---
id: "en/deployment-requirements"
title: "Server configuration and performance requirements"
language: "en"
description: "Configuration recommendations and performance boundaries for business servers, databases, search, transcoding, and storage servers."
canonical_url: "https://www.babel.cc/public/babelwiki/en/deployment-requirements.html"
markdown_url: "https://www.babel.cc/public/babelwiki/ai/en/deployment-requirements.md"
alternates: {"en": "https://www.babel.cc/public/babelwiki/ai/en/deployment-requirements.md", "zh": "https://www.babel.cc/public/babelwiki/ai/zh/deployment-requirements.md"}
reading_note: "文档记录不等于当前部署能力；以实际版本、授权模块与配置为准。保留原文的可选、限制及待确认说明，勿补猜接口。 / Documentation is not a guarantee for every deployment. Preserve optional-module requirements, limitations and unresolved API contracts."
---

# Server Sizing and Performance Requirements

Size the deployment for concurrent users, file counts, conversion jobs, index size, storage growth, and availability needs. Licensed users or storage capacity alone cannot determine performance. These are planning references, not guarantees.

## Services

| Service | Purpose | Deployment considerations |
| --- | --- | --- |
| Application and entry point | Web, API, transfer coordination and permissions | May be combined at small scale; HA needs multiple nodes and traffic routing |
| Database | Users, organization, file metadata and business records | Required; colocated, dedicated, or a validated compatible managed database |
| Search | Full-text indexing and retrieval | Allocate resources by index size and query load |
| File processing | Preview conversion, watermarking and extraction | Size by formats and workload; validate Windows/Linux component compatibility |
| File storage | File content | Mounted disks, NFS, or compatible S3 storage; does not replace the database |
| AI / OCR | OCR, vectorization and inference | Evaluate CPU, memory and GPU separately for enabled models and workloads |

Separating business services from storage is common in production, not a universal two-server minimum. All-in-one deployment can support small evaluations but introduces resource contention and a single point of failure.

## Planning Baseline

The white paper provides a historical small-scale baseline of 8 CPU cores, 32GB RAM, a 500GB system disk and gigabit networking for approximately 100 users and less than 10TB of data. Validate it using representative files and actual concurrency. For larger installations, size application, database, search, processing and storage separately; see the [configuration advisor](https://www.babel.cc/public/babelwiki/en/configuration-advisor.html).

Allow for versions, recycle-bin retention, redundancy, replicas, backups and growth. Raw disk capacity is not usable file capacity. RAID, erasure coding and replication do not replace independent backups and recovery tests.

## System and Network Requirements

- Use a security-maintained OS and container environment validated for the delivered BabelBird version, such as a suitable Ubuntu or openEuler release. Not every Linux distribution is automatically compatible.
- Old OS names in historical white papers are not recommendations for new deployments. See the [official CentOS lifecycle notice](https://www.centos.org/centos-linux-eol/).
- Validate the OS, format components and licenses for Windows conversion services. Consult [file preview](https://www.babel.cc/public/babelwiki/en/preview-annotation-review.html) and the delivery specification.
- Do not expose databases or storage services directly to the public Internet. Apply minimum access rules using the [port requirements](https://www.babel.cc/public/babelwiki/en/deployment-ports.html) and an HTTPS entry point.
- Before production, test transfers, permissions, indexing, conversion, backup recovery and peak load. HA environments also require failover and failback tests.