Why Cloud Computing Changes Database Design: Benefits, Drawbacks, and Core Principles

Cloud computing's fundamental principle is using multiple interchangeable machines, which directly impacts the functionality of cloud‑based database systems.

Traditional databases fall into parallel‑first (e.g., MongoDB, Teradata) or single‑storage‑first (e.g., PostgreSQL, MySQL) categories, each with inherent design limits that affect certain architectural decisions.

Examples: Greenplum supports sequences while Redshift does not; BigQuery lacks sequences whereas Teradata provides them.

Cloud databases belong to the same category, with new systems leaning toward parallel‑first. Cloud systems emphasize scalability and replaceability of machines.

In the single‑storage‑first group, cloud offerings focus on managing costs, upgrades, and reliability of traditional single‑node products, such as Heroku PostgreSQL, Amazon Aurora, Google Cloud SQL, and Azure SQL.

In the parallel‑first group, there are two sub‑categories: SQL/relational (BigQuery, Snowflake, Redshift, Spark, Azure Synapse) and DHT/NoSQL (BigTable, DynamoDB, Cassandra, Redis). The distinction relates to physical data layout and access patterns rather than SQL language support.

Parallel‑first relational clouds often rely on native cloud storage with a limited get/put API, which restricts random high‑performance access and thus limits advanced persistent data structures like indexes.

Consequently, cloud implementations using native storage favor sequential reads/writes over indexed access; indexes must be built outside the underlying storage, and even then the storage API may hinder address‑based operations.

Advantages of Cloud Computing

Cloud platforms manage the entire infrastructure stack, relieving users of deployment, reliability, and management concerns.

Free trials let users start experiments quickly and scale when needed, which is hard with on‑premises systems.

Standardized integration processes with third‑party SaaS let users focus on core business.

Resource utilization can be maximized; clouds expose usage controls, enable load smoothing, and prioritize interactive or critical workloads.

Large‑scale procurement and high utilization lower storage costs, making long‑term data retention, time‑travel queries, and backup more economical.

Cost accounting at the query level lets users make informed cost‑benefit decisions and sometimes choose to scale out rather than heavily optimize queries.

Weaknesses of Cloud Computing

Distributed cloud platforms have broader failure domains; a single component failure can cause temporary service disruption.

Many cloud databases are less mature than on‑premises products, lacking features that may never be added.

Parallel relational clouds can suffer performance penalties on INSERT/UPDATE/DELETE operations.

Additional latency between clouds and on‑premises systems may force users to lock into a specific provider and region.

Cost curves differ: scaling is easy but controlling spend is harder; hitting cost caps can cause job rejections, adding operational complexity.

Vendor lock‑in is common; migrating between cloud platforms is often more difficult than moving between on‑premises solutions.

Compiled by Tao Ran. Further reading: Which Cloud Database Should You Use?