eho@word.Princeton.EDU (Eric Ho) (08/30/89)
Does anyone know what the variable SPP.USER.TIMEOUT (in koto) really stands for and how does it really affect the system ? The default value is 15000 but I don't know what unit it is in ? Any pointers appreciated. -- Eric Ho Cognitive Science Lab., Princeton University voice = 609-987-2987 email = eho@confidence.princeton.edu 609-987-2819 (messages) eho@bogey.princeton.edu regards. -eric-
welch@CIS.OHIO-STATE.EDU (Arun Welch) (08/30/89)
[We took a power hit as I was sending this out, hope it makes it] SPP is the Sequenced Packet Protocol, part of the XNS suite. It provides a reliable transmission protocol, and corresponds to the TCP protocol in the TCP/IP suite. My guess is that variable sets the time that the system waits for an ACK before it assumes that the packet got dropped somewhere enroute and sends another one. If so, mucking with it will affect your network behaviour, by sending more duplicates if it's lowered, or waiting longer if it's raised. If you have to cross many hops between your hosts and you're experiencing timeouts then I'd raise it, if you're not experiencing any problems I'd leave it alone. ...arun
vanMelle.pa@XEROX.COM (08/31/89)
This variable is used mainly for two, perhaps incompatible, measures: 1) It is an upper bound on the "round trip time estimate" that Lisp makes. An SPP connection maintains a round trip estimate based on how long it takes the other end to ack a packet we send. This estimate is used so as to have a guess about how long we should wait without a response before deciding that the other end missed our packet and we should try again. 2) Lisp waits at least this long when trying to initially establish a connection before deciding that the other server is not there. The actual timeout can be larger if the server is far away: it is the larger of spp.user.timeout and 3000*#hops. So raising it might be good if you're having a hard time connecting to servers; if not, raising it only means you'll wait longer for Lisp to give up on trying to connect to a server. Also, any temporary slowdowns in existing connections may take longer to speed back up if Lisp has cranked the round trip estimate needlessly high, but I suspect this effect is minor. The units are milliseconds. Bill