@device[x2700] @make[report] @flushleft[For the attention of the E.C.U.C.] @majorheading[Report on Communications] @section[General] The most significant change to the network since October has been the phasing out of most of the TCPs. There are now only 2 active TCPs left and the total number of PADs has risen to 76. Most of the transition took place, as planned, over the Xmas vacation. @section[Switches] @subsection[GEC Switches] Since the last report @begin[verbatim] S/W H/W S/W & H/W Total Crashes Crashes Environmental MTBF MTBF ------- ------- ------------- ------- ---- KB GEC 4190 2 1 2 33 days 20 days AT GEC 190 1 1 1 50 days 33 days AT GEC 4160 5 1 1 17 days 14 days @end[verbatim] For both the major switches, these figures represent an improvement over the figures in the October report. Unfortunately, the overall picture has been marred by crashes on the smaller switch in the Appleton Tower. This switch's instability has been largely caused by the introduction of new software in November on the King's Buildings switch with an attempt to manage all switches in a more coherent manner. A further version of software, Type 2 Release 8, was installed at the beginning of term which has improved, but not completely cured the situation. @subsection[TelePAC Switches] A new release of software was installed in the two main TelePAC switches which has cured a number of problems, including a significant improvement in the time taken to connect a call. The MTBF figures since October are as follows:- @begin[verbatim] S/W H/W S/W & H/W Total Crashes Crashes Environmental MTBF MTBF ------- ------- ------------- ---------- ----- TelePAC 1 (KB) 0 0 1 >101 days 101 days 2 (KB) 0 0 2 >101 days 50 days 3 (Bush) 1 0 1 101 days 50 days @end[verbatim] There has, however, been a software problem with the largest switch which causes individual lines to be reset and all calls cleared from time to time. The total incident rate has been about 4 per month over the entire switch. The problem has been identified and a new release is expected within the next couple of weeks. Where possible with all switches any software maintenance has been done at the weekends. The main KB GEC switch is still taken out at lunch time (twice in the reporting period) because of the observed 'knock-on' effects to other switches. @section[PADs] The overall period between incidents on PADs is about 12 days. The problem has been raised with CAMTEC and it has been established that it is a more general problem in the community. We now have a further set of patches which ought to fix at least some of the problems. The most frequent cause of crashes is lost space, to help avoid crashes during the day the PADs are polled at 5 a.m. and if a PAD is low in space a reload to recover space is performed immediately. @section[Front Ends] Overall the performance of the EMAS and BUSH front ends have been satisfactory although they have suffered from a short spate of problems on one front end on each system. The cause of this is still being investigated. During the peak of last term the two front ends on EMAS-A became overloaded because of an inadequate communications board. To spread the load, a third front end was added improving performance although editing response was still being delayed more by the FEPs than the Amdahl. During the Xmas vacation a new X.25 board was introduced that can support line speeds in excess of 48 kb/sec; the higher line speed avoids queueing for interactive calls and much faster file transfers. This board is now in place on two of the three FEPs. As the user load built up in the first week of term, a number of crashes have occurred because of software faults, but these have been largely fixed and a much better quality of service is expected for this term. @section[INFO] Info suffered from a head crash at the beginning of term. Soon after that there were processor and memory faults, probably caused by disruption to the hardware. This meant that Info was out of service for several days. Over the past months a new Info has been under development which will offer the users more information than the last one was capable of, particularly on EMAS-A and UNIX hosts. The new Info has been released to the operators in the first instance and when it has been upgraded it will be released to users as a whole. Access to it will be direct, rather than through XCALL. @section[PSSE, XCALL] Service has been satisfactory. @flushright[Brian Gilmore 22nd January 1987.]