
Tracert was cancelled

UK Server 2 (www.bbc.co.uk):
24ms - Normal - it's about right!

UK Server 3 (www.ic24.net):
24ms - Normal - it's about right!

USA East Coast Server (www.yahoo.com):
0ms - *** PING timed out ***
Tracert to specified location:
Tracert was cancelled

USA West Coast Server (www.lycos.com):
101ms - Fast - very fast response

European Server (www.marcopoly.com):
24ms - Excellent - better than expected

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Section 2: Custom Server Ping Results
These results are to four user-definable servers.

Blueyonder Counter-Strike 1:
0ms - *** PING timed out ***
Tracert to specified location:

Tracing route to 194.159.254.213 over a maximum of 20 hops
1 9 ms 8 ms 8 ms 10.147.23.254
2 7 ms 7 ms 7 ms 80.4.64.65
3 7 ms 8 ms 7 ms 62.254.191.65
4 8 ms 7 ms 10 ms 62.253.184.81
5 14 ms 14 ms 13 ms 62.253.185.162
6 14 ms 15 ms 20 ms 62.253.187.185
7 19 ms 19 ms 20 ms 62.253.185.238
8 66 ms 61 ms 63 ms 62.253.185.82
9 * * * Request timed out.
10 * * * Request timed out.
11 * * * Request timed out.
12 * * * Request timed out.
13 * * * Request timed out.
14 * * * Request timed out.
15 * * * Request timed out.
16 * * * Request timed out.
17 * * * Request timed out.
18 * * * Request timed out.
19 * * * Request timed out.
20 * * * Request timed out.
Trace complete.


Blueyonder Team-Fortress Leage 1:
0ms - *** PING timed out ***
Tracert to specified location:
Tracert was cancelled

BarrysWorld Master Server (all games):
0ms - *** PING timed out ***
Tracert to specified location:
Tracert was cancelled

Barrysworld Quake III Arena:
21ms -

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Section 3: Download Speed Results
Three downloads are tested; one from one of your ISPs servers, one through your Transparent Cache and one through an alterntive cache. This will give you an idea of how capable your connection is. Note: You should take the highest download speed into account.

ntl:World - 1Mb service Server Download:
114KB/s (912Kb/s) - Average, though much higher should be achieved

Download through Transparent Cache:
6KB/s (48Kb/s) - Dismal - this is extremely poor

Download through alternative Cache (cache1-glfd.server.ntli.net):
119KB/s (952Kb/s) - Perfect - running at full speed!

* It appears that the alternative cache is working better than the ntl:World - 1Mb service cache. To improve Web Browsing, set the Proxy Server in your Web Browser to cache1-glfd.server.ntli.net on port 8080*

Test Statistics:
Information on the test

Time and Date: 02/01/2004 20:26:39
Test duration: 297 secondsIP address:
Username: [Not disclosed]
Test Version: 2.81

Broadband Speed Test - www.vantage.pwp.blueyonder.co.uk

As you can see I've changed to alternative cache to be Guildford. My own local cache at Renfrew on port 8080 was equally as bad.

I'm sorry NTL but this is just not good enough.

Originally Posted by jem16

My own proxy using port 80 (the transparent cache) can be fine off-peak but during peak times is absolutely terrible.

My own proxy using port 8080 ( hence avoiding the transparent cache ) is fine. I've pasted the relevant section from the speed test below;

You are not avoiding your cache by switching to port 8080, you are simply accessing it properly as a proxy server rather than relying on the port 80 interception. For some reason this often works better even on the same cache.

[jem16]

Originally Posted by pem

You are not avoiding your cache by switching to port 8080, you are simply accessing it properly as a proxy server rather than relying on the port 80 interception. For some reason this often works better even on the same cache.

Ah I see.

However even the Renfrew inktomi on another port has given up now. I've switched to Guildford.


Speed Test Report

Broadband Speed Test - for ntl:World - 1Mb service
By Daniel Elwell - dannyelwell2001@msn.com
Get this test from http://www.vantage.pwp.blueyonder.co.uk

*** THIS TEST IS NOT OWNED OR PAID FOR BY BLUEYONDER ***
If you like this Speed Test, please make a donation to help me
meet the development costs. No ISPs pay me to maintain this test.
More information: http://www.vantage.pwp.blueyonder.co.uk/donate.htm

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
These results are only a snapshot of this particular moment. They are dependant
on any other Internet use on this computer and any delays
on the Internet. To improve accuracy, run the test several times.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Section 3: Download Speed Results
Three downloads are tested; one from one of your ISPs servers, one through your Transparent Cache and one through an alterntive cache. This will give you an idea of how capable your connection is. Note: You should take the highest download speed into account.

ntl:World - 1Mb service Server Download:
118KB/s (944Kb/s) - Perfect - running at full speed!

Download through Transparent Cache:
5KB/s (40Kb/s) - Dismal - this is extremely poor

Download through alternative Cache (inktomi1-ren.server.ntl.com):
0KB/s (0Kb/s) - **** SERVER ERROR: The attempt to connect timed out****

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Test Statistics:
Information on the test

Time and Date: 02/01/2004 21:41:25
Test duration: 291 secondsIP address:
Username: [Not disclosed]
Test Version: 2.81

Broadband Speed Test - www.vantage.pwp.blueyonder.co.uk

Somehow I think Renfrew has a problem!!

[rdhw]

Originally Posted by pem

You are not avoiding your cache by switching to port 8080, you are simply accessing it properly as a proxy server rather than relying on the port 80 interception. For some reason this often works better even on the same cache.

Well, partly. The other factor is that, when you set an explicit proxy, you use only that proxy, for all your traffic. When you do not configure an explicit proxy, your traffic gets allocated to one of a local cluster of proxies, chosen on the basis of the requested HTTP address. If one proxy of the cluster is sick, you notice poor performance on all the addresses that it handles. When you configure an explicit proxy, you probably choose a different member of the cluster (one that is working normally). This is why it often appears that explicit use of a local proxy works when the default transparent does not. If you were to check out every member of the local cluster, you would probably find which one(s) were sick and which were fine.

[jem16]

Originally Posted by rdhw

Well, partly. The other factor is that, when you set an explicit proxy, you use only that proxy, for all your traffic. When you do not configure an explicit proxy, your traffic gets allocated to one of a local cluster of proxies, chosen on the basis of the requested HTTP address. If one proxy of the cluster is sick, you notice poor performance on all the addresses that it handles. When you configure an explicit proxy, you probably choose a different member of the cluster (one that is working normally). This is why it often appears that explicit use of a local proxy works when the default transparent does not. If you were to check out every member of the local cluster, you would probably find which one(s) were sick and which were fine.

So far inktomi 1 and 3 are sick, almost fatally I think - both about 40kb/s
Inktomi 2 is wonderful - about 1092kb/s

4 and 5 will have to wait as I'm too tired to wait another half hour or so!

Originally Posted by rdhw

Well, partly. The other factor is that, when you set an explicit proxy, you use only that proxy, for all your traffic. When you do not configure an explicit proxy, your traffic gets allocated to one of a local cluster of proxies, chosen on the basis of the requested HTTP address. If one proxy of the cluster is sick, you notice poor performance on all the addresses that it handles. When you configure an explicit proxy, you probably choose a different member of the cluster (one that is working normally). This is why it often appears that explicit use of a local proxy works when the default transparent does not. If you were to check out every member of the local cluster, you would probably find which one(s) were sick and which were fine.

Interesting, I didn't know it was address based (or IP based ?) allocation. What device actually makes the decision and routes you to the appropiate machine ?

[MadGamer]

Try http://www.adslguide.org.uk/tools/speedtest.asp and post back the results.

[Chrysalis]

does using a diff dns server disable the proxy?

I use a diff dns server as NTL's is unusable and I am actually not going through any transperent proxies, am I just in a good area? or is the dns bypassing it?