Dilum Bandara

Work on workload, resource, and cost aware VM allocation in IaaS and PaaS clouds have resulted in several publications. Our key contributions include the following:

• A proactive, workload, resource, and cost-aware auto-scaling solution for PaaS cloud that combines a predictive model, cost model, and smart killing. An ensemble workload prediction mechanism is introduced based on time series and machine learning techniques for making accurate predictions on drastically different workload patterns.
• Given that cloud operator exposes a small, dynamic fraction of its infrastructure (corresponding resource specifications and constraints), we propose a dynamic and computationally efficient VM reconfiguration scheme which comprises an Application Performance Model, a Cost Model, and a Reconfiguration algorithm. This enables cloud hosted user applications to dynamically scale while consuming just the adequate amount of resources along different types of resource dimensions, more than just CPU and memory.
• A resource and policy aware VM scheduling solution for medium-scale IaaS clouds that enables the deployment of VMs based on a predefined set of policies and user priorities, while being aware of the resource utilization of the cloud.

Work related to Complex Event Processing (CEP) span across both the architectural aspects of CEP engines and use of CEP for multidisciplinary applications. Some of the examples include:

• Porting WSO2 Siddhi CEP engine to GPU
• Automated CEP query generation
• Scatter and gather based approach for scaling stateful CEP operators
• Out-of-order event processing in CEP
• CEP and machine learning for weather detection
• CEP engine for IoT devices

Resource-rich computing devices, decreasing communication costs, and Web 2.0 technologies are fundamentally changing the way distributed applications communicate and collaborate. With these changes, we envision Peer-to-Peer (P2P) systems that will allow for the integration and collaboration of peers with diverse capabilities to a virtual community thereby empowering it to engage in greater tasks beyond what can be accomplished by individual peers, yet are beneficial to all the peers. Collaborative P2P systems are applicable in a wide variety of contexts such as:

• Distributed Collaborative Adaptive Sensing (DCAS) systems such as Collaborative Adaptive Sensing of the Atmosphere (CASA)
• Grid and cloud computing
• P2P clouds, i.e., community-based cloud computing
• Mobile and social P2P networks
• Global Environment for Network Innovations (GENI)

Collaborations involving application-specific resources and dynamic quality of service goals will stress current P2P architectures that are designed for best-effort environments with pair-wise interactions among nodes with similar resources. These systems will share a variety of resources such as processor cycles, storage capacity, network bandwidth, sensors/actuators, services, middleware, scientific algorithms, and data. Goal of this project is to address all related challenges of key phases of resource collaboration such as resource advertising, selection, matching, and binding as well issues related to incentives, trust, security, and privacy. Initial work started while I was at CNRL. Some of the sub-projects include:

• ResQue — A multi-attribute resource and range query generator. Generates random computing nodes with multiple static and dynamic attributes over a given time period and multi-attribute range queries.
• BitTorrent Search Engine/Site Survey – A survey conducted to understand whether users access multiple BitTorrent search engines and how frequently.
• Secure patch dissemination over P2P