centers) can have arbilxary probability distribution.
In our model, the first delay center (Center 1)
models the thinking period between the time a
caller thread finishes an RPC and the time it starts a
new WC. We assurne that the thinking times
follow the sane arbitrary probability distribution.
Thus this center has a single class of jobs. The
second delaly ceinter (Center 3) models the time a
component thread spends waiting for a response for
an invoked primitive lkom its peer component. As
shown in Figure I,th ere are three different waiting
actions. We assume that the waiting times for them
follow three independent arbitrary probability
distributions with different mean service times.
Thus this ceinter has three different classes of jobs.
From the abiove description and assumptions, it can
be seen that the queueing model is a closed single
chain multiclass queueing network. The network
consists of' three serviice centers, of which one has
the PS queueing discilpline and the other two have
the IS queueing discipline. The network has no
externsell job arrivals. lJnder the assumption that N
caller threaids in the application process are
invoking RPCs or thinking, there are N jobs
circulating in a closed single chain in the network.
These jobs belong to 18 classes and change their
class after visiting (a service center to reflect
differenlt service requirements. Different classes of
jobs have different different service time
distributions. Bared or1 the descriptions of queueing
discipliines, job classes, and service time
distributions about the network, it can be shown
that the queuenng network is a . product form
network [6].
centers) can have arbilxary probability distribution.In our model, the first delay center (Center 1)models the thinking period between the time acaller thread finishes an RPC and the time it starts anew WC. We assurne that the thinking timesfollow the sane arbitrary probability distribution.Thus this center has a single class of jobs. Thesecond delaly ceinter (Center 3) models the time acomponent thread spends waiting for a response foran invoked primitive lkom its peer component. Asshown in Figure I,th ere are three different waitingactions. We assume that the waiting times for themfollow three independent arbitrary probabilitydistributions with different mean service times.Thus this ceinter has three different classes of jobs.From the abiove description and assumptions, it canbe seen that the queueing model is a closed singlechain multiclass queueing network. The networkconsists of' three serviice centers, of which one hasthe PS queueing discilpline and the other two havethe IS queueing discipline. The network has noexternsell job arrivals. lJnder the assumption that Ncaller threaids in the application process areinvoking RPCs or thinking, there are N jobscirculating in a closed single chain in the network.These jobs belong to 18 classes and change theirclass after visiting (a service center to reflectdifferenlt service requirements. Different classes ofjobs have different different service timedistributions. Bared or1 the descriptions of queueingdiscipliines, job classes, and service timedistributions about the network, it can be shownthat the queuenng network is a . product formnetwork [6].
การแปล กรุณารอสักครู่..
