Products overview¶
Compute and storage building blocks for running GPU and CPU workloads on Verda — from single instances to multi-node clusters, serverless containers, and hosted inference APIs.
Compute¶
Verda offers several ways to run compute workloads, from dedicated instances to fully managed serverless containers.
-
Instances
Dedicated CPU and GPU instances for development, training, and hosted workloads.
-
Private Networks
Isolated private IPv4 networks so your instances communicate over private addresses, off the public internet.
-
Clusters
Multi-node GPU clusters with Slurm or Kubernetes for distributed training and HPC.
-
Serverless Containers
Deploy containerized inference workloads with autoscaling, health checks, and batch jobs.
-
Inference
Call open models through an OpenAI-compatible API, billed per token - no GPUs to deploy or manage.
Storage¶
Persistent and shared storage options for data used by instances, clusters, and containers.
-
Block Volumes
Persistent disk volumes for instance and cluster storage.
-
Shared Filesystem
A shared filesystem that multiple instances or cluster nodes can access simultaneously.
-
Container Registry
A private registry for storing and deploying container images.
-
Object Storage
S3-compatible storage for datasets, model weights, and backups.