Leaderless by design
Every peer is equal. Send a request to any node and it becomes the coordinator—no elections, terms, or dedicated write leader.
INTERACTIVE PORTFOLIO · 2026
SELECT A PRESENTATION
Two stories. One throughline: reliable systems built for the real world.
DISTRIBUTED SYSTEMS, TRIMMED TO THE ESSENTIALS
A compact, leaderless JSON key-value store that keeps three Rust nodes in lockstep—through crashes, partitions, and rough water.
01 / THE IDEA
SailboatKV narrows the problem on purpose: one inspectable protocol, one fixed crew, and one clear consistency promise.
Every peer is equal. Send a request to any node and it becomes the coordinator—no elections, terms, or dedicated write leader.
Three deliberate operations: GET, PUT, and DELETE. Values are validated JSON objects; tombstones keep deletes ordered.
A minority returns 503 instead of guessing. Successful reads and writes always intersect a majority.
02 / THE PROTOCOL
A write is not successful because one node saw it. It is successful only after durable majority acknowledgement.
03 / CONSISTENCY
Every node converges on the same totally ordered version.
A completed read can’t return a value older than an earlier completed write.
Older or equal versions are never allowed to overwrite newer records.
Tombstones travel through the same quorum path, so deletion never moves backward.
04 / FAILURE MODE
SailboatKV chooses consistency over minority availability. Two nodes keep sailing; one isolated node stops safely.
MAJORITY PARTITION
ISOLATED MINORITY
“No quorum means no stale answer.
That’s the contract—not an outage bug.
05 / DURABILITY + RECOVERY
Acknowledged state crosses a deliberate persistence boundary. Stale peers heal when they return—without replaying a log.
06 / MEASURED PERFORMANCE
Fresh three-node Docker clusters · 10,000 measured linearizable reads · median of five runs on Apple Silicon.
operations / second
20.1% lower in this workload
≈12.6× lower per container
! A deliberately narrow native HTTP vs. etcd HTTP/JSON gateway comparison—not a claim of general superiority.
07 / GET UNDERWAY
Docker, Cargo, or Nix. Every package starts the same fixed three-node cluster with the same observable guarantees.
$ docker compose -f docker/docker-compose.yml \
up --build --detach --wait
✓ Network sailboat created
✓ Container node-1 healthy
✓ Container node-2 healthy
✓ Container node-3 healthy
QUORUM ONLINE · 3 ACTIVE NODES
$ curl localhost:3003/boats
{ "name": "Alcyone", "length": 32 }
SYSTEMS SOFTWARE · BUILT FOR THE FIELD
JACK PALA / SYSTEMS ENGINEER
Fifteen years across Linux, networks, security, automation, physical infrastructure, and software—focused on making the entire system work.
01 / OPERATING RANGE
The throughline isn’t a tool. It’s ownership across boundaries—from the cable and kernel to the service and deployment.
Automation, backend services, APIs, testing, deployment tooling, and deliberate low-dependency engineering.
Production administration, reproducible systems, storage, virtualization, recovery, and lifecycle operations.
Software that reads, controls, and responds to real equipment instead of assuming a clean cloud boundary.
SDN, segmentation, remote access, identity, least privilege, and production troubleshooting.
02 / PERSONAL INFRASTRUCTURE
My homelab is where I explore systems because I genuinely enjoy the craft—not because a ticket asked me to. It is a working hyperconverged environment for learning, building, breaking, and recovering infrastructure end to end.
LAB PHILOSOPHY“I learn systems by owning their interactions.”
Virtualization, clustered workloads, services, and lifecycle experiments.
Compute and resilient storage designed as one working environment.
Different system models, networking paths, and operational tradeoffs.
03 / LIQUIDSTAR
Solar-powered edge micro-datacenters operating across unreliable WANs, constrained power budgets, and remote international sites.
04 / CYBER-PHYSICAL SYSTEMS
At Liquidstar, control logic connects compute, networking, telemetry, and renewable-energy infrastructure. The design principle is local autonomy: safe behavior continues when the uplink disappears.
“Control remains local. Connectivity is useful—not required.”
05 / REPRODUCIBLE FLEET OPERATIONS
NixOS turns remote machines into reviewable, reproducible system definitions—with a cleaner path to deploy, roll back, and recover.
Services, networking, access, telemetry, and recovery live together.
The machine is produced from a declared configuration—not tribal knowledge.
Repeatable changes across similar hardware operating thousands of miles apart.
Rollback and rebuild reduce the cost of failure at inaccessible sites.
06 / OPERATE WHAT YOU SHIP
Reliability is more than uptime: observe the system, constrain access, understand the network, and make failure diagnosable.
07 / SOFTWARE ENGINEERING DEPTH
SailboatKV is deliberate practice in the formal failure modes behind the production systems I’ve operated for years.
08 / ROLE ALIGNMENT
A systems-software generalist for autonomous platforms: comfortable where application code, Linux, networks, hardware, and field operations meet.
Rust and Python engineering backed by deep Linux, backend, automation, and deployment experience.
Production integration with PLCs, MQTT, BACnet, Victron, metering, contactors, and network equipment.
Remote edge recovery, power-aware control, monitoring stacks, secure access, and distributed-systems testing.
Work across physical infrastructure, software, electrical controls, customer constraints, and international deployments.