Anton Borisov – Medium

Anton Borisov

Pinned

Published in

How We Accidentally Became One of UK’s First StarRocks Production Pioneers

We moved home-page analytics off Postgres to StarRocks, cutting p95 from ~20s to ~200 ms with minute-level freshness. Kafka→Flink powers…

Aug 18

Diagram of client → FE (Leader/Follower/Observer) → CN execution/caches → distributed storage.

Aug 18

Published in

Freeze the River’s Flow: Evolution of Streaming Computation Models

Streaming evolved from checkpoints to progress tracking. We perfected coordination, then discovered we didn’t need it all the time

4d ago

Freeze the River’s Flow: Evolution of Streaming Computation Models

4d ago

Published in

Iceberg CDC: Stream a Little Dream of Me

Iceberg v3 gave CDC identity and precision, v4 makes change detection light and metadata compact. Streaming icy edges finally start to…

Oct 27

Iceberg CDC: Stream a Little Dream of Me

Oct 27

Published in

The Good, The Bad and The AutoMQ

Slack’s KIP-1176, Aiven’s Diskless 2.0, and AutoMQ’s KIP-1183 and how Kafka’s storage frontier shifts from disks to the cloud.

Oct 16

The Good, The Bad and The AutoMQ

Oct 16

Published in

Switching me Softly

Zero-downtime PostgreSQL 12→17 upgrade at Fresha: RDS snapshots, logical replication, PgBouncer tricks & Debezium orchestration.

Sep 29

Switching me Softly

Sep 29

Published in

What the Fuss with Fluss: Flink Delta Force

Released July 31, 2025, Flink 2.1 dropped a bombshell on the streaming world: DeltaJoin and MultiJoin. These aren’t your typical join…

Sep 22

Picture presenting otters as a mascot for Apache Fluss

Sep 22

Published in

The SELECT FOR UPDATE Trap Everyone Falls Into

Using SELECT FOR UPDATE when you don’t need — kills concurrency. Postgres row-level locks, the right to choose, and how to spot the…

Sep 8

The SELECT FOR UPDATE Trap Everyone Falls Into

Sep 8

Published in

Iceberg MoR the Hard Way: StarRocks Code Dive

How I Learned to Stop Worrying and Love the Delete Files: A Journey into StarRocks’ Iceberg MoR Implementation

Sep 3

Diagram showing how StarRocks handles Iceberg merge-on-read deletes. Positional deletes use the FileScan node and IcebergDeleteBuilder to create bitmaps that filter rows during scan. Equality deletes use key rows in an anti-join processed by a hash joiner. Both paths merge with data files to produce the final output rows.

Sep 3

Published in

How Tables Grew a Brain: Iceberg, Hudi, Delta, Paimon, DuckLake

Not a feature bingo — an idea flow across Iceberg, Delta/Hudi, Paimon, and DuckLake with practical guidance on fit.

Aug 28

Timeline of table-format ideas: Iceberg (batch snapshots), Hudi (incremental COW/MOR), Delta Lake (transaction log & deletion vectors), Paimon (stream-native LSM), DuckLake (catalog-first metadata).

Aug 28

Published in

Divide And Partition: pg_partman online partitioning gotchas

Problem

Sep 23, 2024

Divide And Partition: pg_partman online partitioning gotchas

Sep 23, 2024

Anton Borisov

Anton Borisov

Engineer with 10+ years across data platforms and databases. Now building Fresha’s next-gen data stack with Flink, Spark, Paimon, Iceberg, and StarRocks.

Following

Help

Status

About

Careers

Press

Blog

Privacy

Rules

Terms

Text to speech