jotaBase storage lab

One write, through every place a document is kept. Change the persistence strategy and the same writes are costed against a different design.

Simulated. Runs the engine's own algorithms and key formats in this page — the revision hash (verified byte-for-byte against a running Worker), the write order, and the tail, segment and manifest shapes. Nothing is sent anywhere and no database is involved. Bytes and operation counts are exact; latency is not modelled.

Write a document to watch it land in every store Seals at 5 documents, so the tail fills and seals while you watch.
Browser

Client A — IndexedDB

store … · table docs

table meta


    


    

      
Worker
Stateless

      
room …

      

    


    

      
Durable Object
SyncRoom — in memory

      
seq 0 · tail 0 / 5

      

      
tail — a Map keyed by id, so a second write to one document
        replaces its entry rather than adding one

      

      
versions — one counter per document, what the revision is derived from

      

    


    

      
Durable Object
DO storage — survives eviction

      
local to the machine · 0ms even at 81k entries — but capped at 2 MB per key

      

      

    


    

      
Workers KV
SYNC_KV

      
…

      

      

      
…

      

    


    

      
Browser
Client B — IndexedDB

      
store … · table docs

      

      
table meta