Start page
Ukranian
Golub Sergey Vladimirovich
Speciality "Software of AS"
E-mail: s_v_golub@ukrtop.com
Bybliotek Search results Links
Resume
Avtoreferat
Master dissertation theme:
Scientific chief:
Biography
Common information
I was born in 18 September 1980 in Shachtarsk.
In 1986 I went to school ¹20 of Kirovskoe town in Donetsk state.
In 1990 I passed to school ¹1 of the same town. I studied well. I won in town's olympiads by mathematics,
chemistry, informatics. In 1997 I took a part in regional olympiad by informatics. Took an active part
in â public life of my school, played in school-teams in volleyball, basketball
and football. In 1997 I graduated from school ¹1 with gold medal, in the same year
entered in Donetsk technical state university with speciality "Software of automated systems". I studied well, took a part in scientific conferences. I well graduated military chair of our university. I have a grants of science convocation of our university and state administration.
Motives of specialty choice
In 1990 I was in ASÑ of coalmine "Komsomolets Donbassa", where in that time my mother worked. First contact with
computer made a great impression for me and I decided: I would be a programmer.
Master dissertation
My master dissertation is about processor's dynamic load balancing in multiprocessor systems (scheduling). I chose this problem, because it is very interesting.
The scheduling problem is one of the most widely-studied problems, with almost limitless number of variants. Here is the first variant of task: we have m machines è n unpreemtable jobs. Completion time of job executionis positive real value, these reals are known before job execution. We need to schedule all jobs to minimize makespan, last job completion time.
Formally the problem is this: Given sequence of positive reals a1,a2, …, an and an integer m; for each j assign aj to machine i, 1 <= i <= m, so as to minimize th maximum, over i, sum of all aj assigned to machine i. There are many other variants of this problem, but we choose on-line version: as soon as job j arrives, it must be assigned immediately to the one of machines. For sequence of jobs s let A(s), a random variable, denote the makespan of the schedule, generated by (randomized or deterministic) algorithm A , and let OPT(s) denote the minimum makespan among all m-machine shedules for s. A's competitive ratio is then:
CA= lim(A(s)/OPT(s)),
where supremum is over all sequences of jobs.
The natural question is: how small CA can be?
Begin of the page
Bybliotek Search results Links
Resume
Avtoreferat
Start page
Ukranian
Last update: 25.04.2002