Google has introduced a new feature in its NoSQL database service, Bigtable, that could change the way enterprises manage data storage costs. This new capability is a fully managed tiered storage system that automatically moves less frequently accessed data from high-speed SSDs to cheaper, infrequent access storage. The goal is to lower costs while still keeping the data accessible for applications and queries.

This kind of storage tiering isn’t new outside of databases, but integrating it directly into Bigtable is a game-changer. It means companies won’t need to juggle multiple systems or deal with delays when accessing cold data. Instead, both hot and cold data are available within the same database environment. This simplifies data management and can lead to significant cost savings.

The Cost and Complexity of Traditional Storage Tiers

Many cloud providers, like Google, Amazon, and Microsoft, offer different storage tiers—hot, cold, and archive—that can be used alongside their database services. Usually, enterprises move infrequently accessed data to cheaper storage options to save money. But this comes with drawbacks. Accessing cold data often requires switching systems, which adds latency and complexity.

Typically, these integrations mean the database offloads cold data to external storage systems. Managing these separate systems and data movement pipelines can be a headache. Plus, retrieving data from cold or archive tiers incurs additional costs. For example, Google charges $0.02 per GB for cold storage access and $0.05 per GB for archive data retrieval, on top of the storage and network fees.

The Impact on AI and Data-Intensive Tasks

This new storage tiering feature is particularly useful for AI workloads, especially those involving agentic AI, which generates huge amounts of data. As data volumes grow rapidly, companies often struggle with managing costs. They tend to focus only on frequently accessed data, like updated vector embeddings, to keep expenses manageable.

With automated tiered storage, enterprises can now explore storing more complex data types—vector indexes, context logs, and other AI-related data—without incurring the high costs of always-on SSD storage. This opens up new possibilities for scaling AI applications efficiently.

Stephanie Walter from HyperFRAME Research highlighted that this feature offers a practical way for companies to handle large, vector-heavy workloads. Instead of paying premium prices for all their storage needs, they can rely on tiering to keep costs in check. Google already offers automated storage tiering in its distributed database, Spanner, which was introduced earlier this year.

In sum, Google’s move to embed tiered storage into Bigtable could make managing large datasets more affordable and less complex for companies working with AI and other data-intensive tasks. This development aligns with the broader trend of making cloud storage smarter, more integrated, and cost-effective.

Inspired by

• https://www.infoworld.com/article/4081636/google-adds-tiered-storage-to-nosql-bigtable-to-reduce-complexity-costs.html

Sources

• gmpg.org
• futurumgroup.com
• google.com
• google.com
• google.com