3.5.1. Storage Provisioning and Partitioning
💡 First Principle: Provisioning is allocating storage capacity to a workload. The key decisions are how much to allocate (capacity planning), how to organize it (partitioning), and where to place special files (page/swap, scratch).
Storage Provisioning Decisions
- Formatting: After partitioning, a file system must be created (formatted) before data can be stored. Choose the correct file system for the OS and workload (ext4 for Linux, NTFS for Windows, VMFS for VMware).
- Connectivity: Verify storage is accessible (correct cabling, correct initiator configuration for SAN, correct network path for NAS) before formatting.
- Partitioning: Divide the disk into logical sections. Best practice for servers: separate OS, applications, data, and logs onto different partitions/volumes. This prevents a log file explosion from filling the OS partition and crashing the server.
- Page/swap/scratch location and size: Place on separate, fast storage. Windows page file: Microsoft's current guidance is to leave it system-managed and size it by crash-dump requirement, not by a multiple of RAM — the old "1.5× physical RAM" rule dates from an era of far smaller memory and wastes hundreds of gigabytes on a modern server. If you must capture a complete memory dump, the page file needs to be at least RAM + 257 MB; a kernel dump needs far less. Linux swap: varies by use case; for servers with abundant RAM, 2GB may suffice.
Disk Quotas
Disk quotas enforce storage limits per user or group. Benefits:
- Prevent single users from consuming all available space
- Force housekeeping behavior
- Enable chargeback (billing departments for storage consumed)
In Windows, disk quotas are configured per volume. In Linux, quotas are configured via the operating system and enforced at the file system level.
Compression and Deduplication
- Compression: Reduces the size of stored data by encoding redundant patterns. Useful for log files, documents, and other compressible data. Adds CPU overhead during read/write.
- Deduplication: Identifies and eliminates duplicate data blocks, storing only one copy and using references elsewhere. Dramatically reduces storage consumption for backup targets, VDI environments, and file servers with many similar files. Requires significant CPU and RAM for the deduplication process.
⚠️ Exam Trap: Deduplication savings depend entirely on the data type. Virtual machine images with many identical blocks can see 80%+ reduction; already-compressed media files (MP4, JPEG) show near-zero savings because they have little redundancy.
Reflection Question: A file server runs out of disk space at 2 AM, crashing all client connections. The OS and data are on the same partition. What architectural change would prevent this from happening in the future?
