1 / 11

Open Queueing Network and MVA

Open Queueing Network and MVA. Cheng-Fu Chou. Jackson Networks. Assume each queue has one or more servers with expo. distributed service time, and Poisson arrival of jobs to the network

lynley
Télécharger la présentation

Open Queueing Network and MVA

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Open Queueing Network and MVA Cheng-Fu Chou

  2. Jackson Networks • Assume each queue has one or more servers with expo. distributed service time, and Poisson arrival of jobs to the network • If the network has Q queues and if ni is the number of jobs at queue I, then a Jackson network in steady state has the surprising property that • A product of marginal prob., such a network is called product-form or separable

  3. Each queue behaves like an independent M/M/K queue or M/G/1 queue • We can use a Markov chain derive an expression for Prob(ni=ki) separately and substitute it into the above eqn. • Note that the arrival process at some queue may not be Poisson

  4. Ex • Consider the open network in Fig. 3.6 with single-server queues A, B, and C with branching prob. pA, pB, pC, respectively.

  5. Let pdone = 1 – (pA +pB +pC), then, the number of visits to A has distribution Geometric(pdone). • Let VA, VB, and VC be the expected number of visits to A, B, and C, respetively. • VA =1/ pdone , VB =pB VA = pB / pdone , and VC = pC / pdone • Let lA, lB, abd lC be the job arrival rates at A, B, and C respectively. • lA = l VA , lB = l VB , lC = l VC • After we can get prob(nA=i), prob(nB=j), and prob(nC=h) , we get prob(nA=i, nB=j, nC=h)

  6. Closed Queueing Network • Closed systems are used for the interactive systems • In a closed network, we can model a set of users submitting requests to a system, waiting for results, then submitting more requests • human users interacting with a system, • threads contending for a lock, • processes blocking for I/O, • networked servers waiting for a response message.

  7. Product Forum Queueing Networks • A PFQN consists of a collection of queueing and delay centers. It satisfies the following conditions. • All queueing centers : FCFS, PS, or LCFSPR • Any delay centers • FCFS with exponential distribution • If a FCFS center has multiple service classes, they must all have the same average service time • External arrivals, if any, are Poisson • Routing is state-independent

  8. Arrival Theorem • For a separated closed network with N jobs, an arrival at a queue sees a network state that is (distribution-wise) the same as that seen by an outside observer of the same network with N-1 jobs.

  9. Mean Value Analysis • The Iterative Solution Method

More Related