Incident sheet · /infrastructure
- System
- Infrastructure & Cloud
- Filed
- Read
- 4 min · 750 words
- Severity
- Routine
Hitachi Vantara boosts VSP One Object small-object performance 5.5x
Hitachi Vantara announced a four-node VSP One Object start, up to 5.5x small-object performance and direct queries for Iceberg, Parquet, JSON and CSV.

By Rebecca Stone
Hitachi Vantara announced a smaller, four-node starting configuration for VSP One Object, with enhancements delivering up to 5.5x the performance for small objects. The announcement also covered improved multisite and regional replication (copying data across locations) and direct queries over Iceberg, Parquet, JSON and CSV data.
What did Hitachi Vantara announce?
Hitachi Vantara made the announcement last week. Its VSP One Object expansion covers three areas:
- A smaller, four-node starting configuration, with up to 5.5x performance for small objects.
- Improved multisite and regional replication capabilities.
- Support for querying open-format data directly on VSP One Object: Iceberg (table format), Parquet (columnar file format), JSON (structured text format) and CSV (delimited text format).
The announcements align with an industry-wide transformation across block, file and object storage. Hitachi Vantara links that transformation to growing use of agents in AI systems and the need to modernize storage infrastructure to maximize return on investment (ROI).
What does Omdia research show?
Storage integration remains a measured obstacle. Omdia, a division of Informa TechTarget, found that 71% of IT decision-makers agreed that “effectively integrating storage with data pipelines and AI workflows is a major challenge.”
Another 70% agreed that “storage-related challenges are a significant barrier to AI success.” Planned modernization is broad: Omdia found that 83% of organizations expect to upgrade their storage environments within 24 months to support growing AI demands.
Scott Sinclair is Omdia’s Practice Director covering the storage industry. Omdia disclosed that its analysts have business relationships with technology vendors.
Why does AI change storage requirements?
An end-to-end AI storage environment spans data preparation, model training, tuning (parameter adjustment) and inference (model prediction), which typically requires object, file and block storage rather than one storage type.
Object storage (a flat-address repository for unstructured data) supports massive, multi-petabyte repositories and high resilience. Hitachi Vantara says its use in data lakes (large analytical repositories) supports large volumes of text, video, image and audio.
Direct queries for open formats simplify dataset management within VSP One Object. Hitachi Vantara says the feature accelerates time to value (the period before an AI project produces measurable benefits) for larger projects.
Object storage also suits AI data preparation and provides cost-effective capacity for model checkpointing (saving model state during training). Traditional object storage, however, is better optimized for sequential, read-heavy environments than for the high-frequency, random read/write patterns associated with AI workloads.
Cold archives (low-frequency, long-term storage) no longer meet AI storage requirements on their own. Platforms need transactional performance (rapid handling of many storage requests) so larger data volumes do not delay project value.
The rise of agentic AI (AI workflows using agents) makes access more frequent and less predictable. Hitachi Vantara says this raises performance demands across block, file and object storage and should lead buyers to demand more from storage providers.
Can existing storage support AI?
Organizations can use a mix of existing and new storage early in an AI program. Hitachi Vantara says a legacy environment, however, will not fully optimize an organization’s AI investment. The optimal infrastructure depends on the specific AI use cases and approach.
Training an organization’s own model requires a different environment from tuning an off-the-shelf model or using RAG (retrieval-augmented generation, where a model receives information retrieved from a data store).
The first investment priority should be full utilization of GPU or accelerator resources (compute hardware used to accelerate AI workloads). Production performance and faster data preparation should follow. Prepared datasets must remain fit for the intended AI use case.
Hitachi Vantara argues that AI is not simply another enterprise workload. It combines use cases that raise performance, scale and governance demands while increasing complexity and reducing predictability relative to legacy applications.
Positive AI returns require continuous optimization of the entire data path, including GPU investments at scale. Storage must also add stronger data-management capabilities as part of that path.
How are vendors responding?
Hitachi Vantara said other storage offerings increasingly add data-management functions. Storage now manages data rather than merely storing and serving it.
The market response identified by Hitachi Vantara includes Dell Technologies, Everpure, HPE, Hitachi Vantara, NetApp, Vast Data and Weka, each evolving its portfolio for AI.
The result is a rapidly evolving form of storage differentiation that buyers must incorporate into storage decisions. Storage built only for legacy applications must evolve if enterprises expect to maximize returns from AI.
via techtarget.com (Source)
More from Rebecca Stone
Show full bio
Senior reporter covering money, funding and market shifts around enterprise information technology and the leaders who run at CIO Monitor.
208 articles
The newsroom digest
One email a day with the stories that matter. No noise.