RB7
@insightBB.com
| Insightbb just plain suks, period.
I just had a buddy in NY go to this address...
www.mtco.com/techinfo
and hit the speed button. He gets 3500 consistently from NY and pays $35 for his internet connection. The speed test is in Central Illinois. I am a couple town away from it, and I get 2400 - 2600 with a dip down in the 1000-1500 range about every 10 refreshes, and I am paying what, $75 for 384/3000
SPARE ME! |
|
indy0365
join:2001-08-25
Franklin, IN | hmm either move or find something else |
|
crischen
join:2000-11-07
Anderson, IN
| reply to RB7
Are you on a wireless connection? What kind of router are you using? How long is your cable length if not? How new are said cables? What kind of modem do you have? Have you done tweaks? All these could be a factor in reduced speeds. You could also need a filter/booster on your line.
--
-Kinsey |
|
adz133
join:2002-12-03 | reply to RB7
I don't think the "tweaks" really don't do anything...
I get the same speeds from linux boxen, my openbsd box, and my windows boxen. |
|
IGGY
No Guru Just Here To Help
Premium,MVM
join:2001-03-30
Chatham, IL | Well Cisco seems to recommend Dr. TCP. They must feel it has some value. |
|
adz133
join:2002-12-03
| reply to RB7
I personally have never held much respect for cisco... their products are just overpriced for what they are.
As for the tweaks, if anybody wants to do a real statistical analysis of their speeds (take speeds test at regular time intervals, over the course of several months, and then take mean and std deviation for both basic and with tweaks, I'd just love to see the results) that would be great. |
|
IGGY
No Guru Just Here To Help
Premium,MVM
join:2001-03-30
Chatham, IL
| You just love to argue. The point was that company runs many of the nets servers. I'll refrain from posting the link. But Cisco has stated they feel certain tweaks in Dr. TCP can help connection quality. They also include the download link on this page were they discuss this.
Many other users of this forum and others. Have seen a good amount of speed increase after doing tweaking. That fact is proven in this forum and others here on a fairly regular basis.
Do you like anyone who isn't open source?
--
Test Your Security Benefit for Children's Oncology Group Cable Modem Diagnostics |
|
adz133
join:2002-12-03
| reply to RB7
What does open source have anything to do with it?
I just have my doubts that it works, and a lot of the "evidence" has a lot of room to be placebo effect.
Just because Cisco think it "may" help, doesn't mean I think it does.
I don't run software until I have a proven use for it. Stuff like Dr. TCP is useless bloat unless there is well documented evidence that it will improve my connection.
Running a program just for the sake of not arguing is kind of like believing in a witch hunt just so you don't lay suspicion unto yourself. |
|
indy0365
join:2001-08-25
Franklin, IN | reply to RB7
so what are your setting from the tweak test on the front page |
|
BeesTea
Network Janitor
Premium,VIP
join:2003-03-08
00000
| reply to RB7
Dr. TCP and other programs have nothing to do with Cisco or any other manufacturer. It's related to fine tuning your network stack to operate on the real world physical network. All of this is well documented, researched, and proven. IGGY's statement that it "fixes" a number of people's connections is correct. Most systems are not "big delay network" friendly out of the box. Adjusting RWIN, MTU etc to match the network can infact greatly increase performance. Cable modem networks are not ethernet LAN's, they involve queueing and media type changes etc, all of this can degrade performance on a stack setup to only operate over local ethernet by default.
There's plenty of research and documentation about it.
Take a look at RFC's 793 and 1323
»www.faqs.org/rfcs/
Pittsburgh Super Computing has done a massive amount of research about it, going so far as to develop a way to make endpoints adjust their stacks according to network conditions.
»www.psc.edu/networking/
Not bad for make believe.
-BeesT
--
echo 16i[q]sa[ln0=aln100%Pln100/snlbx]sbA0D4D465452snlb xq |dc |
|
adz133
join:2002-12-03 | reply to RB7
If someone has a router, and they use these tweaks on the client computers, doesn't the router still screw them over because it is retransmitting the data using its own stack settings? |
|
BeesTea
Network Janitor
Premium,VIP
join:2003-03-08
00000
| In theory I would guess that could be the case. I don't use any SOHO hardware for routing, so I can't say what options are adjustable on these devices.
I've used a Linux or BSD based router for the last 10 years or so. Once I started using cable I found that the Linux and BSD stacks needed some tuning to operate at maximum performance on the BDP network.
I use the following sysctl changes to get an increase of about a half megabit compared to the default stack settings.
echo 8388608 > /proc/sys/net/core/wmem_max
echo 8388608 > /proc/sys/net/core/rmem_max
echo "4096 87380 5194304" > /proc/sys/net/ipv4/tcp_rmem
echo "4096 65536 5194304" > /proc/sys/net/ipv4/tcp_wmem
echo "4096 87380 5194304" > /proc/sys/net/ipv4/tcp_mem
This adjustment in the stack memory use allows more time for queueing etc. I've also found that doing my own queueing via QoS slightly improves performance.
All of these things are caused by dynamic conditions on the network. Other people's traffic, overloaded routers, bad lines etc all play a part. With that in mind, the settings and gains will also vary.
-BeesT
--
echo 16i[q]sa[ln0=aln100%Pln100/snlbx]sbA0D4D465452snlb xq |dc |
|
adz133
join:2002-12-03
| reply to RB7
Well, I'll give that a shot sometime then. I already do QoS on my router (isn't pf wonderful?) but I didn't get a whole lot out of it. It only meant that while I was gaming, somebody browsing the web or ftping something didn't lag me to death.
So what about the machines on the inside of the network? Is there a way to have alternate settings for each interface?
I have ep0, ep1, ep2, and xl0 on my router. xl0 is the internal card, ep1 goes to the cable modem, ep2 goes to the wireless (tunnel based auth), ep0 goes nowhere. So, is it possible to have the stack treat each interface differently? ep1 being configured for the cable network, ep2 and xl0 being left at default...? |
|
BeesTea
Network Janitor
Premium,VIP
join:2003-03-08
00000
| Your internal wired network should operate fine with the standard 1500 MTU and stack settings as there's probably minimal delay. I would assume your 802.11 network would be the same, so there should be no need to adjust the stack on the backend systems. The only place you'll run into the issue of stack timing is over the cable modem, it's a limitation of the medium.
And yes, I also think pf is great.
-BeesT
--
echo 16i[q]sa[ln0=aln100%Pln100/snlbx]sbA0D4D465452snlb xq |dc |
|
adz133
join:2002-12-03 | reply to RB7
...but how do you adjust these values per if? |
|
BeesTea
Network Janitor
Premium,VIP
join:2003-03-08
00000 | The sysctl options affect all of the interfaces.
-BeesT |
|
adz133
join:2002-12-03 | reply to RB7
Well, I knew that.... but I was wondering if there was a way to have each if handled differently. Guess not.
I'll check up on that sometime in the daylight hours. |
|
