Qumulo is the industry's fastest, most cost-effective way to easily manage exabytes of data anywhere. Cloud Native As shown by the latest SPECstorage Solution 2020 AI_IMAGE benchmark results, Qumulo's Azure Native Qumulo (ANQ) achieved an Overall Response Time (ORT) of 0.84 milliseconds over a five-hour burst period, at a total cost to customers of just $400.
Latest AiThority.com News: Wipro launches on-premise GenAI solution in partnership with Hewlett Packard Enterprise
Cost-effective deployment AI Training Public cloud infrastructure requires data to be transferred from cheap, scalable object storage to limited and expensive file caches, which adds complexity and increases GPU idle time by up to 40% as data is staged from object storage to the local file cache.
ANQ acts as an intelligent data accelerator for object stores, performing parallelized pre-fetch reads that are served from Azure primitive infrastructure via the Qumulo filesystem directly to the GPUs running AI training models. This innovative architecture improves GPU-side performance by reducing load times between the object layer and the filesystem, fundamentally changing the way file-bound AI training in the cloud is architected.
Latest AiThority.com News: Veritone Announces Michael Keathley Joins Board of Directors
Benchmark-proven performance and cost-effectiveness
The SPECstorage Solution 2020 AI_IMAGE benchmark results, as mentioned in our blog, highlight the outstanding performance and cost-effectiveness of the ANQ. Key highlights include:
- We achieved an ORT of 0.84 ms for 700 jobs, which is the fastest benchmark of its kind run on Microsoft Azure infrastructure.
- With the SaaS PAYGO model, measurement stops when performance is not needed, and the list price to run the benchmark is around $400.
Three things set ANQ apart from other cloud-native file solutions that provide AI solutions to customers:
- True Elastic Scalability: ANQ allows customers to focus on business and technology concerns, not cloud-native storage infrastructure. Storage performance is AI Applications Stack demand and save costs when demand isn't. Unlike other cloud file systems, ANQ works without pre-provisioned volumes.
- Breakthrough Pricing: Qumulo brings the savings of cloud economics directly to you – our pricing model is based on actual storage usage (GB) and performance required (throughput and IOPs), with no pre-provisioned capacity required.
- Linear Performance Scaling: ANQ's architecture allows performance to scale linearly as the workload increases. With an average cache hit rate of over 95%, ANQ avoids load times between the object layer and the file system, enabling GPU-side scalability and faster performance.
read more: Blockchain for Good Alliance and Bybit Web3 Join SocialPlus Hackathon to Empower Builders
[To share your insights with us as part of editorial or sponsored content, please write to psen@martechseries.com ]
