Re: [Nbd] NBD, sparse files & TRIM
- To: Wouter Verhelst <w@...112...>
- Cc: Paul Clements <paul.clements@...856...>, nbd-general@...72...
- Subject: Re: [Nbd] NBD, sparse files & TRIM
- From: folkert <folkert@...421...>
- Date: Wed, 5 Jan 2011 17:27:12 +0100
- Message-id: <20110105162712.GF3629@...855...>
- In-reply-to: <20110105160130.GA27042@...510...>
- References: <20091104173920.GJ6589@...421...> <20091106110734.GA9690@...510...> <20091106143449.GN6589@...421...> <20101228152306.GV9658@...855...> <20101229153211.GA19035@...510...> <AANLkTimYH-m5mabiY26VBXnuhomX6ZmAgkK2aU=-HMG4@...18...> <20101229211412.GB19035@...510...> <20110105155250.GE3629@...855...> <20110105160130.GA27042@...510...>
> > - not sending 0x00 pages but some flag in the responsheader indicating
> > an all 0x00 page
>
> I do intend to write the changes that would be required for what
> triggered this thread in such a way that it would (configurably) also
> work even if the server isn't running off a filesystem that supports
> sparse files (so that it would just write out 0x00 pages instead, rather
> than deallocating them). Or did you mean something else?
No what I mean is if either the server or client detects that it is
going to send a block of all 0x00, that it then only sends a
responsheader with a flag indicating that the block would be 0x00. This
reduces network-traffic. Also the server/client can then decide if they
push out the block for generating sparse files. This is different from
TRIM support as TRIM is for deleted blocks and this is for blocks that
are still in use but are filled with 0x00.
Folkert van Heusden
--
MultiTail is a versatile tool for watching logfiles and output of
commands. Filtering, coloring, merging, diff-view, etc.
http://www.vanheusden.com/multitail/
----------------------------------------------------------------------
Phone: +31-6-41278122, PGP-key: 1F28D8AE, www.vanheusden.com
Reply to: