A cloud native platform offering comprehensive control and automation of all HPC tasks via CLI and a user-friendly GUI for designing, editing, and executing HPC jobs. Human-readable, portable workflow files that execute anywhere. Streamlined R&D for HPC & AI Research.
CIQ Fuzzball is a container-first performance intensive computing platform that accelerates innovation by reducing the burdens of workflow development, infrastructure provisioning and management of clustered jobs, even across hybrid environments.
Fuzzball modernizes traditional HPC with an API-first, container-optimized architecture. Operating on Kubernetes, it provides all the security, performance, stability, and convenience found in modern software and infrastructure.
Fuzzball takes an engineer and researcher-first approach and provides intuitive tools to build and deploy without expertise in infrastructure and lowers the barrier for all performance intensive computing. Fuzzball not only abstracts the infrastructure layer but also automates the orchestration of complex workflows, driving greater efficiency and collaboration. Plug it into your existing orchestration and automation, build your own SDKs, or simply point-and-click.
You define compute resource pools and the Fuzzball control plane analyzes the data, compute and storage requirements to automate the provisioning and orchestration of the right resources. Fuzzball makes it simple to embed and run applications with Workflow templates that can be built and designed for any workflow or application.
Fuzzball will deploy and optimize placement of your workflows across disparate clusters that reside in multiple regions or even across on-premise and cloud providers. This optimizes your environment based on data, compute and storage requirements and gives you the flexibility to develop in the cloud and deploy on premise or develop locally and deploy to the cloud for scale. And it comes with integrations for the tools you already use and templates so anyone can enjoy it, including Jupyter, Matlab, OpenRadioss, PyTorch, R Studio, TensorFlow, and it works with your CI/CD pipelines including Jenkins, GitLab, and GitHub Actions.
If you are planning on using CIQ Fuzzball, we strongly encourage you to review the Fuzzball documentation attached to this listing prior to purchase, and contact CIQ if you have any questions before clicking deploy.
Highlights
A user-friendly GUI for designing, editing, and executing HPC jobs. Comprehensive control and automation of all HPC tasks via CLI. Human-readable, portable workflow files that execute anywhere
Automated data ingress and egress with full compliance logs. Native integration with GPU, both on-prem and cloud storage, and your favorite applications and CI/CD pipelines.
Extensive is support included for installation and configuration, bug and security fixes, platform updates and upgrades, troubleshooting and diagnostics, integration, best practices, and general Fuzzball support for usage. See Support information for details.
AWS Marketplace now accepts line of credit payments through the PNC Vendor Finance program. This program is available to select AWS customers in the US, excluding NV, NC, ND, TN, & VT.
Fuzzball bills by usage, charging per minute across three separate dimensions. You pay for Fuzzball Orchestrate by the minute to run the control layer that submits and manages workflows. You also pay per minute for each Fuzzball Substrate Node, the standard compute node running your jobs. For GPU-based work, the Fuzzball Substrate GPU Node carries its own per-minute rate. These dimensions are independent and add together. Your total depends on how long the orchestrator runs and how many standard and GPU nodes you use, plus their runtime.
Top-of-mind questions for buyers
What counts as one Fuzzball Substrate Node versus a Fuzzball Substrate GPU Node for billing?
A Fuzzball Substrate Node is a standard compute node running your jobs, billed per minute. A Fuzzball Substrate GPU Node is a node with GPU hardware enabled, billed at its own per-minute rate. You pay the GPU rate only for nodes running GPU-based training or inference work.
Am I charged for nodes when jobs are not running?
Charges accrue per minute while the Orchestrate control layer runs and while Substrate Nodes or GPU Nodes are active. When nodes are not running jobs, their per-minute metering does not accrue. Your bill reflects actual runtime across each dimension.
Which dimension usually drives the largest part of my bill?
All three dimensions bill independently and add together on one invoice. Fuzzball Orchestrate meters the time the control layer runs. Node charges scale with how many standard and GPU nodes you run and for how long. GPU-heavy workloads shift more cost toward the Fuzzball Substrate GPU Node dimension.
ciq.com
Helpful?
Vendor refund policy
CIQ does not offer refunds at this time.
How can we make this page better?
Tell us how we can improve this page, or report an issue with this product.
Give us feedbackReport a problem with this product or seller
Legal
Vendor terms and conditions
Upon subscribing to this product, you must acknowledge and agree to the terms and conditions outlined in the vendor's End User License Agreement (EULA).
Content disclaimer
Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.
Helm charts are Kubernetes YAML manifests combined into a single package that can be installed on Kubernetes clusters. The containerized application is deployed on a cluster by running a single Helm install command to install the seller-provided Helm chart.
Version release notes
Changed
[fuzzball-substrate-extension/fuzz-8033] Services now publish declared ports on randomly assigned host ports so same-port services can share a node; in-cluster discovery via DNS SRV records and dynamic-config ip:port arrays. Clients must not assume node:declared-port reachability, and node orchestrate extensions must be upgraded before or with orchestrate.
[fuzzball-scheduler/fuzz-8046] Provisioner definition policies now apply to internal jobs, with fallback to unrestricted placement when no definition's policy matches.
Fixed
[fuzzball-scheduler/fuzz-8032] Fixed all scheduling for a definition stalling when its nodes are fully held by services; bounded-walltime work now backfills while the blocked head waits.
[fuzzball-scheduler/fuzz-8044] Fixed uncordoned nodes never returning to the scheduler until substrate restart.
[fuzzball-scheduler/fuzz-7982] Fixed preemption fit math under-counting task-array victims running multiple ranks on one node.
[fuzzball-scheduler/fuzz-7983] Fixed preempted task arrays re-dispatching out-of-range task ids and re-placing ahead of the allocation that evicted them.
[fuzzball-scheduler/fuzz-8034] A failed autoscaled service replica above replicas.min is now returned to the pending pool for a later scale-up retry instead of erroring the whole workflow and tearing down healthy replicas.
[fuzzball-ui/fuzz-8051] Fixed non-admin users being unable to create volumes from provisioners in the web UI.
[fuzzball-scheduler/fuzz-8031] Autoscaled service replicas are now checked against provisioner pool capacity -- a submission whose replicas.min cannot fit the pool is rejected, replicas.max above capacity is accepted with a warning and effective cap, and scale-up is skipped with an event when the pool is at capacity instead of leaving replicas pending forever.
[fuzzball-cli/fuzz-8049] Fixed AWS deploy and update failing when --version is passed with a leading "v" prefix.
Application-specific assistance to run workloads on Fuzzball and professional services to create and manage workflows are available from CIQ at an additional cost. Contact us for details.
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.
Be the first to review this product. We've partnered with PeerSpot to gather customer feedback. You can share your experience by writing or recording a review, or scheduling a call with a PeerSpot analyst.