docs: add comprehensive hardware requirements documentation for CSPs #780
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,142 @@ | ||
| # Hardware Requirements for Bud-Stack Platform | ||
|
|
||
| ## Executive Summary | ||
|
|
||
| Bud-Stack is a comprehensive multi-service platform for AI/ML model deployment and cluster management. This document provides infrastructure requirements for Cloud Service Providers (CSPs) and organizations planning to deploy the platform. | ||
|
|
||
| ### Platform Overview | ||
|
|
||
| The platform consists of: | ||
| - **14 Microservices** (Application, cluster management, ML optimization, model registry, etc.) | ||
| - **Core Infrastructure** (Databases, message queues, object storage, authentication) | ||
| - **Observability Stack** (Metrics, logging, distributed tracing) | ||
| - **High-Performance Gateway** (Rust-based API routing) | ||
|
|
||
| --- | ||
|
|
||
| ## Infrastructure Requirements Summary | ||
|
|
||
| ### AI-In-A-Box - OEM | ||
|
|
||
| | Resource | Requirement | | ||
| |----------|-------------| | ||
| | **CPU Cores** | 32 cores | | ||
| | **Memory (RAM)** | 64 GiB | | ||
| | **Storage (SSD)** | 200 GiB | | ||
| | **Operating System** | Linux (Ubuntu 22.04+, RHEL 8+, or OpenShift 4.12+) | | ||
| | **Kubernetes** | Version 1.29+ | | ||
|
|
||
| **Max concurrency**: Upto 100 concurrent users | ||
|
|
||
| --- | ||
|
|
||
| ### Enterprise deployment | ||
|
|
||
| | Resource | Requirement | | ||
| |----------|-------------| | ||
| | **CPU Cores** | 96 cores | | ||
| | **Memory (RAM)** | 384 GiB | | ||
| | **Storage (SSD)** | 5 TiB | | ||
| | **Network Bandwidth** | 10 Gbps | | ||
| | **Operating System** | Linux (Ubuntu 22.04+, RHEL 8+, or OpenShift 4.12+) | | ||
| | **Kubernetes** | Version 1.29+ | | ||
|
|
||
| **Max concurreny**: Upto 1000 concurrent users | ||
|
|
||
| --- | ||
|
|
||
| ### CSP Deployment | ||
|
|
||
| | Resource | Requirement | | ||
| |----------|-------------| | ||
| | **CPU Cores** | 120-200 cores | | ||
| | **Memory (RAM)** | 0.5 - 1 TiB | | ||
| | **Storage (SSD)** | 10 - 20 TiB | | ||
| | **Network Bandwidth** | 10-40 Gbps | | ||
| | **Operating System** | Linux (Ubuntu 22.04+, RHEL 8+, or OpenShift 4.12+) | | ||
| | **Kubernetes** | Version 1.29+ | | ||
|
|
||
| **Concurrency**: 10000+ | ||
|
|
||
| --- | ||
|
|
||
| ## Detailed Architecture | ||
|
|
||
| ### Node Pool Breakdown | ||
|
|
||
| Production deployments use specialized node pools for optimal resource allocation: | ||
|
|
||
| | Node Pool | Purpose | Node Spec | Count | Total Resources | | ||
| |-----------|---------|-----------|-------|-----------------| | ||
| | **Control Plane** | Databases, state management | 8 vCPU, 32GB RAM, 500GB SSD | 3-5 | 24-40 vCPU, 96-160GB RAM | | ||
| | **Application** | Microservices, APIs | 16 vCPU, 32GB RAM, 200GB SSD | 5-10 | 80-160 vCPU, 160-320GB RAM | | ||
| | **Data Plane** | Analytics, storage, messaging | 16 vCPU, 64GB RAM, 1TB SSD | 3-5 | 48-80 vCPU, 192-320GB RAM | | ||
| | **Gateway** | API gateway, ingress | 8 vCPU, 16GB RAM, 100GB SSD | 2-3 | 16-24 vCPU, 32-48GB RAM | | ||
|
|
||
|
|
||
| --- | ||
|
|
||
| ### Persistent Storage Breakdown | ||
|
|
||
| | Component | Size (Min) | Size (Recommended) | Performance | | ||
| |-----------|------------|-------------------|-------------| | ||
| | **Databases** (PostgreSQL) | 10 GiB | 100-200 GiB | 3,000-10,000 IOPS, <10ms latency | | ||
| | **Analytics** (ClickHouse) | 30 GiB | 200-500 GiB | 5,000-20,000 IOPS, <5ms latency | | ||
| | **Object Storage** (Models, Datasets) | 50 GiB | 500 GiB-1 TiB | 1,000-5,000 IOPS, <20ms latency | | ||
| | **Message Queue** (Kafka) | 20 GiB | 100-200 GiB | 2,000-10,000 IOPS, <10ms latency | | ||
| | **Application Data** | 50 GiB | 100-200 GiB | Standard SSD | | ||
| | **Backups** | - | 500 GiB-1 TiB | Standard/Archive | | ||
|
|
||
| **Total Storage**: | ||
| - **Minimum**: 256 GiB | ||
|
Contributor
There was a problem hiding this comment. The 'Minimum' total storage is listed as |
||
| - **Recommended**: 2 TiB | ||
|
|
||
| ### Storage Type Requirements | ||
|
|
||
| - **Premium SSD/NVMe**: Required for databases (PostgreSQL, ClickHouse) | ||
| - **Standard SSD**: Acceptable for application data, metrics | ||
| - **Network Storage**: Supported for shared volumes (NFS, Azure Files, EFS) | ||
|
|
||
| --- | ||
|
|
||
| ## Network Requirements | ||
|
|
||
| | Traffic Type | Minimum | Recommended | Notes | | ||
| |--------------|---------|-------------|-------| | ||
| | **Inter-Node** | 5 Gbps | 10 Gbps | Between cluster nodes | | ||
| | **Internet Ingress** | 1 Gbps | 5 Gbps | API traffic, model uploads | | ||
| | **Internet Egress** | 1 Gbps | 5 Gbps | Model downloads, webhooks | | ||
|
|
||
| --- | ||
|
|
||
| ## High Availability Scenarios | ||
|
|
||
| ### Standard HA Configuration | ||
|
|
||
| | Component | Configuration | Failover Time | | ||
| |-----------|--------------|---------------| | ||
| | **Kubernetes Masters** | 3 nodes (multi-zone) | <30 seconds | | ||
| | **PostgreSQL** | 1 master + 2 replicas | <1 minute | | ||
| | **ClickHouse** | 3-node cluster | <2 minutes | | ||
| | **Redis** | 3-node Sentinel | <10 seconds | | ||
| | **Microservices** | 3+ replicas | Immediate | | ||
| | **Gateway** | 3+ replicas | Immediate | | ||
|
|
||
|
|
||
| ### Key HA Features | ||
|
|
||
| - **Auto-Scaling**: HPA enabled for all stateless services (CPU/memory threshold: 75%) | ||
| - **Health Checks**: Liveness/readiness probes on all pods (5-second intervals) | ||
| - **Anti-Affinity**: Pods distributed across zones to prevent single point of failure | ||
| - **PodDisruptionBudget**: Minimum 50% pods available during updates | ||
| - **Backup Schedule**: Daily database backups, 30-day retention, WAL archiving | ||
|
|
||
|
|
||
|
|
||
| ## Required Software | ||
|
|
||
| - **Kubernetes**: Version 1.29+ | ||
| - **Helm**: Version 3.10 or higher | ||
| - **Container Runtime**: containerd 1.6+ | ||
| - **kubectl**: Matching Kubernetes version | ||
| - **Operating System**: Ubuntu 22.04+, RHEL 8+, or OpenShift 4.12+ | ||
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The resource requirements in this summary table are inconsistent with the totals derived from the 'Detailed Production Architecture' section below. For example, you specify 120-200 CPU cores here, but the detailed breakdown sums to 168-304 vCPU. Similar discrepancies exist for RAM and Storage. To avoid confusion for capacity planning, this summary table should be updated to accurately reflect the totals from the detailed breakdown (after correcting the calculation errors in that section).