0

Is there a possibility to use a flash drive as a speed up for conventional hard disks? I got the idea to redirect all read ops to the flash drive if the data is already stored there, and to read from the conventional disks if the data is not found there (and during idle time the freshly accessed data from the conventional disk is stored on the flash disk).

Is this already possible with linux standard tools?

Improve this question

1 Answer 1

Reset to default
2

What you're talking about is called a HSM (Hierarchical Storage Manager). There are some such functions built into ZFS, but generally speaking there isn't anything in standard Linux tools.

One thing that works fine is to put the filesystem journal and/or metadata on a flash disk. This is mostly useful if you have a RAID array, of course. Using XFS, a 6 TB RAID-6 and one small SSD allows blazingly fast speeds.

Another option is to use the newest Adaptaec RAID adapters, that allow to use an SSD as a huge cache, doing precisely what you hoped for, but at the RAID controller firmware level.

Last option, is using the SSD as swap space with high priority.

Improve this answer
2
  • +1: Thanks for pointing me into the right direction. The current ZFS kernel implementation looks very promising, even if only the block layer is implemented yet. But I still miss the true HSM option there. Sadly most HSM options seem to be commercial, and I would love to have an freely available option. If no better solution is added, I will accept your answer.
    – Daniel
    Jun 12, 2010 at 16:15
  • I think using an SSD as swap with high swappiness in the kernel can work great with some workloads. You'll have to benchmark it with your applications.
    – wazoox
    Jun 13, 2010 at 17:15

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .