[PAPER] Efficient Barrier Synchronization

[Thomas Jungblut <th...@googlemail.com>]

Hey all,

I'm sure you already know this paper [1], but I wanted to discuss it a bit.

Besides that the "relaxed synchronization" looks quite cool and the newly
introduced "commit" method, I have seen the put and get operators.
It seems to be very useful that we provide some kind of distributed cache
which is storing the user data. I call it cache, because we can implement
some kind of swapping logic if the data is too large to fit in RAM.
The user can now get stuff from other machines or put it.

I know that this is widely used in enterprise solutions, e.G. EHCache by
terracotta.
I totally think that Apache DirectMemory could be useful for it. Sadly I
don't know if it is distributed nor that they provide some kind of swapping.

So first off all, what do you think about the paper?
What do you think about the caching idea?
And what do you think about using another framework for this, instead of
implementing our own?

[1]
http://people.apache.org/~tjungblut/Efficient%20Barrier%20Synchronization%20Mechanism.pdf

-- 
Thomas Jungblut
Berlin <th...@gmail.com>

Re: [PAPER] Efficient Barrier Synchronization

[Thomas Jungblut <th...@googlemail.com>]

Ah great, I subscribed the dev list. I've read the stuff on the mailing
list about the distributed cache.

Let's hear some other opinions on it, but I would really love to see this
in DirectMemory as well as Hama.

2011/11/25 Tommaso Teofili <to...@gmail.com>

> 2011/11/25 Thomas Jungblut <th...@googlemail.com>
>
> > Hey all,
> >
> > I'm sure you already know this paper [1], but I wanted to discuss it a
> bit.
> >
> > Besides that the "relaxed synchronization" looks quite cool and the newly
> > introduced "commit" method, I have seen the put and get operators.
> > It seems to be very useful that we provide some kind of distributed cache
> > which is storing the user data. I call it cache, because we can implement
> > some kind of swapping logic if the data is too large to fit in RAM.
> > The user can now get stuff from other machines or put it.
> >
> > I know that this is widely used in enterprise solutions, e.G. EHCache by
> > terracotta.
> > I totally think that Apache DirectMemory could be useful for it. Sadly I
> > don't know if it is distributed nor that they provide some kind of
> > swapping.
> >
> > So first off all, what do you think about the paper?
> >
>
> I didn't know it but I'll read it and share my thoughts.
>
>
> > What do you think about the caching idea?
> > And what do you think about using another framework for this, instead of
> > implementing our own?
> >
>
> I am a committer in DirectMemory and it'd be happy to support integration
> between Hama and DM.
> The distributed cache is being discussed but not provided yet but it's sure
> in its roadmap.
> Tommaso
>
>
>
> > [1]
> >
> >
> http://people.apache.org/~tjungblut/Efficient%20Barrier%20Synchronization%20Mechanism.pdf
> >
> > --
> > Thomas Jungblut
> > Berlin <th...@gmail.com>
> >
>



-- 
Thomas Jungblut
Berlin <th...@gmail.com>

Re: [PAPER] Efficient Barrier Synchronization

[Tommaso Teofili <to...@gmail.com>]

2011/11/25 Thomas Jungblut <th...@googlemail.com>

> Hey all,
>
> I'm sure you already know this paper [1], but I wanted to discuss it a bit.
>
> Besides that the "relaxed synchronization" looks quite cool and the newly
> introduced "commit" method, I have seen the put and get operators.
> It seems to be very useful that we provide some kind of distributed cache
> which is storing the user data. I call it cache, because we can implement
> some kind of swapping logic if the data is too large to fit in RAM.
> The user can now get stuff from other machines or put it.
>
> I know that this is widely used in enterprise solutions, e.G. EHCache by
> terracotta.
> I totally think that Apache DirectMemory could be useful for it. Sadly I
> don't know if it is distributed nor that they provide some kind of
> swapping.
>
> So first off all, what do you think about the paper?
>

I didn't know it but I'll read it and share my thoughts.


> What do you think about the caching idea?
> And what do you think about using another framework for this, instead of
> implementing our own?
>

I am a committer in DirectMemory and it'd be happy to support integration
between Hama and DM.
The distributed cache is being discussed but not provided yet but it's sure
in its roadmap.
Tommaso



> [1]
>
> http://people.apache.org/~tjungblut/Efficient%20Barrier%20Synchronization%20Mechanism.pdf
>
> --
> Thomas Jungblut
> Berlin <th...@gmail.com>
>