Desktop workload study with implications for desktop cloud resource optimization

Desktop cloud is a new delivery model in which end users connect to virtual desktops running in remote data centers. This paradigm offers multiple benefits both in terms of manageability as well as efficiency improvements. However, realizing this potential requires better understanding of desktop workload and its implications for desktop consolidation. We analyze CPU and memory usage on a sample of 35 desktops using a fine-grained 10 second averaging interval. Results provide insights into achievable efficiency improvements from desktop consolidation as well as detailed autocorrelation and variability behavior as a function of number of aggregated desktops. We also propose an interactivity classification method leading to functional form suitable for estimating residual durations of interactivity states. This finding can be leveraged in on-line proactive management algorithms for desktop cloud optimization.

[1]  Teunis J. Ott,et al.  Load-balancing heuristics and process behavior , 1986, SIGMETRICS '86/PERFORMANCE '86.

[2]  Edward D. Lazowska,et al.  The limited performance benefits of migrating active processes for load sharing , 1988, SIGMETRICS '88.

[3]  Philip H. Dorn,et al.  The Soul of a New Machine , 1982, Annals of the History of Computing.

[4]  Andrzej Kochut,et al.  On Strategies for Dynamic Resource Management in Virtualized Server Environments , 2007, 2007 15th International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems.

[5]  Mor Harchol-Balter,et al.  Exploiting process lifetime distributions for dynamic load balancing , 1995, SIGMETRICS.

[6]  Vanish Talwar,et al.  Architecture and Environment for Enabling Interactive Grids , 2003, Journal of Grid Computing.

[7]  Helmut Hlavacs,et al.  Modeling user behavior: a layered approach , 1999, MASCOTS '99. Proceedings of the Seventh International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems.

[8]  Rajdeep Bhowmik,et al.  Managing responsiveness of virtual desktops using passive monitoring , 2009, 2009 IFIP/IEEE International Symposium on Integrated Network Management.

[9]  Carl Staelin,et al.  Idleness is Not Sloth , 1995, USENIX.

[10]  S ChaseJeffrey,et al.  Managing energy and server resources in hosting centers , 2001 .

[11]  Robert P. Goldberg,et al.  Survey of virtual machine research , 1974, Computer.

[12]  Robert J. Creasy,et al.  The Origin of the VM/370 Time-Sharing System , 1981, IBM J. Res. Dev..

[13]  Jeffrey C. Lagarias,et al.  Convergence Properties of the Nelder-Mead Simplex Method in Low Dimensions , 1998, SIAM J. Optim..

[14]  Amin Vahdat,et al.  Managing energy and server resources in hosting centers , 2001, SOSP.

[15]  Jeffrey K. Hollingsworth,et al.  Exploiting Fine-Grained Idle Periods in Networks of Workstations , 2000, IEEE Trans. Parallel Distributed Syst..

[16]  David J. Bradley.  The creation of the IBM PC , 1990 .

[17]  Karl Aberer,et al.  Stochastic resource prediction and admission for interactive sessions on multimedia servers , 2000, ACM Multimedia.

[18]  Arun Kumar,et al.  Stream-Packing: Resource Allocation in Web Server Farms with a QoS Guarantee , 2001, HiPC.

[19]  K. Shin,et al.  Performance Guarantees for Web Server End-Systems: A Control-Theoretical Approach , 2002, IEEE Trans. Parallel Distributed Syst..

[20]  Andrzej Kochut,et al.  Dynamic Placement of Virtual Machines for Managing SLA Violations , 2007, 2007 10th IFIP/IEEE International Symposium on Integrated Network Management.

[21]  Günter Haring,et al.  On Stochastic Models of Interactive Workloads , 1983, Performance.

[22]  Peter H. Gum,et al.  System/370 Extended Architecture: Facilities for Virtual Machines , 1983, IBM J. Res. Dev..