I received B.S. and M.S. degrees from Bilkent University, Ankara, Turkey, in 2002 and 2005, in electrical and electronics engineering. I received Ph.D. degree from Osaka University, Japan, in information science and technology, in 2008. I was the recipient of Monbukagakusho Graduate Scholarship from the Ministry of Education, Science, Sports and Culture of Japan between 2005–2008. I am a post-doctoral researcher at Osaka University since 2008. My research interests are network function virtualization (NFV), neural networks, optical networks, traffic engineering and simulation. I have worked on several government funded projects on computer networks and telecommunications in Europe and Japan. I was the recipient of "Best Paper Award" at the First International Conference on Evolving Internet (INTERNET 2009). I am the editor of IEEE Internet Initiative Newsletter. IEEE Internet Initiative connects technology experts and policymakers from around the world to foster a better understanding of, and to improve decisions affecting, Internet governance, cybersecurity, and privacy issues. I am serving in the steering committee of International Conference on Evolving Internet. I am a Senior Member of IEEE.
Ph.D. in Information Science and Technology
Dissertation: Design and Performance Evaluation of Small-buffered Optical Packet Switched Networks
M.S. in Electrical and Electronics Engineering on Computer Networks Area
Thesis: Combined Use of Prioritized AIMD and Flow-Based Traffic Splitting for Robust TCP Load Balancing
Summer School organized by European Union
Topic: Routing and Multi-Layer Traffic Engineering in Next Generation IP Networks
B.S. in Electrical and Electronics Engineering
• Research Assistant
• Teaching Assistant
• Administrator of computer systems of Department of Electrical and Electronics Engineering
• Administrator of Bilkent University Information Networks Laboratory
My research topics and some selected journal publications are as follows:
One of the key emerging areas in computer networks is Network Function Visualization (NFV). In traditional networks, many network services like firewall, domain name service (DNS) etc. are implemented in dedicated hardware appliances. NFV decouples such functions from proprietary hardware appliances so they can run in software. NFV is expected to reduce capital and operating costs, improve network flexibility, manageability and reduce the time to deploy new networking services. I am working on orchestration and management of virtual network functions and improving their reliability.
I was involved in Yuragi project of Osaka University, which proposes applying mathematical neural network models derived from biological systems to computer networks. Yuragi models the gene regulatory network of cells as an attractor selection based neural network for controlling VNT (virtual network topology) in optical networks. The attractor VNT candidates are stored in a Hopfield network, which is a kind of recurrent artificial neural network. I designed heuristic algorithms for designing attractor VNTs for Yuragi that are robust against large scale network failures.
I designed a hybrid optical network architecture, which uses both path (circuit) and packet switching. It carries the long TCP flows over path switching wavelengths in order to maximize average the throughput of the flows. It dynamically changes the ratio of path and packet switching wavelengths network-wide according to the traffic characteristics. I implemented a large scale optical packet and path(circuit) switching network simulator for this research. To the best of my knowledge it is currently the fastest and most scalable simulator for packet level simulation of TCP flows on optical networks in the literature. I have simulated mesh optical networks with 80Gbps link speed using around 10^10 TCP flows.
Using a M∕G∕c∕c Markov chain queuing model I proposed an analytical framework for calculating the path (circuit) blocking probabilities and reservation delays in path switching optical WDM networks. There are numerous analytical models in the literature, but to the best of my knowledge my analysis is the most accurate one in the literature.
In my Ph.D. thesis I challenged the famous rule of thumb, which states that network routers require a buffer size of Bandwidth*RTT in order to achieve high utilization with TCP flows. I modelled the packet drop rate of different buffer architectures with TCP traffic by analysis and simulation. I proposed novel network architectures, which greatly decrease the buffering requirements of both electronic and optical networks.
We studied the effects of the number of justifiers on TCP performance in an optical burst switching (OBS) network. Our simulations show that increasing the number of assemblers per destination reduces the negative effects of synchronization between TCP flows occurring as a result of burst losses. I designed and implemented an advanced optical burst network simulator for this project and released its source code on the Internet available here. Some independent survey papers selected my simulator as the most advanced OBS simulator among all related free and commercial simulators.
I proposed a QoS based network architecture for electronic networks using TCP load balancing traffic engineering methodology and a random early rerouting algorithm that controls the queuing delay difference between the two alternative paths. I showed that avoiding packet reordering by flow level splitting significantly improves the network performance.
My M.S. Thesis on this project won the Best M.S. Thesis Award of Electrical and Electronics Department of Bilkent University.
The source code of OBS (optical burst switching) extension that I implemented for ns-2 simulator is here. The new version is updated for ns2.35. It was presented and used in "nOBS: an ns2 based simulation tool for performance evaluation of TCP traffic in OBS networks" journal paper available here.
The ns-2 simulator module and simulation scripts that I implemented for the simulations in the paper "Combined use of prioritized AIMD and flow-based traffic splitting for robust TCP load balancing," are here. The paper is here.
The source code and executables of the encoder and the differential cryptanalysis key finder programs that I implemented for the MacGuffin block cipher algorithm are here. The encoder program creates 100000 random cyphertext plaintext pairs by encrypting with 8-block MacGuffin cipher algorithm with 128-bit input key and the findkey program calculates the secret subkeys of the encrypted ciphertexts by differential cryptanalysis.
Advanced Network Architecture Laboratory,
Graduate School of Information Science and Technology,
1-5 Yamadaoka, Suita,
Osaka 565-0871, Japan