db faq about RAM

README.md

@@ -19,6 +19,7 @@ Erigon is an implementation of Ethereum (aka "Ethereum client"), on the efficien
   + [JSON-RPC daemon](#json-rpc-daemon)
   + [Run all components by docker-compose](#run-all-components-by-docker-compose)
   + [Grafana dashboard](#grafana-dashboard)
+- [FAQ](#faq)
 - [Getting in touch](#getting-in-touch)
   + [Erigon Discord Server](#erigon-discord-server)
   + [Reporting security issues/concerns](#reporting-security-issues-concerns)
@@ -222,6 +223,17 @@ XDG_DATA_HOME=/preferred/data/folder docker-compose up
 
 `docker-compose up prometheus grafana`, [detailed docs](./cmd/prometheus/Readme.md).
 
+FAQ
+================
+
+### How much RAM do I need
+
+- on SSD: 16Gb RAM sync takes 5 days, 32Gb - 4 days, 64Gb - 3 days
+- on NVMe: 16Gb RAM sync takes 4 days, 32Gb - 4 days, 64Gb - 3 days
+- File systems: Ext4 50% faster than ZFS. ZFS 50% faster than BTRfs. ZFS 50% faster than "zfs compression=on"
+
+Detailed explanation: [./docs/programmers_guide/db_faq.md](./docs/programmers_guide/db_faq.md)
+
 Getting in touch
 ================
 

cmd/rpcdaemon/README.md

@@ -384,7 +384,7 @@ Reduce `--private.api.ratelimit`
 
 ### Read DB directly without Json-RPC/Graphql
 
-[./docs/programmers_guide/read_db.md](./docs/programmers_guide/read_db.md)
+[./docs/programmers_guide/db_faq.md](./docs/programmers_guide/db_faq.md)
 
 ### Batch requests 
 

docs/programmers_guide/read_db.md → docs/programmers_guide/db_faq.md

-# How to read DB directly - not by Json-RPC/Graphql:
+# Database FAQ
+
+### How to read DB directly - not by Json-RPC/Graphql:
 
 There are 2 options exist:
 
@@ -21,8 +23,20 @@ option 2 using - `kv_mdbx.go`
 
 Erigon using MDBX database. But any articles in internet about LMDB are also valid for MDBX.
 
-We have Go, Rust and C++ implementations of `RoKV` interface. 
+We have Go, Rust and C++ implementations of `RoKV` interface.
 
 Rationale and Architecture of DB interface: [./../../ethdb/Readme.md](./../../ethdb/Readme.md)
 
-MDBX docs: [erthink.github.io/libmdbx/](https://erthink.github.io/libmdbx/) and [https://github.com/erthink/libmdbx/blob/master/mdbx.h](https://github.com/erthink/libmdbx/blob/master/mdbx.h)
+MDBX docs: [erthink.github.io/libmdbx/](https://erthink.github.io/libmdbx/)
+and [https://github.com/erthink/libmdbx/blob/master/mdbx.h](https://github.com/erthink/libmdbx/blob/master/mdbx.h)
+
+
+### How RAM used
+
+Erigon will use all available RAM, but this RAM will not belong to Eroigon’s process. OS will own all this
+memory. And OS will maintain hot part of DB in RAM. If OS will need RAM for other programs or for second Erigon instance
+OS will manage all the work. This called PageCache. Erigon itself using under 2Gb. So, Erigon will benefit from more
+RAM and will use all RAM without re-configuration. Same PageCache can be used by other processes if they run on same
+machine by just opening same DB file. For example if RPCDaemon started with —datadir option - it will open db of
+Erigon and will use same PageCache (if data A already in RAM because it’s hot and RPCDaemon read it - then it read it
+from RAM not from Disk). Shared memory.