Student Spotlight: Anthony Byrne is advancing the security, energy efficiency, and cost transparency of cloud computing
The cloud computing industry has faced several significant data breaches, impacting millions of people. Furthermore, the industry currently consumes between 1% and 2% of the world’s energy supply, and consumption increases yearly.
Anthony Byrne, working with collaborators, is developing an innovative new cloud computing architecture, MicroFaaS, to improve security, energy efficiency, and cost transparency. Byrne is a researcher with the Boston University Performance and Energy Aware Computing Lab (PEACLab) and is advised by CISE Director, Professor Ayse K. Coskun. At the 2022 CISE Best Student Paper Competition, Byrne received the 2nd place award for his paper on this new architecture.
Byrne’s MicroFaaS architecture is built on the serverless cloud computing paradigm. Current cloud computing data centers resemble warehouses filled with hundreds of racks housing dozens of rack servers. The x86 processors inside these servers were built for computers that remain plugged in, allowing them to operate with little regard for energy efficiency.
“We’re excited about our proposal, which uniquely targets the serverless cloud computing paradigm. MicroFaaS demonstrates a 5.6x increase in energy efficiency over conventional, broadly deployed cloud systems,” says Byrne. “Moreover, financial projections show that with MicroFaaS, cloud providers can realize 33% cost reduction over a five-year lifespan compared to traditional cloud data centers performing the same tasks.”
MicroFaaS differs from conventional cloud computing methods by relying on ARM-based single-board computers instead of x86-based rack servers. ARM chips were designed for portable devices like cell phones, which have to be aware of energy efficiency because of their limited battery life. “This switch automatically cuts the cost of cloud computing because ARM chips are significantly more efficient than x86-based processors,” adds Byrne.
MicroFaaS’ technology brings the benefit of energy proportionality, allowing it to be a cost-transparent option for providers. This means the data center’s energy will be proportional to the computational power needed: for 5 units of computational work, only 5 units of power will be used. Similarly, MicroFaaS has a 1:1 relationship between the number of nodes in the system and how much work can be done simultaneously. As a result, cloud providers know exactly how much hardware to purchase.
Traditional x86 processors don’t have the same linear relationship between the amount of work performed and the power consumed. Additionally, x86 processors run constantly so that they are always ready for incoming workloads or user requests. MicroFaaS is unique because its worker nodes take only 1.5 seconds to start up, so when they’re not in use, they shut off. These features allow MicroFaaS to be a more energy-efficient option than conventional cloud computing architectures.
Traditional data centers are also unable to estimate the amount of hardware required to ensure a smooth user experience, often causing them to overcompensate and purchase more hardware than necessary. Both of these factors lead to a lot of wasted power. Alternatively, MicroFaaS’ novel cloud computing architecture decreases energy consumption and waste, while offering providers a more cost-efficient option.
The multi-tenant nature of conventional cloud computing methods present significant security concerns. Public clouds, like Amazon and Google, have many different applications sharing one machine. The industry currently relies on software-based isolation to prevent one application from breaching the data of others, but this has proven less than ideal. MicroFaaS uses hardware-based isolation, a more secure option in the face of cyber attacks. This method isolates workloads from each other and drastically reduces security vulnerabilities.
In addition to developing MicroFaaS, Byrne is also working as a Site Reliability Engineer (SRE) at Red Hat, where he works on monitoring and improving the reliability of cloud container orchestration platforms. “Just like today’s software-isolation-based cloud platforms, hardware-isolated cloud technologies like MicroFaaS will need SREs equipped with advanced tools in order to keep things running smoothly in the face of ever-increasing user demand,” says Byrne. “I’m very grateful that my position here at Red Hat gives me a front-row seat to the development of cutting-edge monitoring and diagnostic tools for today’s cloud platforms, and I’m excited to see how the lessons we learn from these innovations can be applied to next-generation platforms like MicroFaaS.” Byrne adds that none of this work would have been possible without his former undergraduate collaborators and co-authors, Yanni Pang and Allen Zou, both of whom graduated this past May.
Anthony Bryne received his B.S. and M.S. in Computer Engineering from Boston University in 2019 and 2022, respectively. While earning his B.S., he completed internships with IBM and Red Hat, where he developed and patented methods of improving and tracking cloud application security and performance. Since joining the ECE Department’s graduate program, his research interests have included increasing energy efficiency in next-generation cloud data centers, as well as using machine learning to detect undesirable components compiled into opaque software binaries. In his free time, he enjoys completing small electronics projects and making mead.