Mailing-List: contact dev-help@asterixdb.incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@asterixdb.incubator.apache.org
MIME-Version: 1.0
In-Reply-To: 
 <CALgBV_fJo8D-FJ714qxXBJggaqgHakxxVh1ODnTNiprnPwkPTQ@mail.gmail.com>
References: 
 <CALgBV_fJo8D-FJ714qxXBJggaqgHakxxVh1ODnTNiprnPwkPTQ@mail.gmail.com>
Date: Sun, 27 Sep 2015 14:29:47 -0700
Message-ID: 
 <CALjEsF61z-_Ay6RjD+0=tMj=SbOfdUM1HNpNYWTMegfdPk1vmw@mail.gmail.com>
Subject: Re: Buffered result status
From: Yingyi Bu <buyingyi@gmail.com>
To: Wail Alkowaileet <wael.y.k@gmail.com>
Cc: dev@asterixdb.incubator.apache.org
Content-Type: multipart/alternative; boundary=001a11c041beb6a35a0520c147f4

--001a11c041beb6a35a0520c147f4
Content-Type: text/plain; charset=UTF-8

Wail,

    Right, that should be an undocumented feature and the memory manager is
only used by the result distribution framework.
    This is different from Hyracks Buffer Cache.  The buffer cache I
referred to was only used by the storage, e.g., B-Tree, LSM B-tree, R-Tree,
etc.

Best,
Yingyi


On Sun, Sep 27, 2015 at 7:36 AM, Wail Alkowaileet <wael.y.k@gmail.com>
wrote:

> Dears,
>
> I was looking in the base code of the Distributed result and I found that
> the result can be buffered. i.e. the result temp file will only be written
> if the number of available memory pages < the result size. The buffer
> policy is LRU as
> in org.apache.hyracks.control.nc.dataset.DatasetMemoryManager.evictPage()
>
> The memory size can be configured in the NCConfig with the
> option: -result-manager-memory
> Is it an undocumented feature?
>
> @Yingyi is that similar to what you referred to as (Hyracks Buffer Cache) ?
>
> Thanks!
> --
>
> *Regards,*
> Wail Alkowaileet
>

--001a11c041beb6a35a0520c147f4--