Self-* Datacenter Management for Business Critical Workloads

Dagstuhl Seminar 15041: Model-driven Algorithms and Architectures for Self-Aware Computing Systems

Download PDF Slides


Multi-cluster datacenters, and, further, multi-datacenter infrastructure, are supporting increasing amounts and types of computer applications. Among the workloads of these datacenters, business-critical workloads, that is, workloads that support business decision and intelligence, and that provide business and operational back-ends, are increasingly important (over 20% of the general IT load, according to an IDC study from 2010). Our goal is to build self-* resource managers for datacenters supporting business-critical workloads, to ensure efficient operation. We present in this talk new results in characterizing business-critical workloads running in multi-cluster multi-datacenters, and advances in scheduling such workloads using portfolio scheduling, scheduling by guessing (but not predicting) workload characteristics, and scheduling by enabling elasticity even for data-intensive workloads (for example, scheduling for elastic MapReduce frameworks).

1. Vincent van Beek, Siqi Shen, and Alexandru Iosup: Statistical Characterization of Business-Critical Workloads Hosted in Cloud Datacenters. CCGRID 2015 (in print, available upon request)
2. Bogdan Ghit, Nezih Yigitbasi, Alexandru Iosup, and Dick H. J. Epema: Balanced resource allocations across multiple dynamic MapReduce clusters. SIGMETRICS 2014: 329-341 ( Also available online: )
3. Lipu Fei, Bogdan Ghit, Alexandru Iosup, and Dick H. J. Epema: KOALA-C: A task allocator for integrated multicluster and multicloud environments. CLUSTER 2014: 57-65 ( Also available online: )
4. Kefeng Deng, Junqiang Song, Kaijun Ren, and Alexandru Iosup: Exploring portfolio scheduling for long-term execution of scientific workloads in IaaS clouds. SC 2013: 55 ( Also available online: )