Packet Loss Causes

Status
Not open for further replies.

a1nerd

Posts: 13   +0
I have extremely fast dsl connection and after it's on for a while i get alot of packet loss. Wehn i ping yahoo i can get a 25% loss and somemtimes it takes a while for a page to download because of the packet loss. What are some ways i can fix this?
 
you can't. Two major factors are
1) DSL is just digital over the phone lines which are subject to all kinds of
'noise'. Sometimes a dial-up can't even sustain 9.6kb. I have experienced even hearing
the cross-talk from a neighbor when calling a friend.

2) the very nature of TCP/IP is 'contention'; the drivers write w/o permission and
then find out after the fact that the packets had collisions and need to be resent.
This is a major reason that TCP/IP only performs to ~70% efficiency.

The interesting issue is 'where do the packets get lost'?
If you are using TRACERT (Unix=traceroute), then it's the losses that are used to find the existence
of the intermediate nodes - - kind of interesting.
In addition, assuming a simple 3 nodes end-to-end (a->b->c), once packets
are received at b, any losses to c are recovered(resent) from b, not a(greatefully).

Ping will report losses from A --> .... -->Z and there's not way to improve the
efficiency end-to-end other that to (somehow) get a totally different routing.

Packet loss is a way of life.
 
Is there any simple way to fix it? It's so bad that i cant even load websites.


jobeard said:
you can't. Two major factors are
1) DSL is just digital over the phone lines which are subject to all kinds of
'noise'. Sometimes a dial-up can't even sustain 9.6kb. I have experienced even hearing
the cross-talk from a neighbor when calling a friend.

2) the very nature of TCP/IP is 'contention'; the drivers write w/o permission and
then find out after the fact that the packets had collisions and need to be resent.
This is a major reason that TCP/IP only performs to ~70% efficiency.

The interesting issue is 'where do the packets get lost'?
If you are using TRACERT (Unix=traceroute), then it's the losses that are used to find the existence
of the intermediate nodes - - kind of interesting.
In addition, assuming a simple 3 nodes end-to-end (a->b->c), once packets
are received at b, any losses to c are recovered(resent) from b, not a(greatefully).

Ping will report losses from A --> .... -->Z and there's not way to improve the
efficiency end-to-end other that to (somehow) get a totally different routing.

Packet loss is a way of life.
 
First, when you see bad performance, test from you to your ISP, if this is poor,
your settings *may* improve things. If this is good, you're in for an up hill battle.

First increase TTL (Time to live) but don't go crazy here
then decrease MTU (eg: dial-ups should NEVER use 1500)

Go to microsoft and search for TCP performance to find this values and
how to set them.
 
try reviewing this post HOWEVER,
if you're not familiar with REGEDIT, just to get a beer and leave things alone.

TTL
HKLM\SYSTEM\CurrentControlSet\Services\Tcpip\Parameters
DefaultTTL="64" (DWORD, recommended setting is 64. Other settings that are widely used are 128 and 32)

MTU of 1480 is a good starting point and work DOWN
Note: For Windows XP PPPoE, there is an additional location for MTU that might need to be adjusted (to 1480, or up to 1492 as per the PPPoE specs), depending on the PPPoE software you use. Check the following location in the Registry:
HKLM\SYSTEM\CurrentControlSet\Services\NdisWan\Parameters\Protocols\0
ProtocolMTU="1480"
 
Status
Not open for further replies.
Back