---
title: "2026-09-16 S3 TIFF metadata feasibility"
description: "Range reads recovered pyramid metadata from the failing Gleason TIFF without downloading it."
image: "https://docs.virdx.dev/img/virdx-social-card.png"
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.virdx.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# 2026-09-16 S3 TIFF metadata feasibility

## Work Done

Read-only probe of the failing AGGC Subset1_Test_1_0.tiff through vxData-signed HTTPS and fsspec HTTPFile into tifffile. Object size 779,326,663 bytes. 64KiB blocks: 9 GETs, 562,375 bytes, 1.95s; 256KiB: 8 GETs, 2,069,703 bytes, 1.25s. Retrieved seven YXS pyramid levels and 0.5um/px resolution, without decoding pixels. Timings are single sequential local-client measurements, not a cluster benchmark. Presign measured 0.32s separately on the successful probe.

## Pitfalls

Current virdx-ome TIFF reader treats S3 URLs as local paths. vxData storage.open downloads the whole file. Installed fsspec lacks aiohttp; s3fs is also absent. Probe used temporary aiohttp dependencies in /tmp, no repository/environment lock edits. Signed URL generation used private SDK control-plane API for investigation only; production needs a supported interface. GET-signed URLs should use known object size instead of relying on HEAD authorization.

## Improvements

Prefer existing vxData auth plus signed HTTP range reads over new raw S3 secrets. Keep bounded block caches, reuse a file handle for all metadata, batch presigning near consumption, and preserve fatal auth/network errors. Validate representative TIFF layouts and local/remote metadata equivalence before deployment. Full-dataset performance and concurrency remain untested.

Source: https://docs.virdx.dev/knowledge/inbox/2026-09-16-histo-s3-tiff-metadata-feasibility/index.mdx