Organizations often face the challenge of balancing the growth of their data sets with cost management. This balance becomes particularly difficult in the realm of object storage, where data access patterns can be unpredictable and missteps in data placement might lead to unexpected cost increases. Instances have been reported where moving data prematurely to a colder storage class, intended to reduce costs, resulted in a surge in operational charges.
Addressing these challenges, a new solution named Autoclass has been introduced, which extends its functionality to existing Cloud Storage buckets. Initially available only for new buckets, Autoclass now offers a managed service for existing data storage, automating the data lifecycle management to optimize costs. This service operates at the bucket level, managing object lifecycles based on their last access time, thus reducing the operational burden of managing data placement rules.
A key benefit of Autoclass is its price-predictability feature. It eliminates surcharges typically associated with colder storage classes, including retrieval, early deletion, and class transition charges. Autoclass provides two configurations for users:
- Default: This configuration transitions objects between Standard and Nearline storage classes.
- Opt-in: This option enables transitions across all four storage classes (Standard, Nearline, Coldline, and Archive), based on object access times.
Each configuration has its implications. While all storage classes offer low latency access, colder classes like Coldline and Archive might have relatively higher read latencies and lower availability SLAs compared to Standard and Nearline classes. The default configuration is thus more suited for latency-sensitive workloads, while the opt-in configuration maximizes savings by including transitions to the lower-priced Coldline and Archive classes. Users also have the flexibility to switch between these configurations as their workload requirements change.
A case study highlighting the effectiveness of Autoclass comes from Konstantin Lalafayan, VP of Infrastructure at Picsart. He notes the critical importance of optimizing storage solutions in the context of Generative AI advancements and daily generation of over two million objects. With Autoclass, Picsart observed a 40% reduction in storage costs within 90 days, expecting further decreases upon reaching the 365-day mark, attributable to transitions to Archive storage.