On 02/09/2015 09:30 PM, Justin Clift wrote:
> On 23 Jan 2015, at 12:43, Mohammed Rafi K C wrote:
>
>> We implemented two approaches, (1) and (2i) to get some
>> performance numbers. The setup was 4*2 distributed-replicated volume
>> using ram disks as bricks to avoid hard disk bottleneck. And t
On Sun Feb 08 2015 at 10:16:27 PM Ben England wrote:
> Avati, I'm all for your zero-copy RDMA API proposal, but I have a concern
> about your proposed zero-copy fop below...
>
> - Original Message -
> > From: "Anand Avati"
> > To: "Mohammed Rafi K C" , "Gluster Devel" <
> gluster-devel@g
On 23 Jan 2015, at 12:43, Mohammed Rafi K C wrote:
> We implemented two approaches, (1) and (2i) to get some
> performance numbers. The setup was 4*2 distributed-replicated volume
> using ram disks as bricks to avoid hard disk bottleneck. And the numbers
> are attached with the mail.
Awesome. :
Avati, I'm all for your zero-copy RDMA API proposal, but I have a concern about
your proposed zero-copy fop below...
- Original Message -
> From: "Anand Avati"
> To: "Mohammed Rafi K C" , "Gluster Devel"
>
> Cc: "Raghavendra Gowdappa" , "Ben Turner"
> , "Ben England"
> , "Suman Debnat
Couple of comments -
1. rdma can register init/fini functions (via pointers) into iobuf_pool.
Absolutely no need to introduce rdma dependency into libglusterfs.
2. It might be a good idea to take a holistic approach towards zero-copy
with libgfapi + RDMA, rather than a narrow goal of "use pre-reg
Rafi, great results, thanks. Your "io-cache off" columns are read tests with
the io-cache translator disabled, correct? What jumps out at me from your
numbers are two things:
- io-cache translator destroys RDMA read performance.
- approach 2i) "register iobuf pool" is the best approach.
-- o
Hi All,
As I pointed out earlier, for rdma protocol, we need to register memory
which is used during rdma read and write with rdma device. In fact it is a
costly operation. To avoid the registration of memory in i/o path, we
came up with two solutions.
1) To use a separate per-registered iobuf_po
