johns@house.house.gov (John Schnizlein) (10/19/90)
We have a repeatable dramatic performance problem using NCSA telnet on Macintosh computers connected to LocalTalk subnetworks gatewayed to an Ethernet backbone by Shiva (Kinetics) FastPath 4 configured with IP subnetting (PROM code 4.1). When the Macintosh telnet client gets a large volume of text from an application on an IBM 3090 running ACCes/MVS connected to the backbone by an Intel FastPath, it seems to function correctly for a while, then it suffers progressively longer and more frequent delays in receiving bursts (packets ?) of text. What is worse, this slowdown condition remains after the text retrieval for the remainder of the logon session. Subsequent retrievals are slow even if they return small or moderate volumes of text. Concurrent telnet sessions through the same Shiva FastPath to different hosts and by clients which have not received large volumes (including tn3270 sessions) do not experience the slowdown. This problem seems independent of other random slowdowns which intermittently quarter the performance of FTP and degrade telnet speed horribly. Help on the random problem would be appreciated also, but solving the repeatable problem is urgent. Has anyone found the same problem and a fix? E-mail help please to johns@house.house.gov