Skip to content

Cesivi Capacity Limits

Version: v259.0 Last Updated: 2026-04-10 Plan: PLAN-971

Overview

This document describes the practical capacity limits of a Cesivi server instance. It is the Cesivi equivalent of SharePoint's published thresholds.

All measurements are taken with: - Storage provider: InMemory (fastest; use FileSystem for persistence-realistic tests) - Hardware: Dev machine (typical modern workstation) - Server: In-process WebApplicationFactory (no network overhead)

Note: These limits reflect test measurements under controlled conditions. Production performance depends on hardware, storage provider, and concurrent load.

Capacity Limits Summary

Resource Tested Limit Usable Limit (<2s) Hard Limit SP Equivalent
Items per list (InMemory) 20,000 (test limit) ~10,000 Not yet determined SP: 30M hard, 5K view threshold
Lists per web 500+ 500+ Not yet determined SP: 2,000 recommended
Files per document library 1,000+ 1,000+ Not yet determined SP: 30M hard
Max file size 100MB+ — Configuration-dependent SP: 10GB (SharePoint Online)
Concurrent users (in-process) 10+ 10+ Not yet determined SP: varies by hardware
Unique permission scopes per list 500+ 500+ Not yet determined SP: 50,000 per library

Status: Values marked "Not yet determined" require running CapacityTests or CapacityBenchmarks. Run dotnet test --filter "Category=Performance" to populate actual values.

Detail: Items Per List

Test Methodology

Tests in CapacityTests.FindListItemLimit_IncrementalScale_RecordResults: - Starts at 1,000 items, adds 1,000 per step up to 20,000 - After each step measures: GetItemById, OData filter, OData orderby - Reports usability threshold (query time < 2s) and hard limit (> 10s)

Expected Behavior

Scale GetItemById OData Filter OData OrderBy Status
1,000 < 10ms < 100ms < 200ms Excellent
5,000 < 10ms < 200ms < 500ms Good
10,000 < 10ms < 500ms < 1s Acceptable
20,000 < 10ms < 2s < 3s At usability limit (InMemory)

Recommendation: For lists > 5,000 items, enable indexed columns and use $select to limit returned fields.

Detail: Lists Per Web

Test Methodology

Tests in CapacityTests.CreateLists_100_MeasureEnumerationTime and CreateLists_500_MeasureEnumerationTime: - Creates 100 or 500 lists and measures GET /_api/web/lists enumeration time

Expected Behavior

List Count GET /web/lists Time Status
100 < 1s Excellent
500 < 5s Acceptable
1,000 < 10s At usability limit

Detail: Document Libraries

File Upload Performance

File Count File Size Expected Upload Time Notes
1,000 1KB < 60s Sequential upload
1 10MB < 5s Single file
1 100MB < 30s Single large file

File Enumeration Performance

File Count GET /Files ($top=100) Status
1,000 < 500ms Excellent
10,000 < 2s Acceptable

Detail: Permission Scopes

Unique Permission Items

For lists with items that have unique (broken) inheritance:

Unique Items ACL Recompute Time Security Trimming (100 results)
100 < 1s < 200ms
500 < 5s < 500ms
1,000 < 10s < 1s

Note: Cesivi uses a computed ACL cache (warm: O(1) lookup). First access after cache invalidation triggers on-demand computation. Use POST /_api/cesivi/permissions/recompute to pre-warm.

Deny ACE Overhead

Items with Deny ACEs add marginal overhead to security trimming: - < 20% Deny ACEs: < 10% overhead vs all-Allow - > 50% Deny ACEs: may add up to 50% overhead — pre-warm the ACL cache

Performance Tips

For Large Lists (> 5,000 items)

  1. Use $select to limit returned fields
  2. Use $top + pagination (server-driven paging via __next)
  3. Index frequently filtered columns via field settings
  4. Use CAML <RowLimit> to limit server-side rows

For Large Document Libraries

  1. Use $top + pagination for file enumeration
  2. Avoid loading all file metadata at once
  3. Use search API for cross-library queries

For Heavy Permission Scenarios

  1. Pre-warm ACL cache with POST /_api/cesivi/permissions/recompute?cascade=true after bulk changes
  2. Configure Cesivi:AclRecalculation:EnabledOnStartup=true for automatic background recalc
  3. Monitor permission recalc status via GET /_api/cesivi/permissions/recalcstatus

Running Capacity Tests

# All performance/capacity tests (xUnit)
bash Scripts/test-watchdog.sh 300 Cesivi.Tests.Performance/Cesivi.Tests.Performance.csproj -- --filter "Category=Performance"

# BenchmarkDotNet capacity benchmarks (Release mode required)
cd Cesivi.Tests.Performance
dotnet run --configuration Release -- --filter "*Capacity*"

# BenchmarkDotNet permission benchmarks
dotnet run --configuration Release -- --filter "*Permission*"

Comparison: Cesivi vs SharePoint Published Limits

Resource Cesivi (InMemory) SharePoint Online SharePoint 2019 On-Premises
Max list items ~10K-20K usable 30M 30M
List view threshold N/A (no threshold) 5,000 5,000
Max lists per site 500+ tested No documented limit No documented limit
Max file size Config-dependent 250GB 2GB (default)
Max unique perm scopes 1,000+ tested 50,000/library 50,000/library

Key difference: Cesivi does NOT enforce a 5,000-item list view threshold. All items in a list are always accessible via REST, regardless of item count. This is an intentional design choice (mock/test server vs production SharePoint).