- n
- 418 apps
- measured
- Nov 2026
- method
- How we test
- dataset
- latestdoi:10.5281/zenodo.22983606
- license
- CC BY 4.0
SelfHost Atlas (2026). Self-hosted apps tagged Rust or Go idle at medians of 23 and 28 MB; within almost every language, each extra data store raises idle RAM. https://selfhostatlas.com/blog/idle-ram-by-language-and-database. Data: doi:10.5281/zenodo.22983606Self-hosted apps tagged Rust or Go idle at medians of 23 and 28 MB; within almost every language, each extra data store raises idle RAM
Nov 9, 2026 · 8 min read
data as of · Nov 2026Data & methodn=418
- n
- 418 apps
- measured
- Nov 2026
- method
- How we test
- dataset
- latestdoi:10.5281/zenodo.22983606
- license
- CC BY 4.0
SelfHost Atlas (2026). Self-hosted apps tagged Rust or Go idle at medians of 23 and 28 MB; within almost every language, each extra data store raises idle RAM. https://selfhostatlas.com/blog/idle-ram-by-language-and-database. Data: doi:10.5281/zenodo.22983606Idle RAM for 418 self-hosted apps grouped by the language, framework and data-store tags in our catalog: Rust-tagged apps idle at a median 23 MB, Go-tagged at 28 MB, Java-tagged at 817 MB, and apps with no database tag at 104 MB against 7,970.5 MB for the four with three data-store tags.
- median idle RAM of the 26 measured apps tagged Rust (2026-11-13 export)
- 23 MB
- apps in the data
- 418
- data as of
- Nov 2026
Correction, 2026-11-14. The figures in this post were updated on 2026-11-14 from dataset release v1.11.0. An earlier version, on the 2026-10-16 export of 194 measured apps, said Go-tagged apps were the light outlier at a median of 25 MB, with Rust-tagged apps at 166.5 MB, and that language mattered far less than data stores; with 224 more apps measured, Rust-tagged apps idle at a median of 23 MB and Go-tagged at 28 MB, and the store count still raises idle RAM within almost every language, but language matters too (Java-tagged apps idle at 638.5 MB even with no store).
Every app in our catalog carries a short list of stack tags: the main language or two, sometimes a framework, and the data stores it runs on. We joined those tags to the idle-RAM measurements for 418 apps to see what, if anything, a stack tells you about memory before you install.
Idle RAM by language tag
These are the language tags carried by at least eight measured apps. An app with two language tags (40 catalog apps, 37 of them measured, are tagged both TypeScript and Node.js, for instance) counts in both rows, and the first number is how many catalog apps carry the tag at all.
| Language tag | Catalog apps | Measured | Median idle (MB) | Lightest–heaviest (MB) |
|---|---|---|---|---|
| Rust | 28 | 26 | 23 | 3–16,824 |
| Go | 105 | 92 | 28 | 4–968 |
| C++ | 12 | 11 | 29 | 3–2,168 |
| C# | 10 | 9 | 111 | 77–327 |
| JavaScript | 25 | 25 | 118 | 2–939 |
| PHP | 71 | 66 | 143 | 16–2,220 |
| Node.js | 60 | 57 | 223 | 22–1,940 |
| TypeScript | 115 | 107 | 401 | 4–16,824 |
| Python | 74 | 65 | 402 | 5–16,824 |
| Ruby | 11 | 10 | 682.5 | 208–2,419 |
| Java | 29 | 28 | 817 | 72–2,998 |
The median across all 418 measured apps is 180 MB. Rust and Go are the light end: 15 of the 26 Rust-tagged apps and 58 of the 92 Go-tagged apps idle below 50 MB. The lightest Rust-tagged apps are Chhoto URL, LLDAP, RustDesk and rustypaste (3 MB each); the lightest Go-tagged ones are OAuth2 Proxy, Screego, Traggo and yarr (4 MB each), then Fusion, Gokapi and Shiori (5 MB). Ten Go-tagged apps reach 256 MB, the heaviest being authentik at 968 MB. Java is the heavy end, at a median of 817 MB across 28 apps.
The 16,824 MB at the top of the Rust, Python and TypeScript rows is the same app, PostHog, which carries all three tags; Sentry (13,284 MB) is tagged Python and Rust. Take those two out and the Rust median moves only from 23 to 19 MB, because most of the row is small. A language row is still not measuring the language. Of the 45 apps that idle above 1 GB, 32 carry at least one data-store tag and 16 carry two or more language tags; their tags describe products made of several services.
Idle RAM by data-store tag
The same join on the data-store and message-queue tags carried by at least five measured apps:
| Data-store tag | Catalog apps | Measured | Median idle (MB) | Lightest–heaviest (MB) |
|---|---|---|---|---|
| SQLite | 58 | 55 | 42 | 3–1,803 |
| MariaDB | 18 | 17 | 191 | 105–1,833 |
| PostgreSQL | 114 | 109 | 449 | 30–16,824 |
| MySQL | 21 | 18 | 470 | 16–1,569 |
| MongoDB | 14 | 13 | 663 | 171–2,998 |
| Redis | 38 | 35 | 858 | 98–2,918 |
| ClickHouse | 10 | 9 | 1,272 | 501–16,824 |
Redis is rarely an app's only store: 27 of the 38 Redis-tagged apps are also tagged PostgreSQL, and the median Redis-tagged app carries four tags in total. That is the pattern behind the whole table. The 215 measured apps with at least one data-store tag idle at a median 259 MB; the 203 with none idle at 104 MB.
Count the stores, as well as the language
Group the 418 apps by how many data-store or queue tags they carry (any of PostgreSQL, MySQL, MariaDB, SQLite, MongoDB, Redis, ClickHouse, Elasticsearch, OpenSearch, Kafka, RabbitMQ):
| Data-store tags | Measured apps | Median idle (MB) | Lightest–heaviest (MB) |
|---|---|---|---|
| 0 | 203 | 104 | 2–2,269 |
| 1 | 171 | 217 | 3–3,842 |
| 2 | 40 | 839.5 | 16–3,259 |
| 3 | 4 | 7,970.5 | 1,665–16,824 |
Each extra store multiplies the median: about 2.1 times from none to one, about 3.9 times from one to two, and about 9.5 times from two to three. The four apps with three tags (Rybbit, Langfuse, Sentry and PostHog) all idle above 1,600 MB, and two of them are the two heaviest apps we have measured. The total tag count tells a similar story: the 64 apps with a single tag (34 of them tagged Go) idle at a median 25.5 MB, the 127 with two at 110 MB, the 144 with three at 242.5 MB, and the 61 with four at 607 MB.
The same holds inside almost every language tag. Go-tagged apps with no store tag idle at a median 22 MB, with one at 42 MB and with two or more at 153 MB; for Python the steps are 246, 405 and 1,603.5 MB, and for Node.js 124.5, 259 and 720.5 MB. TypeScript is the exception at the first step (358 MB with no store tag, 299.5 MB with one), then rises to 878 MB with two or more.
This is not surprising once said out loud. Idle RAM is the sum of every container the install starts, databases and queues included, so the plain reading is that an app which needs PostgreSQL, Redis and ClickHouse is running three servers before its own process does anything. The dataset has no per-container breakdown to prove where the memory goes. On these numbers both matter: Java-tagged apps idle at a median 638.5 MB even with no store tag, while within almost every language the median climbs with each store the app ships.
Frameworks, for completeness
Among server-side framework tags with at least six measured apps (the front-end tags React, Vue, Angular and Svelte are left out): Laravel (18 measured) sits at a median 169 MB, Flask (9) at 227 MB and Symfony (9) at 294 MB; NestJS (8) at 536 MB, Next.js (19) at 543 MB, Spring Boot (6) at 654.5 MB, Rails (6) at 679.5 MB and Django (19) at 830 MB (see the best VPS for Python and Django). Thirteen of the 19 Django-tagged and ten of the 19 Next.js-tagged apps carry at least one data-store tag, so read those rows with the section above in mind.
What these tags are
The stack tags are catalog metadata written by us, not a measurement. Each app gets between one and seven tags (the median is three) naming the main language or languages, sometimes a framework, and the data stores the project depends on. They are a description of the project, not an inventory of the containers our install started, and a tag like "Rust" on PostHog says that part of the product is written in Rust, not that the heavy part is. A project written in several languages may carry more than one language tag.
Idle RAM is the memory docker stats reports for every running container of the app, summed, 30 seconds after the install finishes, on a server pruned of earlier apps. It is a snapshot, not steady-state usage, not a peak and not under load. 412 of the 418 measurements were taken on a Google Cloud e2-standard-2 running Ubuntu 26.04 LTS with Docker; six apps (PostHog, Sentry, Langfuse, RustDesk, AppFlowy and LiteLLM) were run on larger e2-standard machines. How the harness works is described on How we test; the per-app extremes are in the heaviest-and-lightest post, and the same measurements grouped by category are in idle RAM by category.
About the numbers
This post joins two files of our open dataset, both exported on 2026-11-13 and published as release v1.11.0 (457 records each): install-verification.json for the 418 measured idle-RAM figures (measured between 2026-08-28 and 2026-11-12) and catalog.json for the stack tags. The dataset is CC BY 4.0 and cited as DOI 10.5281/zenodo.22983606.
Apps in this post
Related

A 2 GB VPS has the idle RAM for 378 of the 418 self-hosted apps we measured; most AI agent platforms, no-code databases and BI tools do not fit
Applying our 1,126 MB app budget for a 2 GB server to 418 measured idle figures, category by category: 40 of the 65 categories with four or more measured apps fit entirely, while four of six AI agent platforms and three of five no-code databases and BI tools do not, and the 2 GB plans in our catalog listed at $6 to $18 a month on 2026-09-27.

MIT edges out AGPL-3.0 across 457 self-hosted apps, and 53 carry an open-core, source-available, project-specific, proprietary or dual license
The licenses of the 457 apps in our catalog as we record them: 117 MIT, 112 AGPL-3.0, 68 Apache-2.0, 50 GPL-3.0, and a long tail of 54 other strings. 202 are copyleft, 202 permissive and 53 something else, and the measured apps in the something-else group idle at a median 607 MB against 125 MB for the permissive ones.

The median self-hosted app takes 1.3 GB of disk right after install; 36 apps at 5 GB or more hold 37% of the total
Docker disk footprint for 418 self-hosted apps, each installed on a server pruned of earlier apps: median 1,310 MB, from 16 MB (rustypaste) to 27,324 MB (PostHog). Disk tracks idle RAM at the rank level (rho 0.82), but 32 apps that idle under 256 MB still take more than 2 GB of disk.
One email when we publish new measurements.
No digests, no roundups. Just the new numbers and what changed.