System requirements¶

This page describes the operating system, hardware, storage, and network requirements for installing and running MTD Explorer.

Benchmark-based requirements

These requirements are based on a complete warm-cache installation benchmark performed with the full default database set. The benchmark completed successfully on a 20-thread x86-64 computer with 128 GiB of RAM and NVMe storage.

The installation reached approximately 116.6 GiB of process-tree resident memory and 20.2 GiB of swap usage. Systems close to the minimum requirements must therefore have properly configured swap space.

Supported operating systems¶

MTD Explorer is developed and tested on 64-bit GNU/Linux systems.

Currently tested¶

Linux Mint
Ubuntu-based Linux distributions

Not currently supported¶

Native Microsoft Windows
Native macOS

Windows users may be able to use a Linux server, virtual machine, or WSL2, but these configurations have not yet been formally validated.

Hardware requirements¶

The following values apply to a complete MTD Explorer installation, including construction of the default Kraken2 and Bracken databases.

Resource	Minimum	Recommended	Notes
CPU architecture	x86-64	x86-64	ARM64 is not currently validated
CPU threads	8	20 or more	Fewer threads substantially increase installation and analysis time
RAM	128 GB	192 GB or more	256 GB is preferred for additional headroom during large database builds
Swap	32 GB	64 GB	Swap should preferably be located on SSD or NVMe storage
Free storage	1.5 TB	2 TB or more	Includes the installation, databases, Conda environments, and reusable cache
Installation drive	SSD	NVMe SSD	Database construction performs several terabytes of disk I/O
Internet connection	Required	Stable broadband	Used for packages, reference metadata, and database updates

RAM and swap

A full database installation was successfully completed with 128 GiB of RAM, but the benchmark reached approximately 116.6 GiB of resident memory and used approximately 20.2 GiB of swap.

A system with 128 GB of RAM should therefore have at least 32 GB of available swap. Systems without swap may fail during the most memory-intensive Kraken2 or Bracken construction steps.

Storage layout

The benchmark produced approximately:

740 GiB for the MTD Explorer installation and generated databases;
18 GiB for the Conda installation and environments;
373 GiB for the reusable installation cache.

When the cache and installation are stored on the same filesystem, at least 1.5 TB of free space should be available before installation. A minimum of 2 TB is recommended when input data, temporary files, and analysis outputs will also be stored on that filesystem.

The reusable cache may instead be placed on a separate disk. In that configuration, plan for approximately 1 TB on the installation disk and at least 500 GB on the cache disk.

Installation versus analysis

These values describe installation of the complete default database set. Resource requirements for individual analyses depend on the number of samples, FASTQ file sizes, selected databases, and enabled analysis modules.

A separate official clean pipeline benchmark using the 15-sample PRJNA1306560 example dataset completed in 54 h 58 min with 20 threads and reached 117.27 GiB peak process-tree RSS. That analysis also produced approximately 1.39 TB of pipeline output with optional top-5 microbiome-read extraction enabled. These are observed values for that specific workload, not minimum requirements. See Pipeline benchmarking for the complete configuration and stage timings.

Optional Slurm / HPC requirements¶

The hardware table above describes a complete ordinary MTD Explorer installation. The optional Slurm backend has additional cluster requirements.

For HPC execution, prepared compute nodes currently need x86_64 Linux, Slurm access, a consistent node-local MTD Explorer runtime/database layout, and writable local scratch. Nodes may have different CPU counts and RAM; whole-node workers detect the resources exposed by Slurm after allocation.

The default node-local prefix is /MTD_explorer_HPC. The repository, configuration, work directory, and final output directory must be visible where required by the jobs. Raw FASTQs may instead be fetched from the submission host when the remote-input staging option is enabled.

See HPC / Slurm execution for node preparation, database synchronization, Slurm configuration, scratch-space safeguards, and validation.

Storage planning¶

Storage should be available for:

MTD Explorer source code;
Conda environments;
Kraken2 databases;
host reference genomes and indexes;
eggNOG and HUMAnN databases;
the optional reusable installation cache;
input FASTQ files;
analysis outputs and temporary files.

The reusable installation cache may be stored on a separate disk.

Required permissions¶

The user should have:

permission to write inside the MTD Explorer directory;
permission to write to the selected cache directory;
permission to create files in the selected output directory;
sudo access when system packages need to be installed.

MTD Explorer itself should not normally be executed as the root user.

Check your computer¶

Use the following commands before installation.

Operating system¶

cat /etc/os-release

CPU architecture¶

uname -m

Available CPU threads¶

nproc

Available memory¶

free -h

Available disk space¶

df -h

Git availability¶

git --version

Next step¶

Continue to the installation guide.