Tuple Structure

[Atri Sharma]

Can someone please shed some light on tuple Structure we use? Specifically
the headers

[dsetrakyan]

On Thu, Jul 30, 2015 at 12:44 PM, Atri Sharma <[hidden email]> wrote:

> Can someone please shed some light on tuple Structure we use? Specifically
> the headers
>

Are you asking about IgniteBiTuple class? Do you have specific questions?

[Atri Sharma]

I am trying to think of a way to have read only tuples optimised for
maximum oltp throughput. The first location to look into is the tuple
Structure and how it's headers and structure are maintained. So I am
targeting the headers and the mvcc qualification logic in order to cut down
on these two overheads.

The idea is to have data which is bulk loaded and huge and only used for
reads with high throughput and not persist for long.

Thoughts?
On 31 Jul 2015 01:19, "Dmitriy Setrakyan" <[hidden email]> wrote:

> On Thu, Jul 30, 2015 at 12:44 PM, Atri Sharma <[hidden email]> wrote:
>
> > Can someone please shed some light on tuple Structure we use?
> Specifically
> > the headers
> >
>
> Are you asking about IgniteBiTuple class? Do you have specific questions?
>

[dsetrakyan]

On Thu, Jul 30, 2015 at 12:53 PM, Atri Sharma <[hidden email]> wrote:

> I am trying to think of a way to have read only tuples optimised for
> maximum oltp throughput. The first location to look into is the tuple
> Structure and how it's headers and structure are maintained. So I am
> targeting the headers and the mvcc qualification logic in order to cut down
> on these two overheads.
>

I am not sure which tuple structure you are talking about. Are you asking
in regard to some specific Jira ticket you are working on?

>
> The idea is to have data which is bulk loaded and huge and only used for
> reads with high throughput and not persist for long.
>
> Thoughts?
> On 31 Jul 2015 01:19, "Dmitriy Setrakyan" <[hidden email]> wrote:
>
> > On Thu, Jul 30, 2015 at 12:44 PM, Atri Sharma <[hidden email]>
> wrote:
> >
> > > Can someone please shed some light on tuple Structure we use?
> > Specifically
> > > the headers
> > >
> >
> > Are you asking about IgniteBiTuple class? Do you have specific questions?
> >
>

[dsetrakyan]

On Thu, Jul 30, 2015 at 12:53 PM, Atri Sharma <[hidden email]> wrote:

> I am trying to think of a way to have read only tuples optimised for
> maximum oltp throughput. The first location to look into is the tuple
> Structure and how it's headers and structure are maintained. So I am
> targeting the headers and the mvcc qualification logic in order to cut down
> on these two overheads.
>
> The idea is to have data which is bulk loaded and huge and only used for
> reads with high throughput and not persist for long.
>

Atri, I think in Ignite lingo what you really are describing is a
Read-Only-Cache. Such cache would allow inserts (but not updates or
deletes) and all the read/query functionality. Having such semantics would
allow us to add some pretty significant optimizations for data iterations
and querying.

>
> Thoughts?
> On 31 Jul 2015 01:19, "Dmitriy Setrakyan" <[hidden email]> wrote:
>
> > On Thu, Jul 30, 2015 at 12:44 PM, Atri Sharma <[hidden email]>
> wrote:
> >
> > > Can someone please shed some light on tuple Structure we use?
> > Specifically
> > > the headers
> > >
> >
> > Are you asking about IgniteBiTuple class? Do you have specific questions?
> >
>

[Atri Sharma]

Worth working on?
On 1 Aug 2015 00:06, "Dmitriy Setrakyan" <[hidden email]> wrote:

> On Thu, Jul 30, 2015 at 12:53 PM, Atri Sharma <[hidden email]> wrote:
>
> > I am trying to think of a way to have read only tuples optimised for
> > maximum oltp throughput. The first location to look into is the tuple
> > Structure and how it's headers and structure are maintained. So I am
> > targeting the headers and the mvcc qualification logic in order to cut
> down
> > on these two overheads.
> >
> > The idea is to have data which is bulk loaded and huge and only used for
> > reads with high throughput and not persist for long.
> >
>
> Atri, I think in Ignite lingo what you really are describing is a
> Read-Only-Cache. Such cache would allow inserts (but not updates or
> deletes) and all the read/query functionality. Having such semantics would
> allow us to add some pretty significant optimizations for data iterations
> and querying.
>
>
> >
> > Thoughts?
> > On 31 Jul 2015 01:19, "Dmitriy Setrakyan" <[hidden email]> wrote:
> >
> > > On Thu, Jul 30, 2015 at 12:44 PM, Atri Sharma <[hidden email]>
> > wrote:
> > >
> > > > Can someone please shed some light on tuple Structure we use?
> > > Specifically
> > > > the headers
> > > >
> > >
> > > Are you asking about IgniteBiTuple class? Do you have specific
> questions?
> > >
> >
>