reducing TCP overhead
reducing TCP overhead
as many of us know and experience every day there is a thing called TCP overhead.
Its basically a portion of your bandwidth thats reserved/used for all those packets that are needed for tCP to do its thing.
but as bandwidth increases so does the overhead.
right now here is where im at.
Mediacom 5000/256 (yes they are so scared of us having more than 256 that they wont give it to us.)
here are my speeds as of the last few weeks.
2005-06-14 01:45:01 Line quality 0% loss latency 42.6ms View..
2005-06-12 11:57:52 Speed test @ chi.speakeasy.net 4616/235 kbps
2005-06-07 08:40:31 Line quality ping tests cancelled View..
2005-06-07 08:34:22 Speed test @ msp.speakeasy.net 4571/234 kbps
2005-06-07 08:32:09 Speed test @ nyc.speakeasy.net 4483/234 kbps
2005-06-07 08:31:11 Speed test @ chi.speakeasy.net 4600/235 kbps
2005-06-06 19:55:51 Speed test @ stl.speakeasy.net 4173/232 kbps
2005-06-06 19:55:04 Speed test @ msp.speakeasy.net 4440/191 kbps
2005-06-06 19:49:38 Speed test @ ns1.etheric.net 1797/233 kbps
2005-06-06 19:48:32 Speed test @ nyc.speakeasy.net 863/235 kbps
2005-06-06 19:15:42 Speed test @ chi.speakeasy.net 3334/220 kbps
2005-06-03 11:33:33 Speed test @ det.speakeasy.net 4556/234 kbps
2005-06-03 11:33:01 Speed test @ nyc.speakeasy.net 4383/230 kbps
2005-06-03 11:12:27 Speed test @ chi.speakeasy.net 4570/226 kbps
2005-05-28 23:20:22 Speed test @ atl.speakeasy.net 4616/217 kbps
2005-05-28 19:21:17 Speed test @ chi.speakeasy.net 4677/234 kbps
2005-05-28 19:19:06 Speed test @ mia.speakeasy.net 4630/232 kbps
2005-05-28 19:18:19 Speed test @ mia.speakeasy.net 4631/235 kbps
2005-05-27 13:19:26 Speed test @ stl.speakeasy.net 4571/217 kbps
2005-05-27 13:07:55 Speed test @ atl.speakeasy.net 4617/182 kbps
2005-05-27 13:05:43 Speed test @ phx.speakeasy.net 3565/209 kbps
2005-05-27 12:55:54 Speed test @ chi.speakeasy.net 4662/218 kbps
if you go thru mediacom's list over at dslreports.com you will see that pretty much everyone is getting 4600 and below.
this is caused by TCP Overhead and the fact that mediacom wont implement Cap's that are slightly higher than 5000 to make up for the overhead.
so,,, given that,, are there any settings to adjust his and utilize more of the bandwidth thats there for regular usage.
i highly doung that 400k of bandwidth (50kb/sec) is used for that. although 50 isnt much in the scheme of 5000 down (or around 575kb.sec which is the max that i get) trimming that 50kb.sec down to a smaller number would be nice.
steele or anyone else know if this is possible ?
Its basically a portion of your bandwidth thats reserved/used for all those packets that are needed for tCP to do its thing.
but as bandwidth increases so does the overhead.
right now here is where im at.
Mediacom 5000/256 (yes they are so scared of us having more than 256 that they wont give it to us.)
here are my speeds as of the last few weeks.
2005-06-14 01:45:01 Line quality 0% loss latency 42.6ms View..
2005-06-12 11:57:52 Speed test @ chi.speakeasy.net 4616/235 kbps
2005-06-07 08:40:31 Line quality ping tests cancelled View..
2005-06-07 08:34:22 Speed test @ msp.speakeasy.net 4571/234 kbps
2005-06-07 08:32:09 Speed test @ nyc.speakeasy.net 4483/234 kbps
2005-06-07 08:31:11 Speed test @ chi.speakeasy.net 4600/235 kbps
2005-06-06 19:55:51 Speed test @ stl.speakeasy.net 4173/232 kbps
2005-06-06 19:55:04 Speed test @ msp.speakeasy.net 4440/191 kbps
2005-06-06 19:49:38 Speed test @ ns1.etheric.net 1797/233 kbps
2005-06-06 19:48:32 Speed test @ nyc.speakeasy.net 863/235 kbps
2005-06-06 19:15:42 Speed test @ chi.speakeasy.net 3334/220 kbps
2005-06-03 11:33:33 Speed test @ det.speakeasy.net 4556/234 kbps
2005-06-03 11:33:01 Speed test @ nyc.speakeasy.net 4383/230 kbps
2005-06-03 11:12:27 Speed test @ chi.speakeasy.net 4570/226 kbps
2005-05-28 23:20:22 Speed test @ atl.speakeasy.net 4616/217 kbps
2005-05-28 19:21:17 Speed test @ chi.speakeasy.net 4677/234 kbps
2005-05-28 19:19:06 Speed test @ mia.speakeasy.net 4630/232 kbps
2005-05-28 19:18:19 Speed test @ mia.speakeasy.net 4631/235 kbps
2005-05-27 13:19:26 Speed test @ stl.speakeasy.net 4571/217 kbps
2005-05-27 13:07:55 Speed test @ atl.speakeasy.net 4617/182 kbps
2005-05-27 13:05:43 Speed test @ phx.speakeasy.net 3565/209 kbps
2005-05-27 12:55:54 Speed test @ chi.speakeasy.net 4662/218 kbps
if you go thru mediacom's list over at dslreports.com you will see that pretty much everyone is getting 4600 and below.
this is caused by TCP Overhead and the fact that mediacom wont implement Cap's that are slightly higher than 5000 to make up for the overhead.
so,,, given that,, are there any settings to adjust his and utilize more of the bandwidth thats there for regular usage.
i highly doung that 400k of bandwidth (50kb/sec) is used for that. although 50 isnt much in the scheme of 5000 down (or around 575kb.sec which is the max that i get) trimming that 50kb.sec down to a smaller number would be nice.
steele or anyone else know if this is possible ?
Xeon 3210 Quad @ 3.6 - Gigabyte P35 DS3R
2gb Skill HZ - 8800 GTX - 3x 36g Raptors in raid0
3x 250gb Drives and 4x 250gb USB Drives
CoolerMaster 750w psu - 2407 FPW & 2007 FPW LCD's
----------------------------------------------------
E6600 ES @ 3.6 @ 1.36v - Asus P5N-E SLI
2GB Gskill HZ's 1200mhz 5-5-5-15 @ 2.27v
250gb Sata II - 7800 GTX - Dell 2007 FPW
Mediacom Max Cable 15meg/1meg - Tweaks not needed
2gb Skill HZ - 8800 GTX - 3x 36g Raptors in raid0
3x 250gb Drives and 4x 250gb USB Drives
CoolerMaster 750w psu - 2407 FPW & 2007 FPW LCD's
----------------------------------------------------
E6600 ES @ 3.6 @ 1.36v - Asus P5N-E SLI
2GB Gskill HZ's 1200mhz 5-5-5-15 @ 2.27v
250gb Sata II - 7800 GTX - Dell 2007 FPW
Mediacom Max Cable 15meg/1meg - Tweaks not needed
-
- Member
- Posts: 97
- Joined: Sat Nov 06, 2004 10:38 pm
Good example:
Tcp/Ip Accelerator 2.5
-----------------------
Description: Tcp/Ip Accelerator Utilizes The Latest Protocol Compression and ISP Algorythm Technology To Speed Up Your Internet By Up To TEN TIMES AS FAST!!! And Not Only Is It Very Easy To Use But It's TOTALLY FREE! So Download it Today.
------------------------------------------------------------------------
LMAO!
----------------------------------------------------------
Go to google and look up "Tcp/Ip Accelerator 2.5" to find it,..................
(I don't recommend buying compression software, because it's an "Illusion of speed".)Tcp/Ip Accelerator 2.5 can't be downloaded; host stopped hosting program.)
----------------------------------------------------------------
The Tcp/Ip Accelerator 2.5 program is compression, it will slow your connection tremedously, b/c your internet works a lot faster than the compression sever or program.
----------------------------------------------------------------
Compression only works for slower connections, like 56K or anything else slow.... Broadband connections, like DSL or Cable don't work while with compression. (Compressing data does not speed up downloading.)
Tcp/Ip Accelerator 2.5
-----------------------
Description: Tcp/Ip Accelerator Utilizes The Latest Protocol Compression and ISP Algorythm Technology To Speed Up Your Internet By Up To TEN TIMES AS FAST!!! And Not Only Is It Very Easy To Use But It's TOTALLY FREE! So Download it Today.
------------------------------------------------------------------------
LMAO!
----------------------------------------------------------
Go to google and look up "Tcp/Ip Accelerator 2.5" to find it,..................
(I don't recommend buying compression software, because it's an "Illusion of speed".)Tcp/Ip Accelerator 2.5 can't be downloaded; host stopped hosting program.)
----------------------------------------------------------------
The Tcp/Ip Accelerator 2.5 program is compression, it will slow your connection tremedously, b/c your internet works a lot faster than the compression sever or program.
----------------------------------------------------------------
Compression only works for slower connections, like 56K or anything else slow.... Broadband connections, like DSL or Cable don't work while with compression. (Compressing data does not speed up downloading.)
no i would never use those goofy program cuz they dont work.
I've never had luck with cable nut either, its never done anything but mess up my connection. I try to tell people that with Win2k and WinXP right out of the box XP is set up for the best speeds your gunna get off cable.
Gotta remember, that when your hooked to cable, your pc doesnt think its a modem or whatever it see's it as a 10/100 Lan connection and therefor should be using LAN settings, expecially if your on a router.
TCP Optimizer has never done anything either. but..
I did find out that WinXP x64 is using an MTU of 1496 and not 1500
where as Windows XP Pro is using an MTU of 1500.
Also that XP x64 uses a TCP Recieve window of 128480 where as XP PRO usus 256960 according to TCP Optimizer.
odd indeed that it would use those different settings for x64 and xp pro.
Since x64 is simply being used for testing i went ahead and applied TCP Optimizers settings and didnt see any speed changes. XP Pro is my main os and im not a fan of jacking with things lol not for this os at least so i havent applied them yet.
I've never had luck with cable nut either, its never done anything but mess up my connection. I try to tell people that with Win2k and WinXP right out of the box XP is set up for the best speeds your gunna get off cable.
Gotta remember, that when your hooked to cable, your pc doesnt think its a modem or whatever it see's it as a 10/100 Lan connection and therefor should be using LAN settings, expecially if your on a router.
TCP Optimizer has never done anything either. but..
I did find out that WinXP x64 is using an MTU of 1496 and not 1500
where as Windows XP Pro is using an MTU of 1500.
Also that XP x64 uses a TCP Recieve window of 128480 where as XP PRO usus 256960 according to TCP Optimizer.
odd indeed that it would use those different settings for x64 and xp pro.
Since x64 is simply being used for testing i went ahead and applied TCP Optimizers settings and didnt see any speed changes. XP Pro is my main os and im not a fan of jacking with things lol not for this os at least so i havent applied them yet.
Xeon 3210 Quad @ 3.6 - Gigabyte P35 DS3R
2gb Skill HZ - 8800 GTX - 3x 36g Raptors in raid0
3x 250gb Drives and 4x 250gb USB Drives
CoolerMaster 750w psu - 2407 FPW & 2007 FPW LCD's
----------------------------------------------------
E6600 ES @ 3.6 @ 1.36v - Asus P5N-E SLI
2GB Gskill HZ's 1200mhz 5-5-5-15 @ 2.27v
250gb Sata II - 7800 GTX - Dell 2007 FPW
Mediacom Max Cable 15meg/1meg - Tweaks not needed
2gb Skill HZ - 8800 GTX - 3x 36g Raptors in raid0
3x 250gb Drives and 4x 250gb USB Drives
CoolerMaster 750w psu - 2407 FPW & 2007 FPW LCD's
----------------------------------------------------
E6600 ES @ 3.6 @ 1.36v - Asus P5N-E SLI
2GB Gskill HZ's 1200mhz 5-5-5-15 @ 2.27v
250gb Sata II - 7800 GTX - Dell 2007 FPW
Mediacom Max Cable 15meg/1meg - Tweaks not needed
-
- Member
- Posts: 97
- Joined: Sat Nov 06, 2004 10:38 pm
Well, I think the settings make a huge difference. (My oppinion)
My connection runs over 3Mbps (DL). (I'm only suppose to get 3Mb/256Kb)
I tweaked my cable connection and cut all the old wires out, my internet seemed to respond better. After I did all that, I went out and bought a Motorola signal booster, just to make sure my wires worked great.
My Internet has never ran better after the regedit tweaks and wire tweaking.
--------------------------------------------------------------------------
I'm sure you could find some utility for reduced browsing overheads, but it may cost money.
My connection runs over 3Mbps (DL). (I'm only suppose to get 3Mb/256Kb)

I tweaked my cable connection and cut all the old wires out, my internet seemed to respond better. After I did all that, I went out and bought a Motorola signal booster, just to make sure my wires worked great.
My Internet has never ran better after the regedit tweaks and wire tweaking.

--------------------------------------------------------------------------
I'm sure you could find some utility for reduced browsing overheads, but it may cost money.

no there isnt anything that i have ever heard of to reduce tcp overhead. its simply the part so tcp that is used for Ack and such like when your mtu is 1500 there is 28 of that that is reserved for tcp.
i simply cant imagine that all 400k is used for it. but its highly possible.
if you go to dslreports and look up their speed results for Mediacom you will see a solid trend of people getting almost identicle speed results.
that's frankly is impossible. but the speed tests show it. a limitation or a qwerk of the java tests,,, ? possibly. but why sooooo manypeople get identicle results is beyond me.
your connection runs over 3meg simply cuz your not on a loaded node, and that signal booster and your wire "tweaking" (thats funny cuz using good wires isnt tweaking) is what makes your connection so good.
remove the registry tweaks and i bet i get almost the same results.
anyways this isnt a descussion of speed tweaks its about reducing a 400k tcp overhead,, if it cant be done then so be it.
also depending on who your cable company is Emerson,, chances are they are using a Cap file that has the caps set above 3000, like 3250 or something. Comcast does it i believe and a few others.
its to keep people from bitching about slower speeds.
mediacom doesnt do this and never will.
i simply cant imagine that all 400k is used for it. but its highly possible.
if you go to dslreports and look up their speed results for Mediacom you will see a solid trend of people getting almost identicle speed results.
that's frankly is impossible. but the speed tests show it. a limitation or a qwerk of the java tests,,, ? possibly. but why sooooo manypeople get identicle results is beyond me.
your connection runs over 3meg simply cuz your not on a loaded node, and that signal booster and your wire "tweaking" (thats funny cuz using good wires isnt tweaking) is what makes your connection so good.
remove the registry tweaks and i bet i get almost the same results.
anyways this isnt a descussion of speed tweaks its about reducing a 400k tcp overhead,, if it cant be done then so be it.
also depending on who your cable company is Emerson,, chances are they are using a Cap file that has the caps set above 3000, like 3250 or something. Comcast does it i believe and a few others.
its to keep people from bitching about slower speeds.
mediacom doesnt do this and never will.
Xeon 3210 Quad @ 3.6 - Gigabyte P35 DS3R
2gb Skill HZ - 8800 GTX - 3x 36g Raptors in raid0
3x 250gb Drives and 4x 250gb USB Drives
CoolerMaster 750w psu - 2407 FPW & 2007 FPW LCD's
----------------------------------------------------
E6600 ES @ 3.6 @ 1.36v - Asus P5N-E SLI
2GB Gskill HZ's 1200mhz 5-5-5-15 @ 2.27v
250gb Sata II - 7800 GTX - Dell 2007 FPW
Mediacom Max Cable 15meg/1meg - Tweaks not needed
2gb Skill HZ - 8800 GTX - 3x 36g Raptors in raid0
3x 250gb Drives and 4x 250gb USB Drives
CoolerMaster 750w psu - 2407 FPW & 2007 FPW LCD's
----------------------------------------------------
E6600 ES @ 3.6 @ 1.36v - Asus P5N-E SLI
2GB Gskill HZ's 1200mhz 5-5-5-15 @ 2.27v
250gb Sata II - 7800 GTX - Dell 2007 FPW
Mediacom Max Cable 15meg/1meg - Tweaks not needed
-
- Member
- Posts: 97
- Joined: Sat Nov 06, 2004 10:38 pm
Well, I know why my speeds are so high.
(tweak) My Oppinion, unless someone does some testing.....
There is a program called SpeedOptimizer. Which uses a Powerful DNS Cache utility for reduced browsing overheads.
Here's the link, it costs money*** :*(
http://speedoptimizer.com/
Go to the "FAQS" tab, tells you everything about it.
Hope that helps, tell me what you think.
--------------------------------------------------------------------------
What is DNS?
Domain Name System. The classification system by which domain names are identified, located and ultimately translated into website or web page addresses meeting Internet protocol standards allowing information to be sent between computers over the Internet. A domain name is usually an easily remembered or meaningful “nickname” for a combination of network addresses that uniquely identifies a given website.
--------------------------------------------------------------------------
How does DNS Cache work?
The DNS Cache functions as an archive of the translations of local domain names into IP addresses. The IP address is a usually a 32-bit number that identifies each sender or receiver of information that is sent in packets across the Internet. All requests for DNS translations go through the DNS Cache in order to reduce browsing overheads.

There is a program called SpeedOptimizer. Which uses a Powerful DNS Cache utility for reduced browsing overheads.
Here's the link, it costs money*** :*(
http://speedoptimizer.com/
Go to the "FAQS" tab, tells you everything about it.
Hope that helps, tell me what you think.

--------------------------------------------------------------------------
What is DNS?
Domain Name System. The classification system by which domain names are identified, located and ultimately translated into website or web page addresses meeting Internet protocol standards allowing information to be sent between computers over the Internet. A domain name is usually an easily remembered or meaningful “nickname” for a combination of network addresses that uniquely identifies a given website.
--------------------------------------------------------------------------
How does DNS Cache work?
The DNS Cache functions as an archive of the translations of local domain names into IP addresses. The IP address is a usually a 32-bit number that identifies each sender or receiver of information that is sent in packets across the Internet. All requests for DNS translations go through the DNS Cache in order to reduce browsing overheads.
wee96 wrote:TCP overhead isnt something thats tweakable, its just how the protocol was written.
thank wee
i pretty much figured that but heck its worth a shot to ask.
the kernel needs those segments for packets so.. no tweaking there.
im getting a pretty much constant 4650/230 except for those days you see in the list and everyone was having issues after the upgrade so im quite happy with 575kb/sec download.
just wish they would get off their asses and offer 512up.
btw anyone who has mediacom now they ARE coming out with 10meg/1meg and supposedly although its not set in stone is going to be a very low priced 70 bucks. or somewhere in that region.
Xeon 3210 Quad @ 3.6 - Gigabyte P35 DS3R
2gb Skill HZ - 8800 GTX - 3x 36g Raptors in raid0
3x 250gb Drives and 4x 250gb USB Drives
CoolerMaster 750w psu - 2407 FPW & 2007 FPW LCD's
----------------------------------------------------
E6600 ES @ 3.6 @ 1.36v - Asus P5N-E SLI
2GB Gskill HZ's 1200mhz 5-5-5-15 @ 2.27v
250gb Sata II - 7800 GTX - Dell 2007 FPW
Mediacom Max Cable 15meg/1meg - Tweaks not needed
2gb Skill HZ - 8800 GTX - 3x 36g Raptors in raid0
3x 250gb Drives and 4x 250gb USB Drives
CoolerMaster 750w psu - 2407 FPW & 2007 FPW LCD's
----------------------------------------------------
E6600 ES @ 3.6 @ 1.36v - Asus P5N-E SLI
2GB Gskill HZ's 1200mhz 5-5-5-15 @ 2.27v
250gb Sata II - 7800 GTX - Dell 2007 FPW
Mediacom Max Cable 15meg/1meg - Tweaks not needed
RE : PPPoE vs. PPPoA data steam efficiency for speed

its just how the protocol was written ...
unfortunately disagree.
so if
every packet = 'useful data' + mentioned 'overhead' (several components)
there is not exactly as above was explained
e.g. :
¤ TimeStamp ON means +12bits (but from MSS)
¤ PPPoE 'overhead' size : native wXP =20bits vs. regular =8bits
means +12bits 'overhead' (MTU=1480)
It is one of the reason to recommend router (in this case as hdw PPPoE client)
OTOH same of NOT recommended routers
have : MTU stuck; PMTU problems (fragmentation & retransmissions)
but also increase RTT (check latency with every attached PC) etc.
for wXP also should be fixed lan IP addresses etc
¤ PPPoA 'overhead' size is ISP depended - no doubts
and with regular MTU=1492 bits as a rule
'several blind bits' for fulfill 52bits size cells are used
e.g. 1492bits packet / 52bits cells = 28,692307692307692307692307692308
BTW especially for Emerson222 hyperlink file://C:\WINNT\System32\calc.exe was used by me
OTOH if you see on PPPoE vs. PPPoA 'total data flow efficiency'
(in meaning : useful data flow / total data flow)
for speed it importance is significant from connections >>5MBps
IMO for less is only speculated in wishful thinking
I just installed 64 this is my defaults i must have installed it wrong
SiSoftware Sandra
General Information
Windows System : Microsoft Windows XP/2003 Professional (Win64 x64) 5.02.3790 (Service Pack 1)
Full Name : Microsoft Windows XP 5.2.3790 : Service Pack 1
MTU = 1500
MTU is fully optimized for broadband.MSS = 1460
Maximum useful data in each packet = 1460, which equals MSS.
Default Receive Window (RWIN) = 65535
RWIN Scaling (RFC1323) = 0 bits
Unscaled Receive Window = 65535
Note: Under Windows 9x, if you have RWIN set to any other value, and the Analyzer reports 65535 you might need to install the MS Vtcp386 fix.
For optimum performance, consider changing RWIN to a multiple of MSS.
Other values for RWIN that might work well with your current MTU/MSS:
513920 (MSS x 44 * scale factor of 8)
256960 (MSS x 44 * scale factor of 4)
128480 (MSS x 44 * scale factor of 2)
64240 (MSS x 44) bandwidth * delay product (Note this is not a speed test):
Your RcvWindow limits you to: 2621.4 kbps (327.675 KBytes/s) @ 200ms
Your RcvWindow limits you to: 1048.56 kbps (131.07 KBytes/s) @ 500msMTU Discovery (RFC1191) = ONTime to live left = 118 hops
TTL value is ok.Timestamps (RFC1323) = OFFSelective Acknowledgements (RFC2018) = ONIP type of service field (RFC1349) = 00000000 (0)
SiSoftware Sandra
General Information
Windows System : Microsoft Windows XP/2003 Professional (Win64 x64) 5.02.3790 (Service Pack 1)
Full Name : Microsoft Windows XP 5.2.3790 : Service Pack 1
MTU = 1500
MTU is fully optimized for broadband.MSS = 1460
Maximum useful data in each packet = 1460, which equals MSS.
Default Receive Window (RWIN) = 65535
RWIN Scaling (RFC1323) = 0 bits
Unscaled Receive Window = 65535
Note: Under Windows 9x, if you have RWIN set to any other value, and the Analyzer reports 65535 you might need to install the MS Vtcp386 fix.
For optimum performance, consider changing RWIN to a multiple of MSS.
Other values for RWIN that might work well with your current MTU/MSS:
513920 (MSS x 44 * scale factor of 8)
256960 (MSS x 44 * scale factor of 4)
128480 (MSS x 44 * scale factor of 2)
64240 (MSS x 44) bandwidth * delay product (Note this is not a speed test):
Your RcvWindow limits you to: 2621.4 kbps (327.675 KBytes/s) @ 200ms
Your RcvWindow limits you to: 1048.56 kbps (131.07 KBytes/s) @ 500msMTU Discovery (RFC1191) = ONTime to live left = 118 hops
TTL value is ok.Timestamps (RFC1323) = OFFSelective Acknowledgements (RFC2018) = ONIP type of service field (RFC1349) = 00000000 (0)
Comptia a+ n+