paid off~ The 6600 is very sensitive and requires regular tuning. Power transJ.ents even with redundant conneetions~ flywheels and generators occasionally cause system problems. Efforts have been made to determine dynamically what system is doing and then automate the allocation of resources. Rather than confuse users with acct,rate accounting charges with all the variables~ a consistent calculated charge (even if larger) based upon data. handled has made the users happier The centre] facility may suffer from ~mproper programming of users. VII A COMPARISON OF OPERATING PROBLEMS FOR TIME SHARING AND BATCH SYSTEMS by J~ Harper A Review by Warren P. Burrell Jo Harper~ Chief of the ESSA Computer Center at Boulder, described operations for a batch and a T/S computer facility. The batch operations have evolved from 650~ 1604~ 7090 and 3600 to a 65 K 3800 with a million word drum configuration. For the order of 450 active users (12% at NBS) batch jobs have averaged two minutes at $350 per minute User feedback has been excellent Interrupts were used to determine what goes on in the processor and have permitted FORTRAN job times to be cut by 50% Based upon what appeared to be successful experience in 1965 at G.E. in Phoenix and Ccm-share and Tymshare with SDS gear~ a 940 T/S system for 24 lines was axquired. Experience in separation of the facts from propaganda has been unsaEisfactory since the system has been down most of the time. The 24 lines at Boulder were farther complicated by requirinz two separate exchanges in the local telephone company. At the beginning of May 1968~ new software and renewed support by SDS turned the situation around. Since mid-May~ the system was up approximately 99% of the time. On 35 TTYVs, 450 active users have been averaging 850 to 900 Lseful terminal hours per month with i000 system time. The system is still considered experimental with efforts continuing to collect data for its justification~ There is no basis for comparison of T/S with batch. Good experience is in the minority for T/S. Generally~ the users are of a different type with research applications. Much greater system reliability is needed because when the system goes down~ all users know it right away and have lost their files. The operator assumes many of the functions of the systems programmer, while the systems programmer becomes more of an analyst. Better analysts are needed to improve the user feedback which currently is very poor compared to batch systems. I m p r o v e d manuals and education in general is really needed for T/S. Most of the apparent success in commerical T/S the fact they are not completely sharing the operati sational interactive T/S system concept is still an r.o~eonse. As the load upon resources increases~ the ___~_. poor and all conversation ceases. Therefore~ the diversity of resource allocati~n: for T/St response time over efficiency; for batch~ efficiency over response. Sepaation of pricing for software and more open competi.tion for services may force the issue and resolve the basis to compare T/S with batch. -47-

/lp/association-for-computing-machinery/review-of-a-comparison-of-operating-problems-for-time-sharing-and-0QZXV2fF4K