--- title: Deployment tiers url: https://www.tines.com/docs/self-hosted/reference-architecture/sizing-and-scaling/deployment-tiers/ updated: 2026-03-10T12:51:41+00:00 --- *[tines.com](https://www.tines.com/llms.txt) › [Docs](https://www.tines.com/llms.txt) › [Self-Hosted](https://www.tines.com/llm/docs/self-hosted.md) › [Reference Architecture](https://www.tines.com/llm/docs/self-hosted/reference-architecture.md) › [Sizing & Scaling](https://www.tines.com/llm/docs/self-hosted/reference-architecture/sizing-and-scaling.md)* # Deployment tiers *[View on tines.com](https://www.tines.com/docs/self-hosted/reference-architecture/sizing-and-scaling/deployment-tiers/)* Choosing the right deployment size starts with understanding your workload. This guide covers five reference tiers from single-node evaluation deployments through to high-availability Kubernetes clusters for production, using daily action runs as the primary scaling dimension. All Tines self-hosted deployments consist of the same core components — **tines-app** (web server), **tines-sidekiq** (background worker), **tines-command-runner** (optional), **PostgreSQL**, **Redis**, and a **load balancer / reverse proxy** — scaled according to workload. ## Deployment Options and Sizing Overview | Component | Tier 1: Small | Tier 2: Medium | Tier 3: Large | Tier 4: X-Large | Tier 5: XX-Large | | --- | --- | --- | --- | --- | --- | | **Daily Action Runs** | Up to 100K | Up to 1M | Up to 5M | Up to 10M | 25M+ | | **Deployment Model** | Docker Compose | Docker Compose | Kubernetes (HA) | Kubernetes (HA) | Kubernetes (HA) | | **Host / Cluster** | Single node | Single node + managed data tier | K8s cluster | K8s cluster | K8s cluster | | **High Availability** | No | No | Yes | Yes | Yes | | **Recommended For** | Evaluation, PoC, small teams | Small-to-mid production | Mid production | Large production | Enterprise / high-throughput | > **NOTE:** These tiers are reference architectures. Actual resource requirements vary depending on story complexity, action payload sizes, webhook volume, concurrent users, and whether AI / command runner features are enabled. Start at the recommended baseline and scale up based on observed utilization. ## Tier 1: Small **Docker Compose (up to 100K actions/day)** All containers run on a single host. Suitable for evaluation, proof-of-concept, and small production workloads. ### Host Specifications | Resource | Specification | | --- | --- | | **vCPU** | 4 | | **RAM** | 16 GB | | **Disk** | 500 GB SSD | | **OS** | Linux (Ubuntu 22.04+, RHEL 8+, Amazon Linux 2) | ### Container Layout (single host) | Container | CPU Allocation | RAM Allocation | Instances | | --- | --- | --- | --- | | **tines-app** | shared | shared | 1 | | **tines-sidekiq** | shared | shared | 1 | | **tines-command-runner** | shared | shared | 1 | | **tines-nginx** | shared | shared | 1 | | **postgres** | shared | shared | 1 | | **redis** | shared | shared | 1 | ### Notes - All services share the single host’s resources; no dedicated infrastructure required. - TLS terminated at the bundled nginx container. - Postgres and Redis run as local containers; regular backups via `pg_dump` cron job recommended. - **Not recommended** for production workloads requiring uptime SLAs. ## Tier 2: Medium **Docker Compose (up to 1M actions/day)** Single application host with externalized (managed) data services recommended. Suitable for small-to-mid production workloads. ### Application Host Specifications | Resource | Specification | | --- | --- | | **vCPU** | 8 | | **RAM** | 32 GB | | **Disk** | 500 GB SSD | | **OS** | Linux (Ubuntu 22.04+, Amazon Linux 2) | ### Container Layout (application host) | Container | CPU Allocation | RAM Allocation | Instances | | --- | --- | --- | --- | | **tines-app** | shared | shared | 2 | | **tines-sidekiq** | shared | shared | 2 | | **tines-command-runner** | shared | shared | 2 | | **tines-nginx** | shared | shared | 1 | ### Data Tier (managed services recommended) | Service | Specification | Example (AWS) | | --- | --- | --- | | **PostgreSQL 14.x** | 4 vCPU, 32 GB RAM, 500 GB SSD, automated backups | RDS `db.r6g.xlarge` | | **Redis 7.2** | 2 vCPU, 4 GB RAM | ElastiCache `cache.r6g.large` | > If managed services are unavailable, Postgres and Redis may run as containers on the application host. In this case, increase the host to **16 vCPU / 64 GB RAM / 1 TB SSD** to accommodate the data tier locally. ### Recommended ENV Tuning | Variable | Value | Notes | | --- | --- | --- | | `SIDEKIQ_CONCURRENCY` | 12 | Default; suitable for this tier | | `RAILS_MAX_THREADS` | 16 | Default | | `DATABASE_POOL` | 24 | Default | ### Notes - Externalizing Postgres and Redis to managed services simplifies backups, scaling, and maintenance. - TLS terminated at nginx container or an upstream load balancer. - Single point of failure at the application host — schedule maintenance windows accordingly. ## Tier 3: Large **Kubernetes (up to 5M actions/day)** Multi-pod Kubernetes deployment with HA data services. First tier to support horizontal pod autoscaling and data tier redundancy. ### Application Tier (K8s pods) | Component | vCPU (per pod) | RAM (per pod) | Pod Count | Total vCPU | Total RAM | | --- | --- | --- | --- | --- | --- | | **tines-app** | 2 | 8 GB | 3 | 6 | 24 GB | | **tines-sidekiq** | 2 | 8 GB | 5 | 10 | 40 GB | | **tines-command-runner** | 2 | 8 GB | 2 | 4 | 16 GB | | **Subtotal** | | | **10** | **20** | **80 GB** | ### Data Tier (High Availability) | Service | vCPU (per instance) | RAM (per instance) | Instances | Total vCPU | Total RAM | Example (AWS) | | --- | --- | --- | --- | --- | --- | --- | | **PostgreSQL 14.x** | 8 | 64 GB | 2 (Primary + Secondary) | 16 | 128 GB | RDS `db.r6g.2xlarge` (Multi-AZ) | | **Redis 7.2** | 2 | 4 GB | 2 (Primary + Replica) | 4 | 8 GB | ElastiCache `cache.r6g.large` (replica mode) | ### Cluster Sizing Guidance | Resource | Recommendation | | --- | --- | | **Worker nodes** | 3+ nodes (spread across availability zones for resilience) | | **Node size** | 8 vCPU / 32 GB each (minimum) | | **Ingress** | Cloud load balancer with web application firewall (e.g., AWS ALB + WAF, GCP Cloud Armor) | | **Storage** | Managed Postgres and Redis — no persistent volumes required for app pods | ### Notes - HPA (Horizontal Pod Autoscaler) can be enabled for tines-sidekiq and tines-app to handle burst workloads. - `DATABASE_READONLY_ENDPOINT` can be configured to offload read queries to the Postgres secondary. > **NOTE:** For Tiers 3–5, Redis should be deployed in a **ReplicaSet** (primary + replica) configuration — not sharded. Tines does not provide or manage PostgreSQL high availability; customers are responsible for configuring their own Postgres replication and failover mechanisms (e.g., AWS RDS Multi-AZ, Aurora failover, Patroni, or equivalent). ## Tier 4: X-Large **Kubernetes HA (up to 10M actions/day)** High-availability Kubernetes deployment. All data services run in HA pairs. This is the recommended production configuration for large-scale deployments. ### Application Tier (K8s pods with HPA + ReplicaSets) | Component | vCPU (per pod) | RAM (per pod) | Pod Count | Total vCPU | Total RAM | | --- | --- | --- | --- | --- | --- | | **tines-app** | 2 | 8 GB | 6 | 12 | 48 GB | | **tines-sidekiq** | 2 | 8 GB | 9 | 18 | 72 GB | | **tines-command-runner** | 2 | 8 GB | 3 | 6 | 24 GB | | **Subtotal** | | | **18** | **36** | **144 GB** | ### Data Tier (High Availability) | Service | vCPU (per instance) | RAM (per instance) | Instances | Total vCPU | Total RAM | Example (AWS) | | --- | --- | --- | --- | --- | --- | --- | | **PostgreSQL 14.x** | 16 | 128 GB | 2 (Primary + Secondary) | 32 | 256 GB | Aurora `db.r6g.4xlarge` | | **Redis 7.2** | 2 | 4 GB | 2 (Primary + Secondary) | 4 | 8 GB | ElastiCache `cache.r6g.large` (cluster mode) | ### Cluster Sizing Guidance | Resource | Recommendation | | --- | --- | | **Worker nodes** | 6+ nodes (across 3 availability zones) | | **Node size** | 8 vCPU / 32 GB each (minimum) | | **Ingress** | Cloud load balancer with web application firewall (e.g., AWS ALB + WAF, GCP Cloud Armor) | | **HPA** | Enabled on all application deployments | ### Total Resource Summary | Resource | Count | | --- | --- | | **Application pods** | 18 | | **Application vCPU** | 36 | | **Application RAM** | 144 GB | | **Postgres vCPU** | 32 | | **Postgres RAM** | 256 GB | | **Redis vCPU** | 4 | | **Redis RAM** | 8 GB | | **Grand Total vCPU** | 72 | | **Grand Total RAM** | 408 GB | ### Notes - HPA targets should be tuned based on CPU utilization (recommended: 60–70% target). - `DATABASE_READONLY_ENDPOINT` should be configured to leverage the Postgres secondary for read queries. ## Tier 5: XX-Large Enterprise-scale high-availability deployment for the highest throughput requirements. Scales the Tier 4 reference architecture proportionally. ### Application Tier (K8s pods with HPA + ReplicaSets) | Component | vCPU (per pod) | RAM (per pod) | Pod Count | Total vCPU | Total RAM | | --- | --- | --- | --- | --- | --- | | **tines-app** | 2 | 8 GB | 12 | 24 | 96 GB | | **tines-sidekiq** | 2 | 8 GB | 18 | 36 | 144 GB | | **tines-command-runner** | 2 | 8 GB | 6 | 12 | 48 GB | | **Subtotal** | | | **36** | **72** | **288 GB** | ### Data Tier (High Availability) | Service | vCPU (per instance) | RAM (per instance) | Instances | Total vCPU | Total RAM | Example (AWS) | | --- | --- | --- | --- | --- | --- | --- | | **PostgreSQL 14.x** | 32 | 256 GB | 2 (Primary + Secondary) | 64 | 512 GB | Aurora `db.r6g.8xlarge` | | **Redis 6.x** | 4 | 16 GB | 2 (Primary + Secondary) | 8 | 32 GB | ElastiCache `cache.r6g.xlarge` (cluster mode) | ### Cluster Sizing Guidance | Resource | Recommendation | | --- | --- | | **Worker nodes** | 12+ nodes (across 3 availability zones) | | **Node size** | 8 vCPU / 32 GB each (minimum) | | **Ingress** | Cloud load balancer with web application firewall (e.g., AWS ALB + WAF, GCP Cloud Armor) | | **HPA** | Enabled and tuned on all application deployments | ### Total Resource Summary | Resource | Count | | --- | --- | | **Application pods** | 36 | | **Application vCPU** | 72 | | **Application RAM** | 288 GB | | **Postgres vCPU** | 64 | | **Postgres RAM** | 512 GB | | **Redis vCPU** | 8 | | **Redis RAM** | 32 GB | | **Grand Total vCPU** | 144 | | **Grand Total RAM** | 832 GB | ### Notes - HPA targets should be tuned based on CPU utilization (recommended: 60–70% target). - `DATABASE_READONLY_ENDPOINT` should be configured to leverage the Postgres secondary for read queries.