[linux-mm-cc] Swap Compression

[Anderson Briglia]

ext IKEDA, Munehiro wrote:
> Hi Anderson,
> How are you doing?
> 
> Anderson Briglia wrote:
>> Hi all,
>>
>> After some background activity regarding ARM-OMAP drivers and other
>> stuff, I'm back to CCache activities  :) . Ikeda, if you have some
>> update on those patches for ccache shrink operations, please, send
>>  me and I can test them inside an embedded environment.
> 
> Well, I don't have any update from my last post, sorry.
> The latest deliverable is available on
> 
> http://tree.celinuxforum.org/CelfPubWiki/JapanTechnicalJamboree16
> 
> Please download patches from "ccache_avoie-oom_v3.tar.bz2" link.
> Especially, patch 0001 - 0004 are debugging patch so I guess they could 
> help you even if the shrink operation is unnecessary for you.
> 
> 
>> Well, I'm intending to do some performance measurements using Swap
>>  compression on a block device, for instance, using
>> MMC cards. I intend to use the Nitin's code as base and implement
>> a kernel module which will do the swap operations. The main issue
>> I see is the swap reference of the pages since I expect to have
>> more than one compressed page inside a block.
> 
> Hmm...interesting.
> You have a plan to swap out ccache'd data?

Swap out ccache'd data. I mean, I intend to implement (just for research and concept prove) a way to save some data from 
a block device with a swap partition configured. Do you have any references, papers or links that could help? :)

> Or to compress a page data just before swapping it out?
> 
> Anyway, yes, indexing might be an issue as you mentioned.
> Althoug just an idea, how about to define half block, quater block, 1/8 
> block, and so on corresponding to compression ratio, and to use the 
> highest some bits of offset(index) to refer them?

Yes! this is a good idea. I believe it could help with the fragmentation (I'm expecting issues here), and with the page 
address retrieval. My code will be based on this CCache version we have (Nitin's version), but much more simple. Without 
the chunks approach and virtual swap area. Once I have something coded I'll post to this mailing list. :)

Regards,

Anderson Briglia
> 
> 
> Best regards,
> IKEDA, Munehiro
> 
> _______________________________________________
> linux-mm-cc mailing list
> linux-mm-cc at lists.laptop.org
> http://lists.laptop.org/listinfo/linux-mm-cc
>