Discipline: Computer Sciences and Information Management
Subcategory: Computer Science & Information Systems
Alexander Ing - California State University Dominguez Hills
Co-Author(s): Bin Tang, California State University Dominguez Hills, California, Carson. Manar Alqarni, California State University Dominguez Hills, California, Carson.
Middleboxes, such as firewalls and load balancers, are playing an increasingly important role in cloud data centers for security or performance purposes. The recent introduction of Software Defined Network (SDN) and Network Function Virtualization (NFV) in cloud data centers has further facilitated the efficient management of software-based middleboxes in data center networks. In policy-driven cloud data centers, it requires that virtual machine (VM) traffic traverses a sequence of specified middleboxes in order to achieve security and performance guarantee. Much research has been done to study how to place middleboxes inside data centers for cost-efficient VM traffic traversal. However, not much research has focused on load balance of middleboxes. In our research we study the Load-Balanced Middlebox Assignment Problem (LB-MAP), which minimizes the communication energy cost among virtual machines (VMs) in data centers while a), satisfying their policy requirement, as well as b), the capacity constraint of the switches that the middleboxes placed upon. We show that LB-MAP is equivalent to the well-known minimum cost flow problem (MCF), which can be solved optimally and efficiently. We also design a suite of efficient heuristic algorithms based on different criteria viz. VM-Based, MB-Based, and VM/MB-Based. Via extensive simulations, we show that all the heuristic algorithms perform close to the optimal minimum cost flow algorithm, while VM+MB-Based performs best among all the heuristic algorithms. To the best of our knowledge, this is the first work that addresses the energy cost minimization for VM communications as well as load-balancing for middleboxes in policy-driven data centers. This research focuses on load balancing of multiple instances of a single type of middlebox, but further research can be done addressing load balancing for multiple instances with multiple types of middleboxes.
Not SubmittedFunder Acknowledgement(s): This research is funded through the National Science Foundation (NSF) under grant #HRD-1302873 and the Chancellor's Office of the California State University. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation or the Chancellor's Office of the CSU.
Faculty Advisor: Dr. Bin Tang, btang@csudh.edu
Role: My contribution to this research was the creation of the K-ary fat tree simulation, allowing us to change parameters within the simulated datacenter; this robust simulation was used to test the efficiency of the load balancing algorithms. Also, I participated in the collection of data through the experimentation on the three heuristic algorithms as well as the minimum cost flow algorithm. Once the data was collected I also helped analyze and interpret the results of the data from the experiments.