When downloading bodies from one of the larger binary groups (alt.binaries.sounds.lossless), I often find that one of my 3 usenet providers does not contribute to the download while the other 2 providers are see sawing back and forth in terms of download speed.
I have a relatively fast connection so that I can get up to 1.7 MB/sec download speeds from my usenet connection.
However, as I mentioned, one of the providers seems locked out...intermittently at least.
Yesterday, I went so far as to email Tech Support for one of the providers to find out if they were experiencing problems with alt.binaries.sounds.lossless.
Of course they were not.
I believe that either I have a problem with settings or my database is corrupted. I had not noticed this problem until recently, and I had not made any setting changes.
There are over 9 million articles in this newsgroup.
I am runing UE 2.1.4 on Windows XP with a AMD x 2 4400+, 2 GB RAM.
Thanks.
Varying Download Speeds between Various Usenet Servers
no, it must be usenet provider or route to the usenet provider (maybe also to try to reset your router).
for UE all the difference here is a couple of numbers which tell it what header range to get, then whatever server is returning all is thrown into the database exactly in the same way, there is no any difference as to database whatever server you download from.
also database cannot get corrupted, since it verifies/fixes itself every time you load a newsgroup. for compact binary newsgroup destination type (which you must be using for larger newsgroups to prevent RAM depletion) database as to newsgroup doesn't even know from what server what headers came from.
the best you can do as to maintaining database is to check the file system for errors, but it is unrelated to downloading headers.
for UE all the difference here is a couple of numbers which tell it what header range to get, then whatever server is returning all is thrown into the database exactly in the same way, there is no any difference as to database whatever server you download from.
also database cannot get corrupted, since it verifies/fixes itself every time you load a newsgroup. for compact binary newsgroup destination type (which you must be using for larger newsgroups to prevent RAM depletion) database as to newsgroup doesn't even know from what server what headers came from.
the best you can do as to maintaining database is to check the file system for errors, but it is unrelated to downloading headers.
experiment
I decreased the number of tasks for the other two usenet servers to zero each.
When I did that, the usenet server in question then began downloading articles at high speed.
This suggests to me that I have some settings wrong in UE.
Thanks for your reply, Alex.
When I did that, the usenet server in question then began downloading articles at high speed.
This suggests to me that I have some settings wrong in UE.
Thanks for your reply, Alex.
further experiments
I have varied the number of "tasks" of my three servers.
I am assuming that tasks=connections (at least when you are only downloading bodies).
When I increase the tasks for the two OTHER newsgroup servers (usually above 5-10 each), the server in question (although it allows 20 connections) progressively decreases its throughput to near zero.
Is the throughput of the usenet provider being squeezed out by some factor of the other two newsgroup servers? I realize that my "pipe" is only so big, but why do some servers dominate?
Once again, when I effectively turn off the other two providers, I get very high speed throughput from the server I am wondering about.
Someone educate me here.
Thanks.
PS I changed the title of the thread to more accurately reflect the behavior that I am seeing.
I am assuming that tasks=connections (at least when you are only downloading bodies).
When I increase the tasks for the two OTHER newsgroup servers (usually above 5-10 each), the server in question (although it allows 20 connections) progressively decreases its throughput to near zero.
Is the throughput of the usenet provider being squeezed out by some factor of the other two newsgroup servers? I realize that my "pipe" is only so big, but why do some servers dominate?
Once again, when I effectively turn off the other two providers, I get very high speed throughput from the server I am wondering about.
Someone educate me here.
Thanks.
PS I changed the title of the thread to more accurately reflect the behavior that I am seeing.
but overall you manage to saturate the bandwidth?
did you try to reset the router if you have one?
UE just reads from sockets, from its point of view all connections are equal. what exactly comes in is determined by lower levels, so any disbalance may only come from there.
on your computer it might be firewall, router, then maybe something depending on external routing all the way to the server including the server router, when router buffers are full (i.e. more data is coming when your connection speed allows) some TCP protocol adjustment takes place, so the number of packets which is sent from the server is decreased, maybe different router policies along different routes, if it looks like abnormal, it is not your local problem and it wasn't the case maybe it will fix itself later.
did you try to reset the router if you have one?
UE just reads from sockets, from its point of view all connections are equal. what exactly comes in is determined by lower levels, so any disbalance may only come from there.
on your computer it might be firewall, router, then maybe something depending on external routing all the way to the server including the server router, when router buffers are full (i.e. more data is coming when your connection speed allows) some TCP protocol adjustment takes place, so the number of packets which is sent from the server is decreased, maybe different router policies along different routes, if it looks like abnormal, it is not your local problem and it wasn't the case maybe it will fix itself later.
router
I reset the router yesterday.
Yes, I do manage to saturate my connection, regardless of the number of connections per server.
The lesson to me is to unsubscribe from one or more of my servers.
Thanks again, Alex.
Yes, I do manage to saturate my connection, regardless of the number of connections per server.
The lesson to me is to unsubscribe from one or more of my servers.
Thanks again, Alex.
two other paths of thought here.
groups like Boneless are huge compared to the others. non premium servers always have problems accessing and keeping decent retention and completeness with these groups.
It's been my experience that some home routers have major problems
when trying to encountering multiple 'file not found' in a row.
We're talking like 4,5,7 hundred here all at once. this is a problem independent of any news client.
It seems the news software sends out the requests and the responses come back so fast the router losees track of which port it's working with. after time the router runs out ports. symptoms: at first the router will slow down then eventually it will lock up.
I'm sure this isn't your problem but it could be if you try to get to many non existence file at once. P2P will also cause this router bug to appear. a simple power down will reset the router.
my semi-solution was to set the F5>server>connect delay to 1 or 2 when i see this happening.
groups like Boneless are huge compared to the others. non premium servers always have problems accessing and keeping decent retention and completeness with these groups.
It's been my experience that some home routers have major problems
when trying to encountering multiple 'file not found' in a row.
We're talking like 4,5,7 hundred here all at once. this is a problem independent of any news client.
It seems the news software sends out the requests and the responses come back so fast the router losees track of which port it's working with. after time the router runs out ports. symptoms: at first the router will slow down then eventually it will lock up.
I'm sure this isn't your problem but it could be if you try to get to many non existence file at once. P2P will also cause this router bug to appear. a simple power down will reset the router.
my semi-solution was to set the F5>server>connect delay to 1 or 2 when i see this happening.
this definately looks like a server issue here. If you think about it it can't be a UE issue, it doesn't have any setting to enable you to vary the bandwith allocated to each server. i bet if you traceroute ping each server you will see that there is a discernable difference between each server. i will also bet that the number of nodes the 'slow' server goes through is significantly more than the others. of course you could ,be living an equal distance from all three farms but i doubt it.........