Installing and configuring artificial intelligence (AI) in production can be a daunting task. Modzy makes it easy to deploy in a wide variety of situations.
What You Need to Know
Most API-based products are deployed in the cloud and are hosted by the company offering the product, meaning that customers needn’t be concerned with the complexities of deployment and maintenance. There are, however, some API-based products that support on-premise deployments where the installation, upgrading, monitoring, and maintenance is handled, at least in part, by the customer. In this scenario, it is highly advantageous for such a product to use existing infrastructure to avoid the need to set up different monitoring, auditing, and security stacks for each internal application.
Modzy fully supports both cloud and on-premise deployments, and every facet of Modzy’s architecture is designed for different deployment scenarios. We only require our customers to provide for a small number of dependencies to run on-premise:
- Kubernetes cluster– version and flavor are up to you as long as it complies with the Kubernetes specification at v1.12 or higher. You can add as many nodes, with whatever hardware (e.g. GPUs) you want
- relational database that supports JDBC
- S3-compatible object store– if you’re on-premise, you can use a tool like Minio to present standard network storage as an S3-compatible data store
- Kubernetes Ingress controller
Modzy is bundled as a package of container images and a Helm chart that will deploy the whole stack with minimal configuration. All the configuration for an on-premise installation is located in a single, documented file. Simply supply the database and object store credentials, the DNS name Modzy under which it will be hosted, and a few other tuning details, then use Helm to install.
Once Modzy is up and running, you’ll want to add monitoring and auditing to your installation. If you already have experience monitoring and auditing a Kubernetes cluster, you can use your preferred tools and applications. All of our services follow best practices for Kubernetes pods and log to STDOUT and STDERR so that tools like fluentd or logstash can ship them off to the logging platform of your choice.
What This Means for You
Modzy is easy to deploy anywhere. Whether you click a button in a cloud provider or rack some servers in an air-gapped datacenter, there are only a few dependencies required to run Modzy, and you’re free to use your existing infrastructure monitoring and security tools to ensure compliance with your organization’s policies.