A Cloud for Machine Learning
The Bigstep Metal Cloud was originally designed for big data purposes, thus it excels at high throughput and 100% CPU utilization types of jobs.
High CPU, network and storage throughput (not latency) is key for typical Hadoop & Spark workloads, and the Bigstep metal cloud is perfectly optimized for these use cases.
Metal Cloud nodes feature 4x10Gbps links per node which can be bonded together for more throughput. Our high capacity L2 Networking with cut-through switching and no virtual switching or overlay networking enables high throughput, thus improving the cluster's overall performance.
Hadoop & Spark based deployments are very storage intensive. We provide nodes with 12 and 24 locally attached self-encrypting SAS drives that provide all the throughput needed for a balanced big data deployment.
Bigstep provides bare metal nodes with CPU cores, network throughput and RAM size and performance characteristics that achieve the best cost-to-performance ratio available with current technology.
Realtime Analytics Pipelines on Bare Metal
Bigstep's Container Platform enables new, Kubernetes based architectures on bare metal for high adaptability, low latency and high throughput for data science and production workloads on the same environment.