hdtodd@eagle.wesleyan.edu (04/22/91)
This is a quick followup to my posting of last week regarding problems booting diskless 2500's over TCP/IP through a bridge. We thought we'd found that the server 2500 put out packets every 2.5msec; the bridge passed them through every 3.5msec; the bridge dropped packets occasionally; the boot protocol is not sufficiently robust for the diskless node to recover from the dropped packets. We replaced the 8KB-buffer WD8003's in the PCBRIDGE bridge (an AT&T 6300 running the public-domain PCBRIDGE program) with WD8003EB's with 32KB buffers: the diskless systems now boot smoothly. Of course there is a performance hit relative to having directly-wired machines with no bridge, but the performance is quite satisfactory. This might help others who run into similar problems with diskless systems. David Todd Wesleyan University