msr@bunker.UUCP (03/05/87)
I'm in the process of looking at some lan software/hardware for some
VAX 780's (yes, there are 780's out there without ethernet) and PC clones.
I'd appreciate if some kind souls would share their experiences with
hardware/software vendors for:
. Ethernet cards for the VAX
. NFS from wherever (including Ultrix 2.0)
. Anyone using NFS on a PC ?
. Anyone using FREEDOMNET ?
Also any words of wisdom about some of the boards with some of the TCP/IP
protocol on them ? Specifically, would I have problems between an
Excelan Exos 204 and a DEC DELUA ?
Thanks in advance,
--
Mike Romaniw - Bunker Ramo/Olivetti, Shelton, CT - (203)337-1407
uucp address: ...{decvax,philabs,ittatc}!bunker!msrron@brl-sem.ARPA (Ron Natalie <ron>) (03/05/87)
In article <1612@bunker.UUCP>, msr@bunker.UUCP (Michael S. Romaniw) writes: > I'd appreciate if some kind souls would share their experiences with > hardware/software vendors for: > > . Ethernet cards for the VAX DEC DEQNA, DELUA Interlan N1010A Excelan > Also any words of wisdom about some of the boards with some of the TCP/IP > protocol on them ? Specifically, would I have problems between an > Excelan Exos 204 and a DEC DELUA ? The TCP/IP on the EXCELAN board is INFERIOR. You'd be happier with the TCP/IP supported on the VAX under ULTRIX (4.2). It doesn't cause that much of a load on the machine. -Ron
jhc@mtune.UUCP (03/07/87)
In article <679@brl-sem.ARPA> ron@brl-sem.ARPA (Ron Natalie <ron>) writes: >The TCP/IP on the EXCELAN board is INFERIOR. Perhaps you could expand on this comment? I have SV and therefore am pretty much stuck, short of going to v9 and using the streams IP/TCP driver. Comparisons with the CMC board? -- Jonathan Clark [NAC,attmail]!mtune!jhc My walk has become rather more silly lately.
ron@brl-sem.ARPA (Ron Natalie <ron>) (03/10/87)
In article <1021@mtune.ATT.COM>, jhc@mtune.ATT.COM (Jonathan Clark) writes: > In article <679@brl-sem.ARPA> ron@brl-sem.ARPA (Ron Natalie <ron>) writes: > >The TCP/IP on the EXCELAN board is INFERIOR. > > Perhaps you could expand on this comment? I have SV and therefore am > pretty much stuck, short of going to v9 and using the streams IP/TCP > driver. Comparisons with the CMC board? > -- Well, first, when compared to the native 4.2 TCP code it is underwelming. There is no real reason not to do it in software. Now for machines that do not have software TCP available, boards like the Excelan and CMCs are the only way to go. I haven't had any experience with the CMC board, but they seem to be a much sharper company. The Excelan boards went through a number of different problems with the firmware. Most of these were supposed to have been resolved (at least it no longer crashes when it receives packets from off the local net), however, our Silicon Graphics Iris processors have these boards still. We have one 310 system with a CMC board and its been pretty well scrutinized by one of the TCP experts, and he doesn't have much problmes with it. -Ron
wsmith@umn-cs.UUCP (03/13/87)
About Excelan cards: In article <679@brl-sem.ARPA> ron@brl-sem.ARPA (Ron Natalie <ron>) writes: >The TCP/IP on the EXCELAN board is INFERIOR. and in response to a question about non-BSD systems writes: >Well, first, when compared to the native 4.2 TCP code it is underwelming. >There is no real reason not to do it in software. > >Now for machines that do not have software TCP available, boards like >the Excelan and CMCs are the only way to go. I haven't had any experience >with the CMC board, but they seem to be a much sharper company. The >Excelan boards went through a number of different problems with the >firmware. Most of these were supposed to have been resolved (at least >it no longer crashes when it receives packets from off the local net), >however, our Silicon Graphics Iris processors have these boards still. >We have one 310 system with a CMC board and its been pretty well scrutinized >by one of the TCP experts, and he doesn't have much problmes with it. > >-Ron It should be noted that the software for the VAX/VMS on an Excelan card is quite a bit better than for the IRIS. When you buy for a VAX, you buy from Excelan and get the newest software. When you get an IRIS (we have several) you get the software SGI gives you (which is AGES old Excelan stuff). Silicon Graphics newest software release supposedly incorporates the newer Excelan code. Randy Smith wsmith@umn-cs.arpa ...!ihnp4!umn-cs!wsmith
vjs@sgi.UUCP (03/15/87)
In article <57700002@umn-cs.UUCP>, wsmith@umn-cs.UUCP writes: > It should be noted that the software for the VAX/VMS on an Excelan card > is quite a bit better than for the IRIS. When you buy for a VAX, you buy > from Excelan and get the newest software. When you get an IRIS (we have > several) you get the software SGI gives you (which is AGES old Excelan > stuff). Silicon Graphics newest software release supposedly incorporates > the newer Excelan code. > > Randy Smith > wsmith@umn-cs.arpa > ...!ihnp4!umn-cs!wsmith The much of the preceding is not true, tho a year ago it was less untrue. - In ages past, we used Excelan's 4.1a-bsd TCP in their board. It was the most current stuff we could get. I trust Excelan has better now, but that no longer affects SGI or our current customers. - Since 3.5/2.5, the IRIS uses 4.3bsd networking. Some of the user code is from the 4.3beta tape, but all of the kernel code is 'real' 4.3, with IF_NET domain sockets, select(2), network ioctl()'s, and everything. It includes the 'three important bug fixes' for 4.3 published last fall. IRIS XNS is still available. (The IRIS 3.5/2.5 kernel is System 5, with lots of 'other' stuff.) - The link-level driver for the Excelan board still used in IRIS 3000's and 2000's is derived from a 4.3bsd VAX Excelan link-level driver. The major changes were to improve performance. - We are currently shipping 3.5r2. - We have measured RCP transfer rates comparable to our XCP rates, which are quite respectable. - Since 3.5, Sun Microsystems' NFS is available for the IRIS. We are pleased with the performance of our NFS implementation. 3000's served by 3030's with Eagles can certainly do lots of compiling and system building. - 3.5 serial I/O is much improved. I have transfered large files using dumb things like 'cat /dev/ttym2 > /tmp/foo' at 38,400 bit/sec, and lost at most 1 or 2 bytes every 30 seconds as update sync's the disk. 9600, even using such stupid 'protocols' as cat-ing to disk, is clean. (Since then, we have reduced the latency caused by disk cache flushing. It also helps to use raw instead of cooked mode.) - Our uucp/usenet gateway is currently a 3030, connected to a network of hundreds of workstations, running SMTP, XNS, UUCP, RIP (routed), and NFS. (For 6 months, it was a 'bubble', which is not even a 1400!) We use IRIS workstations to gateway among ethernets. Please, upgrade your IRISes. Vernon Schryver Silicon Graphics, 2011 Stierlin Rd, Mtn.View, CA 94043 vjs@sgi.com or {pyramid,ucbvax,adobe,sun,ames,mips}!sgi!vjs