When Memory Serves You: Using ramfs and tmpfs

[ Thanks to An Anonymous Reader for this link. ]

"If your read/write performance isn't keeping up with your needs, the least expensive and least time-consuming fix is to place those heavily used files into RAM. Reading and writing to RAM is significantly faster than when using disk-based filesystems. Disk I/O-sensitive data transfers, like those involving databases, reap extreme benefits from moving to RAM-based filesystems.

"Why RAM? RAM is fast. It operates in nanosecond-level access times whereas the fastest disk operates in millisecond-level access times. RAM doesn't spin. Mechanical drives spin, which means their read/write and seek speeds are significantly slower than their RAM-based equivalents. DDR3 RAM, for example, moves data in and out at solder-melting peak rates exceeding 10GB/s. Even Hitachi's hottest 15,000 RPM UltraStar disk, transfers data at a sluggish 119MB/s to 198MB/s sustained and 600MB/s max. RAM has longer MTBF (Mean Time Between Failures). Since RAM isn't mechanical, it doesn't enjoy the high failure rates of spinning disks, therefore giving it a life expectancy well beyond that of a typical disk drive."

Complete Story

Related Stories:

• Opinion: On pramfs and RAM based Linux file systems(Oct 02, 2009)
• The Kernel Newbie Corner: "initrd" and "initramfs"--What's Up With That?(Oct 01, 2009)
• Is There a Perfect Linux Filesystem?(Jun 30, 2009)
• Kernel Log: What's coming in 2.6.30 - File systems: New and revamped file systems(Apr 28, 2009)
• ext4 File System: Introduction and Benchmarks(Mar 30, 2009)
• Overview of RAMFS and TMPFS on Linux(Mar 10, 2009)