katzman@harvard.ARPA (Cynthia Katzman) (01/09/85)
I just finished checking out the overall failure rate of 20/222, pcp cards here at Harvard. (Budget time...) This information is based on an 8 month period, with an average of 302 pcp cards in use. Infant mortality was NOT really a factor as only 40 new cards were factored in over this 8 month period. But, a glance of some information I have obtained from the early days, leads me to believe that infant mortality is really extremely high, and doesn't really represent what happens after GI gets their hands on those cards! Over the past 8 months, I've sent back 39 pcp cards, 2 modems (out of 33) and don't ask about the switch. (ok... its been back to Mountain View 3 times and still doesn't work... verified by the local Sytek tech. support) 2 of the 39 cards had to go back immediately to General Instruments. (A failure that they simply didn't fix. Re-set the one card, and it shorted the whole box into a re-set.) The problems varied a wide range, but we now have a 3 day testing procedure that generally weeds out the cards before they get into the field, either back from GI, or new. Maybe if Sytek Q.C. did a bit more testing, we wouldn't be stuck with this infant mortality problem. However, I must say I have found both Sytek and GI to be most helpful. BOTTOM LINE: Average monthly failure rate is 2%. Projected yearly failure rate is 20%. We have had a number of problems implementing these S-muxes to several hosts. EIA flow control is a major concern to use. Presently, we have it implemented on only 4 of out 20 hosts. We feel we really need EIA because of a healthy and growing EMACS lobby. DTR is often an issue as a short spike of only 380 nano-seconds blows a session away. For some reason, the Accord DH board likes to toggle DTR when using EIA flow control. We have found that after the Able DHs are wrapped for EIA, they work like a charm in all respects. There are a couple of other thorns in our side. One is the Series 1 protocol Converter having problems with both baud rate conversions and dtr (I understand the new version of the Yale package ought to help..). The second has to do with a BBN C70 that simply doesn't want to support both DTR and DCD, with EIA flow control. (EIA is essential with the PEN editor) Other than these few problems, we are rather comfortable with the LocalNet 20 product. We are looking into the possibility of using their bisync product. We have some bisync units presently in test. I'd like to hear about some practical experiences with bisync. Anyone out there? What may we be getting into? Cynthia Katzman harvard!uvax!katzman (617) 495-1272
phil@amdcad.UUCP (Phil Ngai) (01/12/85)
> thorns in our side. One is the Series 1 protocol Converter having > problems with both baud rate conversions and dtr (I understand the > new version of the Yale package ought to help..). The second has Speaking of such things, has anyone else tried the new IBM 7171 protocol converter? We just installed one and it seems to work real nice so far. -- AMD assumes no responsibility for anything I may say here. Phil Ngai (408) 749-5790 UUCP: {ucbvax,decwrl,ihnp4,allegra}!amdcad!phil ARPA: amdcad!phil@decwrl.ARPA