Skip to content

Deployment Cookbook: Static MDS with Compute

Static MDS plus a compute tier - a projector that derives computed and interpolated content from what the other projectors serve, rather than sourcing an external feed. Add it when you want the platform to produce derived content: computed columns, curve interpolation, and functions from its computation libraries.

Audience: Architect, Operator. Start from Static MDS; this page adds one image.

The compute tier is the projector tier used as a compute node, described in Architecture: Advanced.

The same images, three ways to run them

This is the Static MDS deployment with one image added, mf-projector-compute. Everything from Static MDS applies unchanged, including how its MDS projectors find their ETA server; below is the delta.

Method You launch it with Ideal for
Conventional the host ./run launcher the standard install - this is the primary path
Docker docker compose up a demo or lab on a single machine
Kubernetes standard Deployment / Service manifests a cluster-managed fixed install

The image you add

Image Role Release
mf-projector-compute Derives computed and interpolated content from other projectors' output deploy-mf-projector-compute

It joins the Static MDS images: gateway, admin, core services, and one or more MDS projectors.

The compute projector hosts computation libraries. intrinsic (the SQL arithmetic and comparison operators) is always present; basic, math, and interpolation are available, with more to come. See the Image Catalog for what each library provides.

Conventional Static MDS with Compute

Install exactly as Static MDS, then add one more package:

  1. Download and unpack the mf-projector-compute release above.
  2. Edit its deployment-config/ for your environment. Point it at the pub/sub service so it can source the content it computes from - METAFLUENT_PUBSUB_ADAPTER_CONNECTION=mf-session:8900, or the equivalent in deployment-config/.
  3. Set JAVA_HOME and start it with ./run.

Run more than one compute projector the same way when you want more computation throughput or resilience.

Docker Static MDS with Compute

Take the Static MDS Compose file and add this service:

  mf-projector-compute:
    image: ghcr.io/metafluent-dist-nn/mf-projector-compute-cfg-stable:milestone
    container_name: mf-projector-compute
    network_mode: host
    environment:
      METAFLUENT_HOST_UID: "${METAFLUENT_HOST_UID}"
      METAFLUENT_PUBSUB_ADAPTER_CONNECTION: "mf-session:8900"   # where it sources content to compute from
    volumes:
      - ./logs:/app/logs
      - ./data:/app/data

Bring it up with the rest: METAFLUENT_HOST_UID=$(id -u) docker compose up -d.

Kubernetes Static MDS with Compute

An example. Add this to the Static MDS manifests:

apiVersion: apps/v1
kind: Deployment
metadata: { name: mf-projector-compute }
spec:
  replicas: 1
  selector: { matchLabels: { app: mf-projector-compute } }
  template:
    metadata: { labels: { app: mf-projector-compute } }
    spec:
      containers:
        - name: mf-projector-compute
          image: ghcr.io/metafluent-dist-nn/mf-projector-compute-cfg-stable:milestone
          env:
            - { name: METAFLUENT_PUBSUB_ADAPTER_CONNECTION, value: "mf-session:8900" }

Raise replicas, or add another Deployment, for more computation throughput or resilience.

Use a simulator

The same as Static MDS - add the mf-eta-simulator image to feed the MDS projectors, and the compute tier derives its content from what they serve.

Going further

  • Static MDS - the base deployment this extends.
  • Fine-Grained - run each service separately.
  • Scalable MDS - let the projector and compute tiers grow and shrink with load automatically.
  • Serving real, entitled Refinitiv data? See Configure DACS to turn on authentication and entitlement checks.

Where to go next