Member Avatar for RikTelner

Is there any tool (not from system ^1) that I could put in RAM disk and on normal disk. Run both and see the outrageous speed?
Some kind of tool, which will print me Read/Write in MBs?

^1 - Not from system, because I want to see how independant programs are running. That know almost nothing about their environment.

Edited by RikTelner
  • 2 Contributors
  • 2 Replies
  • 311 Views
  • 1 Day Discussion Span
  • Latest Post Latest Post by RikTelner

Dani AI

Generated

Short, practical answer that follows up on and : put the same tiny benchmark binary in both places and have it do a large sequential write and a large sequential read — that will show the raw user-space throughput you care about. Be aware of two things up front: tmpfs (the usual RAM-disk) lives in the kernel pagecache (not a block device), and it does not support O_DIRECT, so tests that rely on direct I/O behave differently on tmpfs vs block devices. (docs.kernel.org)

Quick recipes (safe, repeatable):

Mount a tmpfs:

sudo mount -t tmpfs -o size=2G tmpfs /mnt/ramdisk

Write test with dd (ramdisk):

dd if=/dev/zero of=/mnt/ramdisk/testfile bs=1M count=1024 status=progress

Write test with dd (SSD — bypass pagecache where possible):

sudo dd if=/dev/zero of=/mnt/ssd/testfile bs=1M count=1024 oflag=direct conv=fdatasync status=progress

Notes and troubleshooting:

  • oflag=direct / O_DIRECT avoids the pagecache but may fail or require alignment on some filesystems; see dd docs for flags/limits. (gnu.org)
  • For read tests drop caches after a sync: sync && sudo sh -c 'echo 3 > /proc/sys/vm/drop_caches'. Use this only for testing — it’s non-destructive but can hurt performance while caches repopulate. (docs.kernel.org)
  • For robust, repeatable measurements use fio: run the same job with direct=1 on the block device and direct=0 on tmpfs (fio reports MB/s and latencies). Example joblines below. (manpages.debian.org)
fio --name=dev-write --filename=/mnt/ssd/testfile --size=1G --bs=1M --rw=write --direct=1 --ioengine=libaio --iodepth=16
fio --name=ram-write --filename=/mnt/ramdisk/testfile --size=1G --bs=1M --rw=write --direct=0 --ioengine=sync

Final cautions: don’t run heavy tests on production mounts, avoid overwriting real devices, and remember tmpfs vs a block ramdisk (/dev/ram*) behave differently — use brd if you need a block-device backed RAM disk. The above gives an independent binary/benchmark approach that demonstrates whether your RAM disk is behaving “outrageously” faster.

Recommended Answers

All 2 Replies

Member Avatar for rch1231

Hello,

No matter what you do you are going to have to interact with the system in order to observe the activity. For Linux there is a great tool called sysstat which takes snapshots of the system activity peroiditacally and then lets you compile the reports later using sar.

http://www.thegeekstuff.com/2011/03/sar-examples/

Member Avatar for RikTelner

I don't think you understood what I meant. I meant to check whether RAM Disk is working with such speed as it is supposed to be (faster than SSD). I did not meant to check system statistics.

Be a part of the DaniWeb community

We're a friendly, industry-focused community of developers, IT pros, digital marketers, and technology enthusiasts meeting, networking, learning, and sharing knowledge.