Why Computer Architecture Matters for Bioinformatics

Content developed by Ben Rambo-Martin and Kristine Lacek

Bioinformatics workflows often involve:

  • Processing gigabytes to terabytes of sequencing data
  • Running computationally intensive algorithms)
  • Managing memory for large reference genomes and indices
  • Parallelizing tasks across multiple CPU cores

Understanding the underlying hardware helps you:

Scenario Knowledge Applied
Pipeline runs slowly Is it CPU-bound, Memory “swapping” or I/O-bound?
“Out of memory” errors How much RAM does the tool need? Can you use disk-based alternatives?
Choosing cloud instances How many cores? How much memory? SSD vs HDD?
Optimizing tool parameters CPU/Thread count, memory limits, temp directory location
Key insight: Most bioinformatics bottlenecks come from either memory limitations (not enough RAM) or I/O bottlenecks (slow disk reads/writes), not CPU speed.

Module Objectives

  • Basic computer architecture
  • BIOS/UEFI
  • Introduce Operating Systems (OS)
  • Understand how to use Linux on Windows
  • What is a virtual machine is?
  • How do operating systems differ?
  • What does *nix means for bioinformatics?
  • Install WSL on windows machines, access command prompt from MacOS 

Module Objectives

Computer Architecture Overview

The diagram below shows the main components of a computer and how they interact:

Computer
Operating System
Hardware
Firmware
UEFI / BIOS
Core Compute Hardware
CPU
L1 Cache
↓
L2 Cache
↓
L3 Cache
Physical RAM
Hard Drive / SSD
Virtual Memory
Permanent Storage
Peripherals
Network
Monitor
USB Controller
⌨ Keyboard
🖱 Mouse
Firmware Operating System Hardware Peripherals USB Devices Network
Dashed arrows = boot-time initialization by firmware. Solid arrows = runtime control through the OS. ↔ = bidirectional data flow.

The CPU (Central Processing Unit)

3.1 What the CPU does

The CPU executes instructions — the fundamental operations that make up all software. For bioinformatics, this includes:

  • Comparing nucleotide sequences character by character
  • Calculating alignment scores
  • Evaluating quality score thresholds

3.2 Cores and threads

Modern CPUs have multiple cores — independent processing units that can work in parallel:

Term Definition Bioinformatics relevance
Core A physical processing unit Each core can run one task at a time
Thread A virtual execution stream Hyperthreading gives 2 threads per core
Multi-threading Using multiple threads -t 8 in BWA-MEM uses 8 threads
Rule of thumb: Set thread count to the number of physical cores, not threads. On a 4-core/8-thread CPU, use -t 4 for CPU-intensive tasks.

3.3 CPU cache hierarchy

Data must travel from RAM to the CPU to be processed. CPU caches are small, ultra-fast memory buffers that store frequently accessed data:

Cache Size (typical) Speed Purpose
L1 32-64 KB per core Fastest Currently executing instructions
L2 256-512 KB per core Very fast Recent data per core
L3 8-64 MB shared Fast Shared across all cores
RAM 8-256+ GB Slower Main working memory

The CPU automatically manages this hierarchy — when it needs data not in cache, it fetches it from RAM (a “cache miss”), which is slower.

Bioinformatics insight: Tools that access data randomly (like hash tables) cause more cache misses than tools that process data sequentially (like streaming through a FASTQ file).

3.4 Checking your CPU

Linux/WSL:

# Number of cores
nproc

# Detailed CPU info
lscpu

# Or from /proc
cat /proc/cpuinfo | grep "model name" | head -1
cat /proc/cpuinfo | grep "cpu cores" | head -1

macOS:

sysctl -n hw.ncpu          # Total threads
sysctl -n hw.physicalcpu   # Physical cores

Memory (RAM)

4.1 What RAM does

RAM holds data that the CPU is actively working with:

  • The genome sequence being aligned against
  • Index structures (FM-index, hash tables, dataframes)
  • Reads currently being processed

When you load a reference genome into a tool like BWA, it gets loaded into RAM. The larger the reference or index, the more RAM required.

4.2 Common memory requirements

Task Typical RAM needed
Mira-influenza 8-16 GB
Mira-sc2, Aligning to human genome (BWA-MEM2) 16-32 GB
BEAST phylogenetics, De novo assembly (SPAdes, large genome) 64-256+ GB
Metagenomics classification (Kraken2) 8-64 GB (depends on database)

4.3 Virtual memory and swap

When RAM runs out, the operating system uses virtual memory — disk space that pretends to be RAM:

  • Also called “swap” (Linux) or “page file” (Windows)
  • Much slower than real RAM (100-1000x slower)
  • Causes severe performance degradation (“thrashing”)
Warning: If your bioinformatics job starts using swap heavily, it may take 10-100x longer to complete. Either reduce memory usage or get more RAM.

4.4 Checking memory usage

Linux/WSL:

# Quick overview
free -h

# Detailed memory info
cat /proc/meminfo | head -10

# Watch memory in real-time
htop   # or: watch -n 1 free -h

macOS:

# Memory summary
vm_stat

# More readable
top -l 1 | head -n 10

4.5 Reducing memory usage

When RAM is limited:

  1. Use streaming algorithms — Process reads one at a time instead of loading all into memory
  2. Reduce thread count — Each thread may need its own memory buffer
  3. Use disk-based alternatives — Some tools offer memory-efficient modes
  4. Split input files — Process in chunks, merge results
  5. Use a cluster/cloud — Rent machines with more RAM

Storage (Disk)

5.1 Storage types

Type Speed Cost
NVMe SSD 3,000-7,000 MB/s $$$
SATA SSD 500-600 MB/s $$
HDD 100-200 MB/s $
Network storage Variable (1-1000 MB/s) Variable

5.2 Impact on bioinformatics

Many workflows are I/O bound — they spend more time reading/writing data than computing:

  • Writing millions of SAM/BAM alignment records
  • Reading large FASTQ files
  • Sorting and indexing BAM files
  • Building and querying databases
Tip: If possible, keep your working data on an SSD. The speed difference is dramatic for I/O-heavy operations like sorting BAM files.

5.3 Checking disk space and speed

Linux/WSL:

# Disk space
df -h

# Which disk a directory is on
df -h /path/to/directory

# Simple write speed test
dd if=/dev/zero of=testfile bs=1G count=1 oflag=direct 2>&1 | tail -1

# Clean up
rm testfile

5.4 Storage best practices

  1. Use SSDs for temp files — Set $TMPDIR or tool-specific temp directories to SSD
  2. Compress when possible — Gzipped FASTQ takes 3-4x less space and often process faster
  3. Clean up intermediate files — BAM files from failed runs, temp files, etc.
  4. Archive completed projects — Move to cheaper HDD or tape storage

Operating System

Operating Systems

6.1 Role of the OS

The operating system:

  • Manages memory — Allocates RAM to programs, handles virtual memory
  • Schedules CPU — Decides which processes run on which cores
  • Handles I/O — Coordinates disk reads/writes, network traffic
  • Provides interfaces — File systems, command line, software installation

6.2 Linux dominance in bioinformatics

Most bioinformatics tools are designed for Linux:

OS Bioinformatics support
Linux (Ubuntu, CentOS, etc.) Excellent — most tools native
macOS Good — Unix-based, most tools work
Windows Limited — use WSL for Linux tools
WSL (Windows Subsystem for Linux) lets you run a full Linux environment inside Windows, giving you access to the Linux bioinformatics ecosystem.

6.3 Key OS concepts

Command Line Interface:

Much of bioinformatics is done in a Linux (or Unix) OS — collectively called *nix. Unix and Linux are built around a command-line–focused design for system control, automation, and scripting.

Command Line Interface

CLI Terminology:

CLI Terminology

  • Shell — The program that interprets command-line input and runs commands (bash, zsh, PowerShell)
  • Prompt — Text displayed by the shell indicating it is ready to accept a command
  • Command — A program or instruction typed into the shell to perform an action
  • Flag (or option) — A modifier added to a command that changes how it behaves

Filesystem and Permissions

  • Directory — A location in the filesystem used to organize files
  • Path — The exact location of a file or directory in the filesystem
  • Permissions — Rules that control who can read, write, or execute a file or directory

Linux Filesystem:

Linux treats most things like files — one consistent way to access resources:

  • Regular files (documents, programs)
  • Directories
  • Devices (/dev)
  • System information (/proc)

All files exist under one root directory:

/
├── home/
├── etc/
├── var/
└── dev/

File systems:

  • Linux uses paths like /home/user/data/
  • Windows uses paths like C:\Users\user\data\
  • WSL bridges both: /mnt/c/Users/ accesses Windows files from Linux

Environment variables:

  • $PATH — Where the shell looks for executables
  • $HOME — Your home directory
  • $TMPDIR — Where temporary files go

Package managers:

  • conda/mamba — Cross-platform, recommended for bioinformatics
  • apt (Ubuntu/Debian) — System packages
  • brew (macOS) — macOS packages

6.4 Checking system info

# OS version
cat /etc/os-release    # Linux
sw_vers                 # macOS

# Kernel version
uname -r

# All system info summary
neofetch   # if installed, fun visualization

Putting It All Together

Bioinformatics and the Command Line

Bioinformatics and CLI

7.1 Minimum specs for bioinformatics

Component Minimum Recommended Heavy workloads
CPU 4 cores 8+ cores 16-32+ cores
RAM 8 GB 32 GB 64-256+ GB
Storage 256 GB SSD 1 TB SSD 2+ TB NVMe SSD

7.2 Diagnosing performance issues

When something runs slowly, ask:

  1. Is CPU at 100%? → CPU-bound, consider more cores or faster algorithm
  2. Is RAM full / swapping? → Memory-bound, reduce usage or get more RAM
  3. Is disk activity constant but CPU idle? → I/O-bound, use faster storage
  4. Is network activity the bottleneck? → Download data locally first

Monitoring tools:

htop          # Interactive process viewer (CPU, RAM per process)
iotop         # Disk I/O by process (requires sudo)
nmon          # All-in-one system monitor

7.3 When to use cloud/cluster

Consider external compute when:

  • Job needs more RAM than your machine has
  • Job would take days on your laptop
  • You need to run many jobs in parallel
  • You need specialized hardware (GPUs for ML)

Common platforms:

  • Institutional HPC clusters — Often “free” for researchers
  • AWS, Google Cloud, Azure — Pay per hour, flexible

7.4 Summary checklist

Before running a pipeline:

  • Do I have enough RAM for the largest step?
  • Do I have enough disk space for inputs, outputs, and temp files?
  • Is my working directory on fast storage (SSD)?
  • Have I set appropriate thread counts for my CPU?
  • Do I know where temp files will be written?

Networks and Servers

Networks and Servers

A network is a system that connects computers and devices, allowing them to share data, resources, and services through wired or wireless communication links.

A server is a computer system designed to provide resources or services—such as data, applications, or network functions—to other devices over a network.

  • Optimized for reliability and continuous operation
  • Handle requests from clients, managing tasks:
    • Data storage
    • Application hosting
    • Authentication
    • Communication between systems

Practical: Command-Line on Your Computer

macOS

macOS icon

  1. Press Command (⌘) + Spacebar to open Spotlight
  2. Type Terminal
  3. Press Return to open it
  4. A command prompt will appear on your screen

Windows

Windows icon

  1. Run PowerShell as Administrator
  2. Run the following command in PowerShell: 
    wsl --install
  3. Restart your computer
  4. Reopen PowerShell and enter the following commands: 
    wsl --set-default-version 2
wsl --install -d Ubuntu-24.04
  5. Following successful installation, an Ubuntu terminal should pop up
  6. Enter a username that will be exclusive for WSL. Press Enter and then enter a password
  7. A command prompt will appear on your screen

Linux

Linux icon

  1. Open the Applications or Activities menu (top-left or bottom-left, depending on your system)
  2. Search for Terminal, Console, or Xterm (names vary by distribution)
  3. Click the icon to launch it
  4. A command prompt will appear on your screen

Terminal prompt example