dave@onfcanim.UUCP (Dave Martindale) (09/06/88)
It seems that the A/UX NFS client code never does reads of more than 1024 bytes in one request to the server. Is there any way of getting around this limitation? Our other NFS machines show much better NFS performance when they read large blocks (e.g. 8192 bytes). However, I cannot get the Mac to do the same. I've tried using "rsize=8192" when doing the mount, and using "dd" to make sure that the user code was using a large buffer, to no avail. Is there a tunable kernel parameter that controls this? (I looked and couldn't find one). Or is the problem that the Ethernet card can't handle more than one packet at a time, and forces all requests to be less than or equal to 1024 bytes to ensure that just one packet will come back? In that case, can I get at least a full Ethernet packet (1536 bytes)?