provider analysis

2 posts with the tag "provider analysis"

Cloudinary: Where It Fits and Where It Doesn't

May 30, 2026

What Cloudinary Is

Cloudinary is an image and video platform. You upload a file, you get a URL like res.cloudinary.com/your-cloud/image/upload/w_400,c_fill,f_auto,q_auto/photo.jpg, and Cloudinary handles resizing, format conversion (WebP, AVIF), CDN caching, and delivery. Change a URL parameter, get a new version on demand. The same model covers video, with transcoding and adaptive streaming.

Two other products sit alongside the API: a digital asset management (DAM) tool for marketing teams, sold separately, and MediaFlows for no-code workflow automation. Files are stored on AWS S3 and Google Cloud, in US regions by default.

How the Credit Model Works

Cloudinary does not bill storage, bandwidth, and processing separately. Everything goes through one shared pool of credits (source):

1 credit equals 1,000 transformations, OR 1 GB stored, OR 1 GB delivered, OR 500 seconds of SD video processing, OR 250 seconds of HD video processing.

The pool is flexible: a quiet month uses fewer credits, a viral month uses more. The trade-off is that a single popular video can drain your monthly budget in hours, and you cannot price any one file in isolation.

Two counting details worth knowing:

Transformations count once per unique URL in a month. Repeat views of the same URL are free at the transformation layer (bandwidth still counts).
Bandwidth counts net file bytes served, not HTTPS overhead or retransmits.

The Tiers

All prices monthly. Annual billing saves about 10%.

Plan	Monthly	Annual (per mo)	Credits/mo	Max image	Max video
Free	$0	$0	25	10 MB	100 MB
Plus	$99	$89	225	20 MB	2 GB
Advanced	$249	$224	600	40 MB	4 GB
Advanced Extra	$549	$494	1,350	40 MB	4 GB
Pro PAYG	$1,099	$989	2,750	40 MB	4 GB
Enterprise	Custom	Custom	Custom	Custom	Custom

The Free plan is permanent, not a trial, and Cloudinary states you can run production on it as long as you stay under 25 credits a month. No credit card required to sign up. 25 credits is roughly 5 GB stored, 10 GB delivered, and 10,000 transformations.

Things that surprise people:

Overages exist only on Pro PAYG ($0.45 per extra credit). On every other paid plan, going over triggers warnings, then partial disable, then full disable. There is no grace period.
Custom domain (CNAME) and HTTPS start at Advanced, not Plus. Many teams hit Plus then jump to Advanced just for cdn.example.com.
SAML SSO, multi-CDN, EU data residency, and AI-powered search are Enterprise-only.
DAM is a separate product. A paid API plan does not raise your DAM limits. Full DAM means a second Enterprise contract.
Add-ons (AI tagging, OCR, Rekognition) have free evaluation tiers; production volume requires a paid Cloudinary plan plus paid add-on tiers.

Where Cloudinary Wins

URL-based transformations. Resize, crop, smart-crop on faces, watermark, overlay text, change format and quality, all by editing the URL. No build step.

Automatic format and quality. Add f_auto,q_auto and Cloudinary picks the smallest format and quality the requesting browser supports. Usually 30 to 70 percent smaller than the original with no manual work.

Video without your own pipeline. HLS and DASH adaptive streaming, transcoding, thumbnails, clipping, captions via add-ons. Video burns credits faster than images (250 to 500 seconds per credit), but you do not need to run FFmpeg.

AI features. Auto-tagging, OCR, content moderation, background removal, generative fill. Most are add-ons, integrated into the upload pipeline.

Migration in is well supported. Lazy auto-upload pulls from your existing URLs the first time anyone requests a file, plus the CLI and Upload API for batch jobs. See the Cloudinary migration guide.

Where It Doesn't Fit

Getting files back out. There is no "export everything" button. The Admin API lists assets and lets you download originals one at a time. Backup to your own S3 starts at the Plus tier and only mirrors new uploads. Moving an existing catalog off Cloudinary is a scripting job, or a workflow in Blober.

Cost predictability under traffic spikes. Shared credits mean one popular asset can blow the monthly budget. The only plan that pays overage in cash instead of suspension warnings is Pro PAYG at $989/month annual.

Pure CDN serving. If you have pre-optimized files and just need a CDN, you are paying for the transformation engine you are not using. Cloudflare R2 (zero egress) or Bunny CDN ($0.005 to $0.01/GB) will be much cheaper.

Data residency. US-only storage by default. EU or other regions require Enterprise.

Self-hosted control. Originals live in Cloudinary's AWS and GCS accounts. You can mirror to your own S3 from Plus upward, but the source of truth is theirs.

Cloudinary vs Alternatives

Cloudflare Images is the cheapest serious option. Free plan covers 5,000 unique transformations per month, no card needed. Paid is $5 per 100,000 images stored, $1 per 100,000 images delivered, $0.50 per 1,000 transformations beyond the free 5,000. Less polish on transformations, no DAM, no add-ons. Video is a separate product (Cloudflare Stream, $5 per 1,000 minutes stored, $1 per 1,000 minutes delivered). Much cheaper than Cloudinary at most scales.

ImageKit is the closest direct competitor. Same URL transformation model, integrated DAM, video processing. Free tier with 20 GB bandwidth, paid from around $49/month with explicit storage and bandwidth caps instead of credits. Pulls from your existing origin (S3, GCS, your server), so files do not have to live in their storage. Easier to leave than Cloudinary.

Imgix is transformation and delivery only. No DAM, no widget, no AI. Points at your existing bucket and serves transformed URLs. Per-master-image plus per-GB pricing. Smaller feature set, originals stay where they are.

Bunny Optimizer adds image transformations on top of Bunny CDN. $9.50/month flat per pull zone plus standard Bunny bandwidth. URL-based transformations and format conversion. Weaker on video and AI than Cloudinary, roughly 10x cheaper at scale.

Uploadcare focuses on upload widget plus transformation pipeline. Per-upload and per-GB pricing, more predictable than credits. Smaller feature set.

Imagor and Thumbor are open source. You host them yourself in front of your own storage. Free in software cost, you pay for servers and ops time. No DAM, no AI, no managed CDN.

Using Cloudinary with Blober

Blober supports Cloudinary as a source and a destination. Connect with your Cloud Name, API Key, and API Secret from the Cloudinary Console and Blober can browse your folders, upload from any other supported provider (Google Drive, Dropbox, AWS S3, Backblaze, Cloudflare R2, Wasabi, Rabata, Proton Drive, GoPro Cloud, NAS, local disk), download to any destination, and delete in bulk.

The most useful case is moving out of Cloudinary. Because Cloudinary has no bulk export tool, leaving means listing the catalog, downloading every original, and re-uploading somewhere else. Blober runs that as a single workflow: pick Cloudinary as source, pick the destination, select the files, run it. Transfers stream through your own machine, folders and filenames are preserved, and you can re-run the workflow later to pick up anything new.

The same workflow runs in the other direction if you want to move an existing folder tree from Dropbox or S3 into Cloudinary.

For setup details, see the Cloudinary provider documentation.

Rabata.io: Where It Fits and Where It Doesn't

May 14, 2026

What Rabata Actually Is

Rabata.io is an S3-compatible object storage provider from RCS Technologies (UK) with two products: Hot Storage, general-purpose object storage at $0.01/GB/month in us-east-1, designed for applications, media, and frequently accessed data, and Backup, bulk archival storage at $49/10TB flat in eu-west-2, intended for backups, disaster recovery, and cold data. Both use standard AWS SigV4 authentication, work with any S3 SDK or CLI, and require no code changes to migrate from AWS S3. You swap the endpoint and credentials.

That is the entire product. No compute layer, no managed databases, no dashboard file browser: you cannot preview or view objects through Rabata's web UI, so you need an S3 client or a tool like Blober to actually see what's in your buckets. Just storage with an S3 API.

What the Benchmarks Show

Rabata published benchmarks using MinIO warp v1.0.7 (released January 2025, now superseded by v1.5.0) on a Debian 13 VM in us-east-1 with 8 concurrent threads in September 2025. The methodology is public.

According to their numbers, Rabata wins upload speed by a small margin (1,462 MB/s vs AWS's 1,444) and mixed operations by 2.3x over AWS. It loses on downloads to both Backblaze (2,075 MB/s) and AWS (1,816), and loses small object throughput to iDrive e2 (962 ops/s vs 696).

The mixed operations number is the most relevant for production workloads. Real applications read, write, list, stat, and delete concurrently. Rabata scored 2.3x higher than AWS S3 in that test.

These are same-region tests (us-east-1 to us-east-1). Performance from other geographies is unknown, and Rabata only operates in two regions. The runs are 30 seconds to 10 minutes with 8 threads, so they measure burst, not sustained multi-TB daily throughput over months. The warp version used (v1.0.7, January 2025) was already 8 months old at the time of testing and is now over a year outdated, and newer versions may produce different results. AWS S3 publishes 99.999999999% durability. Rabata publishes no durability SLA, and their terms include a broad "as is" disclaimer with zero liability for data loss.

The Real Use Case

Rabata fits a specific profile:

Write-heavy S3 workloads that need to stay cheap. If you're ingesting backup pipelines, media uploads, log aggregation, or AI training data, and your bottleneck is upload throughput plus cost, Rabata's upload speed at $0.01/GB is competitive, roughly 57% less than AWS's $0.023/GB first-tier pricing (AWS discounts at volume).

The Backup tier at $49/10TB ($0.0048/GB) is priced below Backblaze B2 ($6.95/TB, ~$0.007/GB) and Wasabi ($6.99/TB, ~$0.007/GB, increasing to $7.99/TB in July 2026). Wasabi enforces a 90-day minimum retention. Rabata's Backup tier has no documented minimum retention, but note: egress is capped at 2x your storage amount and billing is in 10TB increments rounded up: store 1TB and you pay for 10TB.

GDPR-compliant EU storage. The eu-west-2 Backup tier gives you EU data residency, which Rabata calls out explicitly. Worth noting: Rabata's parent company (RCS Technologies) operates under UK law, not EU law. Hetzner also offers EU-based S3-compatible storage with three EU regions (NBG1, FSN1, HEL1) versus Rabata's single EU region. For European companies that need S3-compatible storage with data residency guarantees, both are worth evaluating.

No-friction evaluation. 30-day trial, no credit card required per Rabata's signup page.

Where It Doesn't Fit

Download-heavy workloads. If you're serving content to users, Backblaze B2 (2,075 MB/s downloads, ~$0.007/GB) or Cloudflare R2 ($0.015/GB storage, zero egress, weak throughput but free delivery) are better choices depending on whether you're optimizing for speed or cost.
Global distribution. Two regions. If you need worldwide low-latency access, this is not the product.
Enterprise compliance requirements. No published durability SLA, no SOC 2 mention, limited public track record, benchmarks not independently verified.
Ecosystem depth. No lifecycle policies, no event notifications, no cross-region replication, no versioning (or at least none documented), no dashboard file browser. AWS S3 has all of these. Rabata does not.

Where It Sits Competitively

Based on Rabata's own benchmarks (no independent verification available), they offer three things at once that no other single provider does:

Fastest mixed workload performance in their published benchmarks
Simple pricing at $0.01/GB with $0.01/GB egress (Backup tier: egress capped at 2x storage)
No-barrier trial with no credit card required

AWS is faster on downloads but 2-3x more expensive. Backblaze is comparable on storage (~$0.007/GB) but slower on uploads. Cloudflare R2 has zero egress but performs 3-8x worse. Wasabi has no egress fees but enforces 90-day minimums. iDrive wins on small objects but falls behind on mixed workloads.

If your workload is "ingest data via S3 API, store it cheaply, occasionally read it back," Rabata is worth testing. If your workload needs more features, more regions, or a long track record, look elsewhere.

Using Rabata with Blober

Blober supports Rabata.io as a native provider. Connect with your access key and secret key, and Blober detects your buckets across both regions (Hot Storage and Backup). You can use Rabata as a source or destination in any workflow: migrate to it from AWS S3, sync from Dropbox, back up from Google Drive, or download files from Rabata to your local machine. Since Rabata's dashboard has no built-in file browser, Blober is one of the easiest ways to actually see and manage what's in your buckets.

What Blober supports with Rabata:

Browse: list buckets and objects across both regions (something Rabata's own dashboard doesn't offer)
Upload: write files to Hot Storage or Backup buckets
Download: pull files from Rabata to local storage or stream to another provider
Copy/Move: transfer objects between buckets
Delete: remove objects

Blober handles the region routing automatically. If a bucket lives in eu-west-2, operations go through the eu-west-2 endpoint. No manual configuration needed.

For setup details, see the Rabata.io provider documentation.