180 likes | 317 Views
Autonomic microcell assignment in massively distributed online virtual environments. Bruno Van Den Bossche , Bart De Vleeschauwer , Tom Verdickt , Filip De Turck , Bart Dhoedt , Piet Demeester Journal of Network and Computer Applications, vol . 32, no. 6, 2009. Outline. Overview
E N D
Autonomic microcell assignment in massively distributed online virtual environments Bruno Van Den Bossche, Bart De Vleeschauwer, Tom Verdickt, FilipDe Turck, Bart Dhoedt, Piet Demeester Journal of Network and Computer Applications, vol. 32, no. 6, 2009
Outline Overview Introduction Problem description Algorithm description Simulations & Results Conclusions Reference
Overview • In this paper we present a number of algorithms that determine the microcell allocation and runtime adaptations of the microcell allocation to optimize the deployment. • pro-active • reactive • The microcell approach and the algorithms are evaluated through large scale simulations with input parameters based on measurements on a prototype implementation.
Introduction Microcell Hotspot Player migration Inter-server/microcell communication
Problem description • To solve when using the microcell approach is finding an allocation of the microcells over the servers in such a way that the maximum load experienced by these servers is minimized. • The load that an individual microcell generates consists of a number of components. • the actions that are performed by the players • these actions should be forwarded to neighboring microcells or receive from neighboring microcells • players who move between microcells
Problem description (Cont.) The load of each server
Algorithm description Assignment a = currentAssignment; Server s = highestLoadedServer (); cellreassigned = true; while(cellreassigned){ Microcell c = highestMigration (s); Server s = migrationNeighbor (c); Assignment b = reassign (a,c,s); improvement = a.maxLoad - b.maxLoad; if(improvement > threshold) s = highestLoadedServer(); else cellreassigned = false; } Continuous Strong Locality Adjustment (CSLA)
Algorithm description (Cont.) Assignment a = currentAssignment; for(Player p : monitoredplayers){ Microcell c = getMicrocell(p); Neighbors n = getNeighbors(c); Server s = c.getServer(); Assignment b = reassign(a,n,s); if(loadImprovement(a,b)) a = b; } Player Route Adjustment (PRA)
Algorithm description (Cont.) Assignment a = currentassignement; for(iterations){ Assignement b = ILPSolve(a, timeLimit, nrAllowedMoves); if(b.maxLoadoa < maxLoad) a = b; else no improvement: break; } • Linear Limited Microcell Move Algorithm (LLMCM) • Using Integer Linear Programming(ILP) model • NP-hard
Algorithm description (Cont.) • Locality Aware Load Shedding (LALS) • The basic principle of this algorithm is to shed cells from overloaded servers.
Simulations & Results MASON
Conclusions We introduce a number of algorithms to autonomic optimize the microcell assignment. The obtained simulation results show that proactive continuous algorithms perform significantly better than reactive algorithms but at the cost of extra microcell reassignments and thus an additional runtime overhead.
Reference Bruno Van Den Bossche, Bart De Vleeschauwer, Tom Verdickt, Filip De Turck, Bart Dhoedt, and Piet Demeester. 2009. Autonomic microcell assignment in massively distributed online virtual environments. J. Netw. Comput. Appl. 32, 6 (November 2009), 1242-1256.