home *** CD-ROM | disk | FTP | other *** search
- Path: news.ultranet.com!usenet
- From: "Albert P. Belle Isle" <belleisl@cerberus-sys.com>
- Newsgroups: alt.winsock.trumpet,alt.winsock,comp.dcom.modems
- Subject: Re: TCP: Unacceptable Segments
- Date: Sat, 30 Mar 1996 14:05:33 -0500
- Organization: Cerberus Systems, Inc.
- Message-ID: <315D85FD.601E@cerberus-sys.com>
- References: <4jebju$glg@iz.comcat.com>
- NNTP-Posting-Host: apb-p5-90.cerberus-sys.com
- Mime-Version: 1.0
- Content-Type: text/plain; charset=us-ascii
- Content-Transfer-Encoding: 7bit
- X-Mailer: Mozilla 2.01 (Win16; I)
-
- Howard Arons wrote:
- >
- > I'm tuning Trumpet's MSS, MTU and RWIN using Al Belle Isle's
- > outstanding FAQ as a guide. Current settings are MTU=1500, RWIN=4096,
- > MSS=1024. These have given reasonable download speeds, but my TCP
- > trace shows intermittent "Unacceptable Segment" messages that must be
- > slowing things down. Trumpet reports no PPP Frame errors or COM
- > overruns.
- >
- > The Unacceptable Segments do not have the same id, although seg
- > 00000001 is often involved after a resync and connect to a new Home
- > Page. The lines below are typical:
- > 2104.8 80->1039 seq 00003AB2 ack 000000E7 ACK wind 14335 data 512
- > 2104.8 1039->80 seq 000000E7 ack 00003CB2 ACK wind 3584
- > 2104.9 80->1040 seq 00000001 ack 000000E8 PSH ACK wind 14335 data 78
- > Unacceptable segment : seg.seq=00000001 rcv.nxt=000000B3
- >
- > BTW, what is a "PSH ACK"?
- >
- > Can anyone offer any suggestions? Is this just an effect of Net
- > traffic, or can I hope to cure the problem and get better speed?
- >
- > TIA,
- > --
- > Howard Arons
-
- Howard:
-
- I believe that "PSH" denotes a "push" flag included in the TCP header telling any
- relaying TCP servers (like proxies or firewalls) not to hold the segment until an
- RWIN-full has arrived before forwarding it directly to the intended end-recipient.
-
- Received TCP segments are "unacceptable" most frequently because their data has already
- been "accepted" by your winsock. They are re-transmissions from a sender whose RTOmax
- timed-out before he received your winsock's ACK for that data segment.
-
- Often, packets routed through different paths arrive out of sequence, so your winsock
- holds them until the laggard packet arrives, at which time it ACKs all bytes through
- the last sequence number in the last-sent packet.
-
- This tolerance for holding packets (and delaying ACKs) increases with the size of your
- RWIN, which sets the maximum TCP flow-control window size. However, a larger RWIN gives
- better average throughput over routing paths available only in bursts.
-
- Trade-off.
-
- Re-transmissions can also be stimulated by the overhead needed to reassemble fragmented
- IP transmission units; especially if compounded by multi-path routing. That's why the
- "lazy-man's choice" of TCP MSS=536 (the Internet router default MTU of 576 less 40
- bytes of space for the IP and TCP headers) is usually a good "set-it-and-forget-it"
- value for downloads from outside your ISPs Local Area Network (over the Internet).
-
- Once your winsock tuning yields modem-limited speeds (3.2KBytes/sec for uncompressible
- files over a 28.8Kbps connection), no further tuning will make your modem any faster.
- (A direct connection is a different story, of course.)
-
- Regards,
-
- Al
-
- --
- ==================================================================
- Albert P. Belle Isle
- Cerberus Systems, Inc.
-
- Al's Winsock Tuning FAQ -
- http://www.cerberus-sys.com/~belleisl/mtu_mss_rwin.html
- ==================================================================
-