multiprocessor scheduling Flashcards
(9 cards)
explain and compare the multiprocessor scheduling technique
Global scheduling :
aims to find a general strategy to schedule job where jobs are allowed to migrate such that job starts its execution in processor one and finish it in processor 2
partitioned scheduling
aims to find a strategy to partition tasks among processors where job are not allowed to migrate
why Partitioned and Global Scheduling are Incomparable
since there are tasks that Global scheduling can do, which partition scheduling can not do for example
U(t1) > 1
and vise versa
Compare RM and EDF in the context of partitioning
For RM the bound is not tight enough U(π) < nπ (2 1/nπ β 1) which leads to losing 50% of the platform capacity
For EDF itβs way more optimal where the condition of schedulability is limited to U(t) <1
the requirement FFDU are :
u(t) <=(m+1)/2 and Umax <=1
what are the scheduling anomalies
is any intuitive positive change that may lead to jeopardizing the system schedulability,
intuitive positive change might be anything that leads to decreasing the utilization such that increasing the time factor or execution time ..etc
what is the necessary condition for the schedulability of an implicit deadline system
Any Periodic Implicit-Deadline System is feasible iff U(π) β€ m and Umax(π) β€ 1
Explain EDF^k
it is a modification of EDF to fit into smaller numbers of processors such that :
for I < K-1 jobs of tasks are given high priority
for I > K-1 jobs of taski are scheduled using pure EDF
provide a sufficient condition for the schedulability of sporadic multi-p arbitrary deadline system
we can rely on the density;
density πi consists of Ci/ min(Ti, DI)
hence the schedulability test is
πsum(π) β€ m and πmax(π) β€ 1
provide a sufficient condition for the schedulability of EDF^k
U(T) < m(m-1) Umax
however, for EDF this result is quite pessimistic since it requires a big number of processors
explain the condition for the sporadic implicit-deadline system is schedulablbility on m processors
using edf(k)
i
Informally, EDF^k requires processors such that
T1 to P1 β¦..Tk-1 toPk-1 and then assigns the rest such that
Tk to PK β¦Tn to Pn