Machina Rocks Updates AI OS 'Runway' ... Strengthens Performance, Resource, and Cost Management

Machina Rocks announced that it has updated its AI operating system, Runway, to version 2.4.0. The company said the overhaul strengthens features for integrated management of AI performance, resources, and costs.

Runway is a platform that connects data, models, and on-site systems. The platform manages the full lifecycle of AI development, deployment, and retraining, and its defining feature is that it provides the same operating framework even in high-security environments, including closed networks.

The latest update focused on bolstering observability and management functions so users can directly monitor and control changes in AI performance, resource usage, and costs during operations. As a result, it now includes a unified dashboard for token usage, input data drift detection, inference request and response logging, and inference performance monitoring.

It also adds workload Pod-level resource monitoring, resource usage statistics by layer, and storage browser upload and download features.

The expansion centers on usage management, model quality observability, and infrastructure inspection. First, it provides a unified token usage dashboard so organizations can view token consumption by organization, project, and model on a single screen. It also makes it possible to aggregate usage across different deployment environments, such as on-premises and cloud, so it can be used for cost management by department and project.

Model quality observability has also been expanded. It provides input data drift detection, allowing users to compare data distributions from real-world operations with baseline data from training and check whether data has changed. It also sends alerts when configured anomaly conditions are met, and helps identify the possibility of performance degradation based on changes in input data even before ground-truth data is secured.

It also added logging for inference requests and responses, as well as a feature that tracks model accuracy based on ground-truth data. This is intended to help users monitor the status of models in operation more continuously.

Infrastructure management has been refined down to the Pod level. Users can now check GPU, CPU, memory, and storage usage, as well as restart counts, for individual workloads, and use that information to inspect resource waste and signs of failure.

Machina Rocks said the update provides resource usage statistics by layer, including workspace, project, creator, and workload. Operators can use this to check the status of resource allocation. It also provides direct file upload and download features in a web environment.

Machina Rocks described the update as part of its efforts to strengthen Enterprise AI Sovereignty. Enterprise AI Sovereignty means independence from external platforms and direct visibility into and control over AI operating status. Sim Sang-woo, CTO at Machina Rocks, said the update allows users to determine whether performance has deteriorated after AI adoption and whether resources and costs are being used efficiently. He added that securing control over performance, resources, and costs is key to realizing Enterprise AI Sovereignty.

Runway strengthened its open architecture in its 2.0 version earlier this year. In the same version, security policies were also strengthened. Access-based governance was also reinforced.

Since then, Runway has expanded its use cases, mainly in high-security and highly regulated environments such as manufacturing, defense, finance, and the public sector.

Source: TECHWORLD · Kim Seung-ki
Original: https://www.epnc.co.kr/news/articleView.html?idxno=407588

References

• https://www.epnc.co.kr/news/articleView.html?idxno=407588