Re: Dm-cache question
Ming Zhao
Great hearing from you again. I apologize for my late reply because
the end of semester madness and a recent conference trip.
We have made quite a few improvements to dm-cache in addition to
porting it to kernel 3. These include the ability to share cache space
among multiple volumes, store metadata persistently, etc.
But the support for write-back caching has always been kept. Our
recent evaluation results are mainly from write-through caching
because we expect it works better for solid state drives.
Please do not hesitate to contact me if you have any questions.
Best,
- Ming
On Tue, Apr 10, 2012 at 12:55 PM, Mike Young <myo...@wildernessvoice.com> wrote:
> Dear Dr. Zhao,
>
> I hope you are well. It has been quite some time since we last communicated.
> I had taken the very early version of dm-cache that you first published
> while with IBM and converted it to 64-bit and then worked on some utilities
> to make it easier to instantiate and manage the cache.
>
> I see a lot of activity to converge it with the 3.0 kernel and am encouraged
> by that. But when I first started working with it, it looked like more of a
> WB cache or mapper— essentially allowing for a 2-stage file system where
> 100% of the writes are accelerated.
>
> But when I look at the latest diagrams and charts, it looks like a different
> type of cache— one for accelerating hot spots or frequently accessed
> reads/writes. Does it still have the ability to accelerate all writes as it
> once did?
>
> Best regards,
>
> Mike