Getting Started

Installation

Fennel has a client/server model. A Python client is used to define the features, which are then synced with the server. The server is responsible for computing, storing, and serving the actual features.

Installing Fennel's Python Client

The client can simply be pip-installed like this:

1pip install fennel-ai

bash

The client itself ships with an in-memory mock server which has near feature parity with the real server. All the code samples in the documentation can be run just via the mock server. See quickstart for an example.

In other words, once you have pip installed the client, you are all setup to explore Fennel locally without needing a Fennel server.

Deploying Fennel Server

Fennel runs as a single tenant system inside your cloud so that your data and code never leave your cloud perimeter. See deployment model for details.

With this deployment model, here are the steps for deploying Fennel servers in your cloud:

  1. Create a new account in your AWS Organization and note down the account ID. Fennel will run inside this account.
  2. Create a VPC in this account with internet and nat gateways configured so that nodes have internet access. The VPC subnets also need to be appropriately tagged for load-balancer creation. You can do so by following this guide or Fennel support can also provide standard cloud-formation templates to use.
  3. Once VPC is setup, share the account ID from step 1 with the Fennel support. Also inform Fennel support a) whether you want the endpoints to be public or private and b) the email domain(s) that should be allowed to access Fennel web console.
  4. In response, you’d be given a cloud formation template and a cluster ID. Run the cloud formation template from the account chosen in step 1 with the following inputs: a) VPC: VPC ID, publicSubnets, privateSubnets, cidr b) cluster ID
  5. This template provisions the outpost EC2 instance(s) inside the VPC, a security group, couple of IAM roles, a role permission boundary and some other machinery.
  6. Sit back as Fennel Support deploys the platform in this account and manages it from there. When done, Fennel will give you URLs of the Fennel console and the main API servers.
  7. If you want Fennel to ingest data from a private resource (e.g. RDS, Snowflake) in any of your accounts, you may either need to peer the VPC set up in step 2 with your other AWS accounts or add the IP of the NAT gateway to appropriate allow lists (e.g. Snowflake)
  8. To enable laptop access to Fennel for your team, Fennel can either provide PrivateLink endpoint or you can use VPC peering and/or VPN setup yourself.
  9. Once you have access to Fennel console, use it to generate access tokens (you can also create new RBAC roles if needed and/or assign users to roles).
  10. Instantiate Fennel client by providing server URL and access tokens. Use this client to talk to the server.

And you are all set now!

On This Page

Edit this Page on Github