Slide 1

Slide 1 text

AlloyDB Kanji Yomoda (@k-yomo) May 2022

Slide 2

Slide 2 text

Confidential & Proprietary 2021 AlloyDB

Slide 3

Slide 3 text

Confidential & Proprietary 2021 Key features ● Fully compatible with PostgreSQL ● 4x faster for transactional workloads, up to 100x faster for analytical queries than standard PostgreSQL ● 2x faster than RDS ● SLA of 99.99% (inclusive of maintenance) ● non-disruptive instance resizing and database maintenance (several seconds) ● Built-in integration with Vertex AI

Slide 4

Slide 4 text

Confidential & Proprietary 2021

Slide 5

Slide 5 text

Confidential & Proprietary 2021 Architecture transition ● Disaggregation of compute and storage ● => No private storage for replicas ● => Much less replication delay

Slide 6

Slide 6 text

Confidential & Proprietary 2021 Why higher availability and scalability? => persistence is horizontally partitioned into groups of blocks

Slide 7

Slide 7 text

Confidential & Proprietary 2021 Why analytical queries get faster up to 100x? => Make use of columnar format

Slide 8

Slide 8 text

Confidential & Proprietary 2021 Columnar format works well for OLAP OLAP example

Slide 9

Slide 9 text

Confidential & Proprietary 2021

Slide 10

Slide 10 text

Confidential & Proprietary 2021 Pricing https://cloud.google.com/alloydb/pricing

Slide 11

Slide 11 text

Confidential & Proprietary 2021 Pricing [Min Price] 4vCPU * 48.24 + 32 * 8.18 = 454$ (in preview)

Slide 12

Slide 12 text

Confidential & Proprietary 2021 Summary ● High performance ● High availability & scalability Pros Cons (for now) ● Cost ● Low flexibility for instance size ● Can’t be used as Bigquery external datasource

Slide 13

Slide 13 text

Confidential & Proprietary 2021 References ● AlloyDB for PostgreSQL under the hood: Intelligent, database-aware storage ● Introducing AlloyDB, a PostgreSQL-compatible cloud database service ● Introducing AlloyDB for PostgreSQL: Free yourself from expensive, legacy databases

Slide 14

Slide 14 text

Confidential & Proprietary 2021 Thanks!