support for physical ramdisk

support for physical ramdisk

[Cory Coager]

I stumbled upon bcache today and was wondering, how would this work with 
a physical ramdisk instead of a SSD?
--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: support for physical ramdisk

[Kent Overstreet]

It'd function, but performance would be worse than just using the
pagecache, bcache does a whole lot of extra work to get around the
limitations of SSDs and the pagecache is blazing fast. Why would you
want to do that?

On Wed, Jan 19, 2011 at 4:10 PM, Cory Coager <ccoager-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> I stumbled upon bcache today and was wondering, how would this work with a
> physical ramdisk instead of a SSD?
> --
> To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: support for physical ramdisk

[Cory Coager]

I was hoping to enable write back with the use of a battery backed 
ramdisk.  I'm simply substituting the SSD for a physical ramdisk.  The 
ramdisk would outperform the SSD and would probably be safer during 
power loss than a SSD (assuming the battery is in good condition).

On 01/19/2011 07:22 PM, Kent Overstreet wrote:
> It'd function, but performance would be worse than just using the
> pagecache, bcache does a whole lot of extra work to get around the
> limitations of SSDs and the pagecache is blazing fast. Why would you
> want to do that?
--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: support for physical ramdisk

[Cory Coager]

I would think even a low end DDR2 module would saturate the SATA 
interface bandwidth.

How would external journal to bcache help?

On 01/19/2011 08:03 PM, Kent Overstreet wrote:
> Ahh, that makes sense. I'd be curious what performance you get - it
> certainly wasn't designed with ram in mind. You'll want to play around
> with the bucket size and see what works best, too.
>
> Also, in the future I'm going to add support for an external journal
> to bcache - that could be the best of both worlds, for you.
--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: support for physical ramdisk

[Kent Overstreet]

On Wed, Jan 19, 2011 at 5:14 PM, Cory Coager <ccoager-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> I would think even a low end DDR2 module would saturate the SATA interface
> bandwidth.
>
> How would external journal to bcache help?

Well, ram is more expensive per gigabyte than SSDs, particularly if
you've got something with a battery attached - the idea would be to
use your SSD for buffering 80 gb or so of writes, but then buffer all
of bcache's metadata writes in battery backed-up ram, significantly
decreasing the number of write IOPS to the SSD, and making it so
pretty much all the writes to the SSD are big contiguous writes.

>
> On 01/19/2011 08:03 PM, Kent Overstreet wrote:
>>
>> Ahh, that makes sense. I'd be curious what performance you get - it
>> certainly wasn't designed with ram in mind. You'll want to play around
>> with the bucket size and see what works best, too.
>>
>> Also, in the future I'm going to add support for an external journal
>> to bcache - that could be the best of both worlds, for you.
>
--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: support for physical ramdisk

[Cory Coager]

Sounds great, when will it be ready? ;-)

On 01/19/2011 08:27 PM, Kent Overstreet wrote:
> Well, ram is more expensive per gigabyte than SSDs, particularly if
> you've got something with a battery attached - the idea would be to
> use your SSD for buffering 80 gb or so of writes, but then buffer all
> of bcache's metadata writes in battery backed-up ram, significantly
> decreasing the number of write IOPS to the SSD, and making it so
> pretty much all the writes to the SSD are big contiguous writes.
--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: support for physical ramdisk

[Kent Overstreet]

On 01/19/2011 07:34 PM, Cory Coager wrote:
> Sounds great, when will it be ready? ;-)

Well, good question. I've been short funding and thus time for bcache 
lately; not sure what's happening there just yet.

And there's a bunch of things on my todo list; my main priority has been 
the code for a bcache specific block device and superblock for the 
backing device. After that is multiple cache devices, volume management 
and tiered storage, but there's no set order those and 
journalling/external journal support need to be done in.

>
> On 01/19/2011 08:27 PM, Kent Overstreet wrote:
>> Well, ram is more expensive per gigabyte than SSDs, particularly if
>> you've got something with a battery attached - the idea would be to
>> use your SSD for buffering 80 gb or so of writes, but then buffer all
>> of bcache's metadata writes in battery backed-up ram, significantly
>> decreasing the number of write IOPS to the SSD, and making it so
>> pretty much all the writes to the SSD are big contiguous writes.

--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html