Reading

Academic foundations

An annotated bibliography of the papers, works, and specifications that shaped the design of Aeron Cache.

Academic foundations

Aeron Cache didn’t appear from nothing. It stands on decades of work in consensus, concurrency, wire formats, and distributed-systems design. This is the reading that shaped how we think — real papers and works, correctly attributed, each with a note on why it matters to what we built. If you want to understand the why behind a design choice, start here.

Consensus & Replication

In Search of an Understandable Consensus Algorithm (Raft) — Diego Ongaro & John Ousterhout, 2014. Raft was designed to be comprehensible where Paxos is famously not, decomposing consensus into leader election, log replication, and safety. Aeron Cache’s clustered mode runs on Aeron Cluster, which implements RAFT, and the cache is a deterministic replicated state machine applying the committed log in order. This paper is the single most direct influence on how our high-availability story works — see RAFT consensus.

The Part-Time Parliament — Leslie Lamport, 1998. The original (and allegorical) presentation of the Paxos algorithm for reaching agreement among unreliable processors. It established the formal foundations of asynchronous consensus that essentially every later protocol, Raft included, defines itself against. Reading it is how you understand what Raft deliberately simplified.

Paxos Made Simple — Leslie Lamport, 2001. Lamport’s own plain-language retelling of Paxos, written because the parliament allegory obscured the algorithm for many readers. It crystallizes the core invariants of a single-decree consensus and their extension to a replicated log — the exact guarantees a replicated cache relies on to agree on state across nodes after a failure.

Wire Formats

Simple Binary Encoding (SBE) — FIX Trading Community standard; SBE is a wire-format standard born from the financial-messaging world’s need to encode and decode structured messages with minimal latency and no garbage. Its codecs read and write fields in place at known offsets, zero-copy and allocation-free. Aeron Cache’s native gateway transport uses SBE (schema id 7, little-endian) precisely for this property — decoding a message shouldn’t cost you an object graph. (A community standard with a reference implementation, not an academic paper.)

Distributed Systems Foundations

End-to-End Arguments in System Design — Jerome H. Saltzer, David P. Reed & David D. Clark, 1984. A foundational argument that a function is often best implemented at the endpoints of a system rather than in its lower layers, which can only ever do an incomplete job. It informs how we think about where correctness and completeness belong across the cache, its transports, and its clients — for instance, letting the client-side embedded object cache own the reconstruction of a value from streamed deltas.

Dynamo: Amazon’s Highly Available Key-value Store — Giuseppe DeCandia, Deniz Hastorun, Madan Jampani, Gunavardhan Kakulapati, Avinash Lakshman, Alex Pilchin, Swaminathan Sivasubramanian, Peter Vosshall & Werner Vogels, 2007. Dynamo showed how a key-value store could prioritize availability and predictable performance at scale, and it introduced a generation of engineers to consistent hashing, replication, and the practical tradeoffs of distributed state. It’s essential context for any key-value store, and a useful contrast: where Dynamo chose eventual consistency, Aeron Cache’s clustered mode chooses linearizable replication through RAFT.

Consistent Hashing and Random Trees: Distributed Caching Protocols for Relieving Hot Spots on the World Wide Web — David Karger, Eric Lehman, Tom Leighton, Rina Panigrahy, Matthew Levine & Daniel Lewin, 1997. The paper that introduced consistent hashing, the technique for distributing keys across a changing set of nodes while minimizing the keys that have to move when membership changes. It is foundational to how distributed caches partition and rebalance data, and essential background for reasoning about any clustered key-value system.

Harvest, Yield, and Scalable Tolerant Systems — Armando Fox & Eric Brewer, 1999. This work reframed availability-versus-consistency as a tunable tradeoff via the notions of harvest (how complete an answer is) and yield (the probability of completing a request), sharpening the thinking that led to CAP. It’s a clarifying lens for the choices Aeron Cache offers: a RAFT-clustered mode that favors consistency, and an ephemeral single-node mode for when you’d rather trade it away.

The Log: What Every Software Engineer Should Know About Real-Time Data’s Unifying Abstraction — Jay Kreps, 2013. An influential essay (not a peer-reviewed paper) arguing that the append-only, totally-ordered log is the unifying primitive behind replication, stream processing, and data integration. It articulates why a replicated log is such a powerful backbone — the same abstraction at the heart of Aeron Cluster’s replicated log, which drives our deterministic state machine.

Specifications

RFC 7386 — JSON Merge Patch — Paul Hoffman & James Snell (IETF), 2014. A compact specification for applying partial updates to a JSON document: present fields replace, nested objects merge recursively, and a null deletes a field. Aeron Cache implements these exact semantics for patchItem and for deep-merging PATCH_ITEM deltas in the embedded object cache, which is what makes patch-native streaming possible. (An IETF RFC, not an academic paper.) See JSON Merge Patch.

These works shaped how we reason about consensus, concurrency, and distribution.