The massive expansion of the Internet has connected hundreds of millions of people around the world. But it has also stepped up demands on network infrastructure, creating a major challenge that will only grow as more people get online.

Building on the work of their professor and advisor Dr. Michał Aibin, two groups of Khoury College master’s students at the Vancouver campus set out to help address that challenge. Their algorithm-based research projects — which began in the summer and concluded last month at the IEEE UEMCON 2020 conference — mark a key achievement for their academic careers and for the Khoury College graduate program. Their research is a valuable contribution to the growing literature on how algorithms can optimally facilitate data transmission over the Internet, connecting more people more efficiently.

“This is the first batch of students in Vancouver that did this type of project, and they did a great job,” Aibin says. “This proves that the regional campuses can work and can contribute to a body of research. On the first try, we had 100 percent success.”

The Quest for a Better Algorithm

The research journey began during Aibin’s algorithms course during the summer.

“Because they are doing a master’s degree, they need to have a research component,” Aibin explains. “They pick a research topic, or if they don’t have any preference, I help them select topics. They investigate and learn about the topic; they need to understand the problem to be solved, come up with an algorithm, run numerical simulations, and check the results.”

Aibin’s 10 students were split into two teams: Kishan Patel, Siwei Zhao, Sanyami Shah, Nickyta Patel, and Vachana Shetty on one — Bijal Patel, Haiyang Ji, Swati Nayak, Ting Ding, and Yue Pan on the other. Since the coronavirus pandemic had already shifted classes online, the collaborations happened through Zoom and written messages. Though the arrangement was difficult, there was a silver lining.

“If this had been an on-the-ground course, we wouldn’t have been able to work with students from the Boston and Seattle campuses,” says Kishan Patel, an MS student on the Vancouver campus. “That was a great part. We got to know a couple of those students . . . We used to meet once or twice weekly to collaborate and help each other.”

Aibin encouraged them to build on his own research in optical networking.

“When I select the topics, they are pretty broad,” he explains. “[The students] choose the theme, then they narrow it down to a specific problem. I don’t walk them by the hand. But I’ll definitely help them exclude directions that won’t work. They all presented a decent algorithm to start, but I said, ‘Let’s try to push for something unique, something that is yours and adds value.’”

And add they did.

Bijal Patel, another Vancouver MS student, says, “We started by understanding [Aibin’s] work, and we updated his algorithm.” She adds, “It was exciting and challenging to re-do his work, then optimize the algorithm.”

Both research papers investigated dynamic routing algorithms in elastic optical networks (EONs). Dynamic routing is a technique in which routers select paths in response to real-time changes in logical network layout; EONs employ flexible grid technology to increase channel capacity and use bandwidth more efficiently. Kishan Patel explains that while EONs are a great, flexible platform, they introduce new parameters and increase the complexity of the problem.

“You have a bunch of parameters, and you need to find the optimal path,” he says. “You need to handle trade-offs. Your algorithm might not work best for certain parameters. You need to prioritize. What do you want the most?”

Kishan Patel’s group decided to optimize Aibin’s AMRA algorithm, which efficiently handles trade-offs in routing decisions, by introducing a distance factor to improve the regenerator usage. Kishan wrote a distance metric that formed a major part of the new algorithm. Bijal Patel’s group developed new matrices for AMRA and the SPF algorithm, which always finds the shortest available path.

“The new matrix, which was actually giving us the proper output, was the original AMRA matrix with our new calculations added to it,” Bijal Patel says. “We added our calibrated occupied length matrix to it and had quite good results with our new algorithm. Our algorithm [ADRA] outperformed both SPF and AMRA, which was a great achievement for us.”

In both groups, the goal was the same: the most efficient possible use of network infrastructure, and therefore the best internet experience for the most people.

“In networks, the bandwidths are divided into slots,” Kishan Patel says. “Let’s say there is a request for 5 GB of data; the data rate is 5 GB per second, but your transmission line is 50 GB per second. If you assign the whole transmission line to that request, you are essentially wasting 90% of the resources. Having an optimal algorithm which can decide which requests should be assigned to which path is important. And that’s what we set out to do.”

Presentation and Recognition

With the research and the paper complete, the groups shifted their focus to the presentation. Aibin, wanting his students to be as prepared as possible, gathered the Vancouver Khoury faculty to watch a demo presentation one week before the virtual conference.

“We got a lot of positive and negative feedback, and had enough time to incorporate those things into our presentation,” Kishan Patel says. “They’ve published a lot of papers and submitted to a lot of conferences. So their feedback was quite helpful for us — the optimal way to present, how to create a script, which content to include, color schemes to make the presentation more appealing.”

Bijal Patel reports, “In one week we changed our presentation and scripts. The first time we took about 20 minutes, but we had to present it in 12 minutes. Their feedback helped us to make it better and more interesting.” She also noted that the demo helped them to avoid the text-heavy scripts and read-off-the-page style that some other presenters employed.

All their hard work and preparation paid off and then some. Not only were both papers accepted and published — titled “On Efficient Candidate Path Selection for Dynamic Routing in Elastic Optical Networks” and “Adaptive Modulation Regenerator and Distance Aware Algorithm for Dynamic Routing in Elastic Optical Networks” — but Kishan Patel was named the session’s best presenter.

Aibin remarked that the Zoom presentation was well-attended and that the group got to field some questions. However, he lamented that the students didn’t get to physically attend the conference — which is normally held in New York — and benefit from the interactions that in-person conferences encourage.

“There are a lot of companies like Microsoft, IBM, people who will notice you and might even invite you for an interview,” he explains. “During the presentations there is limited time, [but] when there are coffee breaks people talk with each other.”

Nonetheless, Aibin considers the opportunity to present at an IEEE research conference a worthwhile and beneficial experience for his students.

Both Kishan Patel and Bijal Patel noted that their algorithms were successful for unicast and anycast requests, which entail one-to-one and one-to-many transmission, respectively. But for multicast requests (many-to-many) that present more traffic flows and new, complex parameters, research will be a lengthy endeavor. It’s also the natural next step, making the modified algorithms even more applicable to real-world infrastructure.

Remarking on his students’ contribution, Aibin concludes, “They did an amazing job. Both of the teams produced good results, and their solutions can be used to improve the efficiency of existing systems.”