Antonios Katsarakis

My motivation stemmed from enjoying research and engineering that involves software-hardware co-design, the cross-pollination of ideas across fields like databases, systems, networking, and computer architecture, and focusing on the intersection of theory and practice to build next-generation systems. The field of distributed datastores was particularly fascinating as it combined algorithms, distributed systems, databases, networking, and computer architecture -- subjects I enjoyed since my undergraduate studies at the University of Crete.

A major highlight was receiving the EuroSys Roger Needham PhD Award Honorable Mention in 2023, which recognizes an exceptional and innovative contribution to knowledge in the systems area. Other standout moments include the recognition of my research with prestigious accolades such as the PPoPP Best Paper nominee and the IEEE Micro Top Picks Honorable Mention. Additionally, the opportunity to undertake several research internships at Microsoft Research and collaborate extensively with leading academics globally significantly enriched the doctoral experience.

A primary conceptual challenge was tackling the complex intersection of performance, consistency, and fault tolerance in distributed data systems, where traditional protocols often performed poorly during common, fault-free operation. This was overcome by looking inside single-server architecture for inspiration, adapting invalidation-based protocols to distributed systems to achieve high throughput and low latency. I also encountered fundamental theoretical challenges, such as the impossibility result (The LAW theorem) regarding achieving local reads, strong consistency, and safety under asynchrony, which we circumvented by introducing the practical idea of almost-local reads. In non-technical aspects, I overcame the challenge of rejection by developing perseverance, decoupling my ideas from myself, and viewing rejection as an opportunity for improvement and calibration.

I am currently a Principal Researcher at Huawei Research, working at the forefront of designing and developing next-generation databases. My work focuses on crucial areas such as concurrency, consistency, fault tolerance, data structures, and transactional and analytical databases, which are vital for data processing and storage in the modern cloud and AI era. My PhD provided extensive preparation by tackling the intricate issue of data replication and designing innovative systems and algorithms to overcome challenges in performance, consistency, fault tolerance, and transactions in distributed environments. The doctoral training fostered crucial technical approaches like software-hardware co-design and the cross-pollination of ideas across systems, computer architecture, and networking, which are fundamental to my work today in optimizing performance and reducing cost.

The key mindset I rely on heavily today is the principle of "Shooting for the moon" (aiming for 10x improvement, not 2x). Such a radical approach encourages imaginative leaps and rebel ideas necessary for next-generation system design. This mindset is intrinsically linked to the skill of questioning conventional wisdom and seeking to expose and exploit falsely believed trade-offs, particularly by studying the assumptions behind impossibility results and aiming to circumvent them practically. That helps to drive forward research at the intersection of theory and practice.

I would advise aspiring doctoral researchers to aim for "the ideal" solution, defining extreme goals and working backward, and to "Shoot for the moon" by aiming for 10x improvements to encourage radical designs. They should actively question conventional wisdom and study impossibility results to intentionally circumvent their boundaries for practical solutions. It is also crucial to view rejection as a path to improvement, cultivating perseverance and separating personal feelings from written work. Finally, remember the fundamentals: keep your designs simple and lean (the 80-20% rule), and embrace the journey by prioritizing sleep, exercise, social connections, friends, networking, and, of course, plenty of traveling.

I encourage prospective students and the wider community to read the full story of my doctoral research journey, which is detailed in the ACM SIGOPS blog posts: "Thinking outside the box: My PhD Odyssey From Single-Server Architecture to Distributed Datastores". This journey demonstrates how ideas inspired by single-server architecture can illuminate solutions for modern distributed datastores. You can explore my work, including public and reproducible research through open-source contributions, on my personal website: https://antonis.io. Finally, for students in tech, do not let the complexity of the AI era inhibit you; instead, view it as an opportunity to address the immense challenges posed by "cataclysmic amounts of ever-growing data". Shoot for the moon (10x, not 2x!) by focusing on software-hardware co-design and questioning conventional wisdom to build the next generation of highly capable systems.

Thinking outside the box: My PhD Odyssey From Single-Server Architecture to Distributed Datastores (Part 1)

Thinking outside the box: My PhD Odyssey From Single-Server Architecture to Distributed Datastores (Part 2)