Principal Engineer and Independent Researcher specializing in distributed systems. My work centers on two strengths.
I work with consensus algorithms such as Raft end to end, from reading the papers to writing my own implementations in Go. I have designed and built a Raft-based Redis-compatible KVS and an Elasticsearch-compatible distributed search engine that combines Lucene with Raft. I study architectures such as Spanner’s TrueTime, TiDB’s Timestamp Oracle, and the “Log is the Database” and compute-storage separation models behind Aurora and Aurora DSQL, organizing them as trade-offs between scalability and consistency, and I present on them regularly at DB Tech Showcase and cloud-native conferences. Because this knowledge is not tied to a single product, it translates directly into data store selection and design decisions in domains where inconsistency is unacceptable, such as payments and reservations.
I have consistently worked on delivery systems with demanding latency, throughput, and spike-tolerance requirements: AdTech (SSP/DSP), notification infrastructure for smartphone carriers serving tens of millions of users, and large-scale delivery infrastructure for a CX platform. I care about systems that stay cheap as well as fast. That focus produced a published proposal for a low-latency distributed KVS that cuts inter-node synchronization cost using a client-centric consistency model, a proposal for improving DNS cache hit rate and response speed (IPSJ Student Encouragement Award), and substantial delivery-infrastructure cost reduction through redefined SLOs and a framework for running preemptible instances safely. I look for room to reduce both latency and cost at each layer, from caching to name resolution to data placement. On my own time I also design and operate resize.gg, an image delivery CDN for VRChat.
I am currently Principal Engineer at hacomono, leading partial microservice decomposition and the design and implementation of a loosely coupled, schema-driven event bus for its multi-product strategy. I build robust and scalable system infrastructure by continuously cycling between theory (academic research and self-implementation) and practice (large-scale commercial development).
A collection of products created by bootjp.
hashicorp/raft, implementing the datastore, Snapshot, and StateMachine (Apply/Snapshot/Restore).RESP with redcon and supports GET/SET/DEL.MOVED redirection to the leader, validated in a 3-node setup.NewSQL: Scalable Architecture with Storage Disaggregation and Distributed Consensus
The Mechanism Behind NewSQL and Distributed Databases: Raft
Designing an Event Bus for Loosely Coupled Schema-Driven Development
What is Raft? Understanding its Strengths and Weaknesses from the Mechanism
Why Did Spanner Need an Atomic Clock?
Reading the Paper on AWS’s Durable Redis-Compatible KVS: MemoryDB
Building a Low-Latency Distributed KVS for Personalized Content Delivery (VRChat Ver)
Understanding Challenges and Insights Moving from Aurora Serverless to v2 from the Paper
Reading the Paper on AdaptSize Supporting Akamai’s Cache Efficiency
Leading platform modernization to support a multi-product strategy for a vertical SaaS in the wellness industry.
Backend development for distributed search engines and metaverse infrastructure.
Responsible for developing the large-scale delivery and data processing infrastructure supporting the CX platform, where balancing low latency against low cost was the central theme.
Responsible for development and operation of search infrastructure, large-scale data delivery, and notification management for smartphone carriers, designing and operating delivery infrastructure under simultaneous low-latency and high-throughput requirements.
Development of AdTech areas such as SSP/DSP and construction of script-tag based access analysis infrastructure. Learned the fundamentals of low-latency delivery in ad serving, where bid responses face hard time budgets.
Started a rental server business during junior high school, consistently handling service launch and operation.
Operating a study community specialized in distributed systems on VRChat.
Theoretical research and implementation of distributed systems, distributed storage, and distributed KVS, and paper writing.
Product development and tool creation based on technical interests.