Wide-column demo: Bitcoin keyspace point reads, on disk, no hop
You are looking at the Bitcoin address & tx keyspace64-byte keys over ~1KB records, durable on disk in one Semurg substrate. Ask it to fan point reads across every core, cold then warm then hot and it runs the
real query live on this box, on data you can
download and re-run yourself.
Click an example in the agent rail, or ask to READ, or click a point-read example in the agent rail. Every run is measured end to end, on
this box, on the real dataset. Nothing is pre-computed.
A Bitcoin address row is a 64-byte key over a roughly 1KB record of value-flow attributes. The workload is 100% point reads, and that read is a seqlock hit off the RAM-ring, fanned across every core, with the row durable on disk underneath. No partition coordinator, no replica round-trip. Cold, warm and hot, measured live.
the wide-column leader
A NETWORK HOP PER READ. A tuned wide-column node serves on the order of 50 to 100 thousand point reads a second, and every one pays a partition lookup plus a replica round-trip because storage and compute are separate tiers.
semurg, one engine
The compute runs on the data. A point read is a seqlock hit off the RAM-ring, no coordinator, no replica round-trip, no compaction tax, the row durable on disk the whole time.
Fast like a cache, durable like a database, and there is no second tier to cross.
Bitcoin address & tx keyspace, 64-byte keys over ~1KB records. Measured live on this box. Re-pull the chain over Bitcoin P2P and re-run it yourself.
ask the agent to READ, the live cold/warm/hot throughput renders here
All three bands are total round-trip time, no asterisk. We are not pitting our hot against anyone's cold: cold is the first-touch disk pull, warm is the steady-state point read, hot is the cache-resident ceiling. A wide-column node pays a partition lookup plus a replica round-trip on every point read because storage and compute are separate tiers. Here the compute runs on the data, so a read is just a seqlock hit, and every row is durable on disk with no compaction tax. Fast like a cache, durable like a database, in the one engine that also serves graph, full-text and vectors.