Ahmed Ali El Din, highly valued researcher at Elastisys, has successfully defended his doctoral thesis entitled “Workload Characterization, Controller Design and Performance Evaluation for Cloud Capacity Autoscaling” on October 2, 2015 at Umeå University. His work on auto-scaling and workload characterization is at the very core of the Elastisys cloud platform’s predictive and pro-active auto-scaling features.
Ahmed’s post-graduate work started by developing very smart auto-scaling algorithms. He quickly realized, though, that no matter how smart an algorithm may be, it will not be great for all workloads. Thus, he started studying the essential properties of server workloads for cloud services. That research turned into workload characterization, which is about determining what
class a workload pattern belongs to. This is based on similarities with other, already known workloads. Using that information, one can configure a set of auto-scaling algorithms to perform better. Ahmed’s research shows that, e.g., one configuration is better for a web site experiencing rather
normal load, whereas another configuration is better if the load starts to resemble like that of Wikipedia’s more popular articles. He has also, as part of his doctoral thesis, compared various auto-scaling algorithms for efficiency in face of uncertainty using a metric developed solely for that purpose. It shows that there are many gains to be made by choosing a smart auto-scaling system. We are very happy that Ahmed’s research will continue to fuel the products and services offered by Elastisys.
Ahmed will stay at Umeå University for at least the coming two years as a postdoc. During that time, he will continue his work in auto-scaling, workload characterization, and remain a key member in CACTOS. CACTOS is the European Union’s research project on “Context-Aware Cloud Topology Optimisation and Simulation”, which will optimize how cloud data centers operate, today and in the future.
Awesome job, Ahmed! Keep up the great work, and we are really happy for you!