$\textbf{What Do You Want to Do Next?}$ / $\textbf{Things You Should Know}$

• My ideal next role is (most likely) as a Senior/Staff IC on a team working on databases, data infrastructure, observability, streaming systems, capacity planning, or cloud optimization. Industry (excl. “Defense”) is irrelevant, though I prefer working on technical products.

• A year ago, I wrote down some thoughts on software. I’d add the following:

  • Prioritize Simplicity. Well-designed simple systems outperform complex systems. Scalability and performance are downstream of Simple.
  • I work best on teams that develop clear mental models of their systems and are committed to building epistemic infrastructure. Our work cannot exist in a vacuum. Like science, software engineering is a social enterprise.

• I’ve done my best work when working closely with colleagues on solving technical problems that meaningfully affect our team, our colleagues, and our users (e.g. greatly reducing on-call burden, enhancing the user’s experience by dropping p99).

• I get a lot of energy from taking ownership of technical problems that don’t necessarily have clear answers from the outset. For example, I designed and developed DigitalOcean’s DBaaS Scrapable Metrics from a “20% project” to a fully-fledged feature.

• Last year I took a test called the CliftonStrengths. I scored highly on Responsibility, Achievement, and Talent Development and lower on Command and Adaptability.

$\textbf{What Have You Been Doing?}$ — Earlier this year (Late Feb.), I left my role as an IC at Grafana working on the time-series database Mimir. Since then I’ve been quite busy:

• I gave an agent access to a suite of tools that use the Golang profiling API. This agent is quite good at helping me grind CPU time, allocations, and wall time down. Really handy! (post, code).

• Developed a novel (?) algorithm for load-balancing. It’s lightweight and enables dynamic sharding without the need for a dedicated coordinator service. The algorithm shows moderately improved throughput (~10%) and p99 latency (~20%) under high load. (post).

• To benchmark the load-balancing algorithm, I (ahem, Claude) forked the benchmarking client wrk (repo) and added support for open-loop benchmarking. For most systems, open-loop benchmarks provide a more realistic (or more “adversarial”) traffic shape (diff).

• Developed a system for conducting “What-If” style analysis of distributed systems (e.g. “What if my DB’s p99 increases?”) by siphoning a system’s OTel traces. (post WiP).

• Wrote a procedural generator for distributed systems. This was in support of the (much larger) project described above. (post).

• Did some math, a nice suite of little problems (Inverted Index Optimization, others pending…). I’m also trying my damndest to read Bóna’s Combinatorics…

• Realized that Redis is single-threaded (after all these years!). Why has nobody (they actually have, btw) tried pinning multiple Redis instances to different cores of a multi-core machine? Well I tried, and squeezed $\geq20M$ rps out of a 8 vCPU machine. (post).

• After getting hooked on an app called Endel, enlisted Claude to learn some SuperCollider and make generative ambient music. (post).

• Did a little art project where I took a 96vCPU EC2 instance and used $\texttt{stress-ng}$ to “paint” the utilization graph. (post).

• Been reading about Science (e.g. Habermas, Longino, Hamming, and Bush) and Science Fiction (A Canticle for Leibowitz, A lot of Le Guin).