one point optimisation
Locked
 
|
Igniters! What is the point of one phase optimisation?
-- *Best Regards,* *Kuznetsov Aleksey* |
|
On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV <[hidden email]
> wrote: > Igniters! What is the point of one phase optimisation? > Performance |
|
|
Here is a good blog post about 1phase commit impl in Ignite and its advantages:
http://gridgain.blogspot.com/2014/09/one-phase-commit-fast-transactions-for.html <http://gridgain.blogspot.com/2014/09/one-phase-commit-fast-transactions-for.html> Took a reference to it from there: https://apacheignite.readme.io/docs/transactions#section-two-phase-commit-2pc <https://apacheignite.readme.io/docs/transactions#section-two-phase-commit-2pc> — Denis > On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan <[hidden email]> wrote: > > On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV <[hidden email] >> wrote: > >> Igniters! What is the point of one phase optimisation? >> > > Performance |
|
|
thank u !
пт, 31 Мар 2017 г., 21:06 Denis Magda <[hidden email]>: > Here is a good blog post about 1phase commit impl in Ignite and its > advantages: > > http://gridgain.blogspot.com/2014/09/one-phase-commit-fast-transactions-for.html > < > http://gridgain.blogspot.com/2014/09/one-phase-commit-fast-transactions-for.html > > > > Took a reference to it from there: > > https://apacheignite.readme.io/docs/transactions#section-two-phase-commit-2pc > < > https://apacheignite.readme.io/docs/transactions#section-two-phase-commit-2pc > > > > — > Denis > > > On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan <[hidden email]> > wrote: > > > > On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < > [hidden email] > >> wrote: > > > >> Igniters! What is the point of one phase optimisation? > >> > > > > Performance > > -- *Best Regards,* *Kuznetsov Aleksey* |
|
|
in ur blog u texted belonging to the same partition is nessesary for 1
phase commit. But its not guarantee belonging to the same node. Partition may span many nodes вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV <[hidden email]>: > thank u ! > > пт, 31 Мар 2017 г., 21:06 Denis Magda <[hidden email]>: > > Here is a good blog post about 1phase commit impl in Ignite and its > advantages: > > http://gridgain.blogspot.com/2014/09/one-phase-commit-fast-transactions-for.html > < > http://gridgain.blogspot.com/2014/09/one-phase-commit-fast-transactions-for.html > > > > Took a reference to it from there: > > https://apacheignite.readme.io/docs/transactions#section-two-phase-commit-2pc > < > https://apacheignite.readme.io/docs/transactions#section-two-phase-commit-2pc > > > > — > Denis > > > On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan <[hidden email]> > wrote: > > > > On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < > [hidden email] > >> wrote: > > > >> Igniters! What is the point of one phase optimisation? > >> > > > > Performance > > -- > > *Best Regards,* > > *Kuznetsov Aleksey* > *Best Regards,* *Kuznetsov Aleksey* |
|
|
As far as I know a partition is always allocated to a specific node and does not span nodes. Ignite has default 1024 partitions on start that are split equally across nodes.
> On 3 Apr 2017, at 08:10, ALEKSEY KUZNETSOV <[hidden email]> wrote: > > in ur blog u texted belonging to the same partition is nessesary for 1 > phase commit. But its not guarantee belonging to the same node. Partition > may span many nodes > > вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV <[hidden email]>: > >> thank u ! >> >> пт, 31 Мар 2017 г., 21:06 Denis Magda <[hidden email]>: >> >> Here is a good blog post about 1phase commit impl in Ignite and its >> advantages: >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast-transactions-for.html >> < >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast-transactions-for.html >>> >> >> Took a reference to it from there: >> >> https://apacheignite.readme.io/docs/transactions#section-two-phase-commit-2pc >> < >> https://apacheignite.readme.io/docs/transactions#section-two-phase-commit-2pc >>> >> >> — >> Denis >> >>> On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan <[hidden email]> >> wrote: >>> >>> On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < >> [hidden email] >>>> wrote: >>> >>>> Igniters! What is the point of one phase optimisation? >>>> >>> >>> Performance >> >> -- >> >> *Best Regards,* >> >> *Kuznetsov Aleksey* >> > -- > > *Best Regards,* > > *Kuznetsov Aleksey* |
|
|
Regarding IgniteOnePhaseCommitNearSelfTest test , ignite's one phase optimisation works not as you said. I attached picture of message exchange. There are partial prepare phase exists, along with finish phase. пн, 3 апр. 2017 г. в 10:55, Christos Erotocritou <[hidden email]>: As far as I know a partition is always allocated to a specific node and does not span nodes. Ignite has default 1024 partitions on start that are split equally across nodes. -- Best Regards, Kuznetsov Aleksey |
|
|
Aleksey,
The blog describes the 1-phase commit at a high level, but I am still curious about the differences you found. Can you share them here? D. On Mon, Apr 3, 2017 at 2:11 AM, ALEKSEY KUZNETSOV <[hidden email]> wrote: > Regarding IgniteOnePhaseCommitNearSelfTest test , ignite's one phase > optimisation works not as you said. > I attached picture of message exchange. There are partial prepare phase > exists, along with finish phase. > > > > пн, 3 апр. 2017 г. в 10:55, Christos Erotocritou <[hidden email]>: > >> As far as I know a partition is always allocated to a specific node and >> does not span nodes. Ignite has default 1024 partitions on start that are >> split equally across nodes. >> >> > On 3 Apr 2017, at 08:10, ALEKSEY KUZNETSOV <[hidden email]> >> wrote: >> > >> > in ur blog u texted belonging to the same partition is nessesary for 1 >> > phase commit. But its not guarantee belonging to the same node. >> Partition >> > may span many nodes >> > >> > вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV <[hidden email]>: >> > >> >> thank u ! >> >> >> >> пт, 31 Мар 2017 г., 21:06 Denis Magda <[hidden email]>: >> >> >> >> Here is a good blog post about 1phase commit impl in Ignite and its >> >> advantages: >> >> >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast- >> transactions-for.html >> >> < >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast- >> transactions-for.html >> >>> >> >> >> >> Took a reference to it from there: >> >> >> >> https://apacheignite.readme.io/docs/transactions#section- >> two-phase-commit-2pc >> >> < >> >> https://apacheignite.readme.io/docs/transactions#section- >> two-phase-commit-2pc >> >>> >> >> >> >> — >> >> Denis >> >> >> >>> On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan < >> [hidden email]> >> >> wrote: >> >>> >> >>> On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < >> >> [hidden email] >> >>>> wrote: >> >>> >> >>>> Igniters! What is the point of one phase optimisation? >> >>>> >> >>> >> >>> Performance >> >> >> >> -- >> >> >> >> *Best Regards,* >> >> >> >> *Kuznetsov Aleksey* >> >> >> > -- >> > >> > *Best Regards,* >> > >> > *Kuznetsov Aleksey* >> > -- > > *Best Regards,* > > *Kuznetsov Aleksey* > |
|
|
I've attached test that prints messages exchange . Which shows us that there are more messages then you declared in article. Perhaps, implementation has changed. I created it on base of IgniteOnePhaseCommitNearSelfTest пн, 3 апр. 2017 г. в 19:03, Dmitriy Setrakyan <[hidden email]>: Aleksey, -- Best Regards, Kuznetsov Aleksey |
|
|
any thoughts on one phase commit realization ?
пн, 3 апр. 2017 г. в 19:35, ALEKSEY KUZNETSOV <[hidden email]>: > I've attached test that prints messages exchange . Which shows us that > there are more messages then you declared in article. Perhaps, > implementation has changed. > I created it on base of IgniteOnePhaseCommitNearSelfTest > > пн, 3 апр. 2017 г. в 19:03, Dmitriy Setrakyan <[hidden email]>: > > Aleksey, > > The blog describes the 1-phase commit at a high level, but I am still > curious about the differences you found. Can you share them here? > > D. > > On Mon, Apr 3, 2017 at 2:11 AM, ALEKSEY KUZNETSOV < > [hidden email]> > wrote: > > > Regarding IgniteOnePhaseCommitNearSelfTest test , ignite's one phase > > optimisation works not as you said. > > I attached picture of message exchange. There are partial prepare phase > > exists, along with finish phase. > > > > > > > > пн, 3 апр. 2017 г. в 10:55, Christos Erotocritou <[hidden email] > >: > > > >> As far as I know a partition is always allocated to a specific node and > >> does not span nodes. Ignite has default 1024 partitions on start that > are > >> split equally across nodes. > >> > >> > On 3 Apr 2017, at 08:10, ALEKSEY KUZNETSOV <[hidden email]> > >> wrote: > >> > > >> > in ur blog u texted belonging to the same partition is nessesary for 1 > >> > phase commit. But its not guarantee belonging to the same node. > >> Partition > >> > may span many nodes > >> > > >> > вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV <[hidden email] > >: > >> > > >> >> thank u ! > >> >> > >> >> пт, 31 Мар 2017 г., 21:06 Denis Magda <[hidden email]>: > >> >> > >> >> Here is a good blog post about 1phase commit impl in Ignite and its > >> >> advantages: > >> >> > >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast- > >> transactions-for.html > >> >> < > >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast- > >> transactions-for.html > >> >>> > >> >> > >> >> Took a reference to it from there: > >> >> > >> >> https://apacheignite.readme.io/docs/transactions#section- > >> two-phase-commit-2pc > >> >> < > >> >> https://apacheignite.readme.io/docs/transactions#section- > >> two-phase-commit-2pc > >> >>> > >> >> > >> >> — > >> >> Denis > >> >> > >> >>> On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan < > >> [hidden email]> > >> >> wrote: > >> >>> > >> >>> On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < > >> >> [hidden email] > >> >>>> wrote: > >> >>> > >> >>>> Igniters! What is the point of one phase optimisation? > >> >>>> > >> >>> > >> >>> Performance > >> >> > >> >> -- > >> >> > >> >> *Best Regards,* > >> >> > >> >> *Kuznetsov Aleksey* > >> >> > >> > -- > >> > > >> > *Best Regards,* > >> > > >> > *Kuznetsov Aleksey* > >> > > -- > > > > *Best Regards,* > > > > *Kuznetsov Aleksey* > > > > -- > > *Best Regards,* > > *Kuznetsov Aleksey* > *Best Regards,* *Kuznetsov Aleksey* |
|
|
Aleksey,
Can you elaborate on which of the extra messages you observe? --AG 2017-04-04 14:17 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email]>: > any thoughts on one phase commit realization ? > > пн, 3 апр. 2017 г. в 19:35, ALEKSEY KUZNETSOV <[hidden email]>: > > > I've attached test that prints messages exchange . Which shows us that > > there are more messages then you declared in article. Perhaps, > > implementation has changed. > > I created it on base of IgniteOnePhaseCommitNearSelfTest > > > > пн, 3 апр. 2017 г. в 19:03, Dmitriy Setrakyan <[hidden email]>: > > > > Aleksey, > > > > The blog describes the 1-phase commit at a high level, but I am still > > curious about the differences you found. Can you share them here? > > > > D. > > > > On Mon, Apr 3, 2017 at 2:11 AM, ALEKSEY KUZNETSOV < > > [hidden email]> > > wrote: > > > > > Regarding IgniteOnePhaseCommitNearSelfTest test , ignite's one phase > > > optimisation works not as you said. > > > I attached picture of message exchange. There are partial prepare phase > > > exists, along with finish phase. > > > > > > > > > > > > пн, 3 апр. 2017 г. в 10:55, Christos Erotocritou < > [hidden email] > > >: > > > > > >> As far as I know a partition is always allocated to a specific node > and > > >> does not span nodes. Ignite has default 1024 partitions on start that > > are > > >> split equally across nodes. > > >> > > >> > On 3 Apr 2017, at 08:10, ALEKSEY KUZNETSOV < > [hidden email]> > > >> wrote: > > >> > > > >> > in ur blog u texted belonging to the same partition is nessesary > for 1 > > >> > phase commit. But its not guarantee belonging to the same node. > > >> Partition > > >> > may span many nodes > > >> > > > >> > вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV < > [hidden email] > > >: > > >> > > > >> >> thank u ! > > >> >> > > >> >> пт, 31 Мар 2017 г., 21:06 Denis Magda <[hidden email]>: > > >> >> > > >> >> Here is a good blog post about 1phase commit impl in Ignite and its > > >> >> advantages: > > >> >> > > >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast- > > >> transactions-for.html > > >> >> < > > >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast- > > >> transactions-for.html > > >> >>> > > >> >> > > >> >> Took a reference to it from there: > > >> >> > > >> >> https://apacheignite.readme.io/docs/transactions#section- > > >> two-phase-commit-2pc > > >> >> < > > >> >> https://apacheignite.readme.io/docs/transactions#section- > > >> two-phase-commit-2pc > > >> >>> > > >> >> > > >> >> — > > >> >> Denis > > >> >> > > >> >>> On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan < > > >> [hidden email]> > > >> >> wrote: > > >> >>> > > >> >>> On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < > > >> >> [hidden email] > > >> >>>> wrote: > > >> >>> > > >> >>>> Igniters! What is the point of one phase optimisation? > > >> >>>> > > >> >>> > > >> >>> Performance > > >> >> > > >> >> -- > > >> >> > > >> >> *Best Regards,* > > >> >> > > >> >> *Kuznetsov Aleksey* > > >> >> > > >> > -- > > >> > > > >> > *Best Regards,* > > >> > > > >> > *Kuznetsov Aleksey* > > >> > > > -- > > > > > > *Best Regards,* > > > > > > *Kuznetsov Aleksey* > > > > > > > -- > > > > *Best Regards,* > > > > *Kuznetsov Aleksey* > > > -- > > *Best Regards,* > > *Kuznetsov Aleksey* > |
|
|
The test shows as follows:
assertMessageCount(GridNearTxPrepareRequest.class, 1); assertMessageCount(GridDhtTxPrepareRequest.class, 1); assertMessageCount(GridDhtTxPrepareResponse.class, 1); assertMessageCount(GridNearTxPrepareResponse.class, 1); assertMessageCount(GridNearTxFinishRequest.class, 1); assertMessageCount(GridDhtTxFinishRequest.class, 0); assertMessageCount(GridNearTxFinishResponse.class, 1); ср, 5 апр. 2017 г. в 14:53, Alexey Goncharuk <[hidden email]>: > Aleksey, > > Can you elaborate on which of the extra messages you observe? > > --AG > > 2017-04-04 14:17 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email]>: > > > any thoughts on one phase commit realization ? > > > > пн, 3 апр. 2017 г. в 19:35, ALEKSEY KUZNETSOV <[hidden email] > >: > > > > > I've attached test that prints messages exchange . Which shows us that > > > there are more messages then you declared in article. Perhaps, > > > implementation has changed. > > > I created it on base of IgniteOnePhaseCommitNearSelfTest > > > > > > пн, 3 апр. 2017 г. в 19:03, Dmitriy Setrakyan <[hidden email]>: > > > > > > Aleksey, > > > > > > The blog describes the 1-phase commit at a high level, but I am still > > > curious about the differences you found. Can you share them here? > > > > > > D. > > > > > > On Mon, Apr 3, 2017 at 2:11 AM, ALEKSEY KUZNETSOV < > > > [hidden email]> > > > wrote: > > > > > > > Regarding IgniteOnePhaseCommitNearSelfTest test , ignite's one phase > > > > optimisation works not as you said. > > > > I attached picture of message exchange. There are partial prepare > phase > > > > exists, along with finish phase. > > > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 10:55, Christos Erotocritou < > > [hidden email] > > > >: > > > > > > > >> As far as I know a partition is always allocated to a specific node > > and > > > >> does not span nodes. Ignite has default 1024 partitions on start > that > > > are > > > >> split equally across nodes. > > > >> > > > >> > On 3 Apr 2017, at 08:10, ALEKSEY KUZNETSOV < > > [hidden email]> > > > >> wrote: > > > >> > > > > >> > in ur blog u texted belonging to the same partition is nessesary > > for 1 > > > >> > phase commit. But its not guarantee belonging to the same node. > > > >> Partition > > > >> > may span many nodes > > > >> > > > > >> > вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV < > > [hidden email] > > > >: > > > >> > > > > >> >> thank u ! > > > >> >> > > > >> >> пт, 31 Мар 2017 г., 21:06 Denis Magda <[hidden email]>: > > > >> >> > > > >> >> Here is a good blog post about 1phase commit impl in Ignite and > its > > > >> >> advantages: > > > >> >> > > > >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast- > > > >> transactions-for.html > > > >> >> < > > > >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast- > > > >> transactions-for.html > > > >> >>> > > > >> >> > > > >> >> Took a reference to it from there: > > > >> >> > > > >> >> https://apacheignite.readme.io/docs/transactions#section- > > > >> two-phase-commit-2pc > > > >> >> < > > > >> >> https://apacheignite.readme.io/docs/transactions#section- > > > >> two-phase-commit-2pc > > > >> >>> > > > >> >> > > > >> >> — > > > >> >> Denis > > > >> >> > > > >> >>> On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan < > > > >> [hidden email]> > > > >> >> wrote: > > > >> >>> > > > >> >>> On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < > > > >> >> [hidden email] > > > >> >>>> wrote: > > > >> >>> > > > >> >>>> Igniters! What is the point of one phase optimisation? > > > >> >>>> > > > >> >>> > > > >> >>> Performance > > > >> >> > > > >> >> -- > > > >> >> > > > >> >> *Best Regards,* > > > >> >> > > > >> >> *Kuznetsov Aleksey* > > > >> >> > > > >> > -- > > > >> > > > > >> > *Best Regards,* > > > >> > > > > >> > *Kuznetsov Aleksey* > > > >> > > > > -- > > > > > > > > *Best Regards,* > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > -- > > > > > > *Best Regards,* > > > > > > *Kuznetsov Aleksey* > > > > > -- > > > > *Best Regards,* > > > > *Kuznetsov Aleksey* > > > *Best Regards,* *Kuznetsov Aleksey* |
|
|
Do you have a near cache enabled?
2017-04-05 15:00 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email]>: > The test shows as follows: > assertMessageCount(GridNearTxPrepareRequest.class, 1); > assertMessageCount(GridDhtTxPrepareRequest.class, 1); > assertMessageCount(GridDhtTxPrepareResponse.class, 1); > assertMessageCount(GridNearTxPrepareResponse.class, 1); > assertMessageCount(GridNearTxFinishRequest.class, 1); > assertMessageCount(GridDhtTxFinishRequest.class, 0); > assertMessageCount(GridNearTxFinishResponse.class, 1); > > ср, 5 апр. 2017 г. в 14:53, Alexey Goncharuk <[hidden email]>: > > > Aleksey, > > > > Can you elaborate on which of the extra messages you observe? > > > > --AG > > > > 2017-04-04 14:17 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email]>: > > > > > any thoughts on one phase commit realization ? > > > > > > пн, 3 апр. 2017 г. в 19:35, ALEKSEY KUZNETSOV < > [hidden email] > > >: > > > > > > > I've attached test that prints messages exchange . Which shows us > that > > > > there are more messages then you declared in article. Perhaps, > > > > implementation has changed. > > > > I created it on base of IgniteOnePhaseCommitNearSelfTest > > > > > > > > пн, 3 апр. 2017 г. в 19:03, Dmitriy Setrakyan <[hidden email] > >: > > > > > > > > Aleksey, > > > > > > > > The blog describes the 1-phase commit at a high level, but I am still > > > > curious about the differences you found. Can you share them here? > > > > > > > > D. > > > > > > > > On Mon, Apr 3, 2017 at 2:11 AM, ALEKSEY KUZNETSOV < > > > > [hidden email]> > > > > wrote: > > > > > > > > > Regarding IgniteOnePhaseCommitNearSelfTest test , ignite's one > phase > > > > > optimisation works not as you said. > > > > > I attached picture of message exchange. There are partial prepare > > phase > > > > > exists, along with finish phase. > > > > > > > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 10:55, Christos Erotocritou < > > > [hidden email] > > > > >: > > > > > > > > > >> As far as I know a partition is always allocated to a specific > node > > > and > > > > >> does not span nodes. Ignite has default 1024 partitions on start > > that > > > > are > > > > >> split equally across nodes. > > > > >> > > > > >> > On 3 Apr 2017, at 08:10, ALEKSEY KUZNETSOV < > > > [hidden email]> > > > > >> wrote: > > > > >> > > > > > >> > in ur blog u texted belonging to the same partition is nessesary > > > for 1 > > > > >> > phase commit. But its not guarantee belonging to the same node. > > > > >> Partition > > > > >> > may span many nodes > > > > >> > > > > > >> > вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV < > > > [hidden email] > > > > >: > > > > >> > > > > > >> >> thank u ! > > > > >> >> > > > > >> >> пт, 31 Мар 2017 г., 21:06 Denis Magda <[hidden email]>: > > > > >> >> > > > > >> >> Here is a good blog post about 1phase commit impl in Ignite and > > its > > > > >> >> advantages: > > > > >> >> > > > > >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast- > > > > >> transactions-for.html > > > > >> >> < > > > > >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast- > > > > >> transactions-for.html > > > > >> >>> > > > > >> >> > > > > >> >> Took a reference to it from there: > > > > >> >> > > > > >> >> https://apacheignite.readme.io/docs/transactions#section- > > > > >> two-phase-commit-2pc > > > > >> >> < > > > > >> >> https://apacheignite.readme.io/docs/transactions#section- > > > > >> two-phase-commit-2pc > > > > >> >>> > > > > >> >> > > > > >> >> — > > > > >> >> Denis > > > > >> >> > > > > >> >>> On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan < > > > > >> [hidden email]> > > > > >> >> wrote: > > > > >> >>> > > > > >> >>> On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < > > > > >> >> [hidden email] > > > > >> >>>> wrote: > > > > >> >>> > > > > >> >>>> Igniters! What is the point of one phase optimisation? > > > > >> >>>> > > > > >> >>> > > > > >> >>> Performance > > > > >> >> > > > > >> >> -- > > > > >> >> > > > > >> >> *Best Regards,* > > > > >> >> > > > > >> >> *Kuznetsov Aleksey* > > > > >> >> > > > > >> > -- > > > > >> > > > > > >> > *Best Regards,* > > > > >> > > > > > >> > *Kuznetsov Aleksey* > > > > >> > > > > > -- > > > > > > > > > > *Best Regards,* > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > -- > > > > > > > > *Best Regards,* > > > > > > > > *Kuznetsov Aleksey* > > > > > > > -- > > > > > > *Best Regards,* > > > > > > *Kuznetsov Aleksey* > > > > > > -- > > *Best Regards,* > > *Kuznetsov Aleksey* > |
|
|
yes
ср, 5 апр. 2017 г. в 15:07, Alexey Goncharuk <[hidden email]>: > Do you have a near cache enabled? > > 2017-04-05 15:00 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email]>: > > > The test shows as follows: > > assertMessageCount(GridNearTxPrepareRequest.class, 1); > > assertMessageCount(GridDhtTxPrepareRequest.class, 1); > > assertMessageCount(GridDhtTxPrepareResponse.class, 1); > > assertMessageCount(GridNearTxPrepareResponse.class, 1); > > assertMessageCount(GridNearTxFinishRequest.class, 1); > > assertMessageCount(GridDhtTxFinishRequest.class, 0); > > assertMessageCount(GridNearTxFinishResponse.class, 1); > > > > ср, 5 апр. 2017 г. в 14:53, Alexey Goncharuk <[hidden email] > >: > > > > > Aleksey, > > > > > > Can you elaborate on which of the extra messages you observe? > > > > > > --AG > > > > > > 2017-04-04 14:17 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email] > >: > > > > > > > any thoughts on one phase commit realization ? > > > > > > > > пн, 3 апр. 2017 г. в 19:35, ALEKSEY KUZNETSOV < > > [hidden email] > > > >: > > > > > > > > > I've attached test that prints messages exchange . Which shows us > > that > > > > > there are more messages then you declared in article. Perhaps, > > > > > implementation has changed. > > > > > I created it on base of IgniteOnePhaseCommitNearSelfTest > > > > > > > > > > пн, 3 апр. 2017 г. в 19:03, Dmitriy Setrakyan < > [hidden email] > > >: > > > > > > > > > > Aleksey, > > > > > > > > > > The blog describes the 1-phase commit at a high level, but I am > still > > > > > curious about the differences you found. Can you share them here? > > > > > > > > > > D. > > > > > > > > > > On Mon, Apr 3, 2017 at 2:11 AM, ALEKSEY KUZNETSOV < > > > > > [hidden email]> > > > > > wrote: > > > > > > > > > > > Regarding IgniteOnePhaseCommitNearSelfTest test , ignite's one > > phase > > > > > > optimisation works not as you said. > > > > > > I attached picture of message exchange. There are partial prepare > > > phase > > > > > > exists, along with finish phase. > > > > > > > > > > > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 10:55, Christos Erotocritou < > > > > [hidden email] > > > > > >: > > > > > > > > > > > >> As far as I know a partition is always allocated to a specific > > node > > > > and > > > > > >> does not span nodes. Ignite has default 1024 partitions on start > > > that > > > > > are > > > > > >> split equally across nodes. > > > > > >> > > > > > >> > On 3 Apr 2017, at 08:10, ALEKSEY KUZNETSOV < > > > > [hidden email]> > > > > > >> wrote: > > > > > >> > > > > > > >> > in ur blog u texted belonging to the same partition is > nessesary > > > > for 1 > > > > > >> > phase commit. But its not guarantee belonging to the same > node. > > > > > >> Partition > > > > > >> > may span many nodes > > > > > >> > > > > > > >> > вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV < > > > > [hidden email] > > > > > >: > > > > > >> > > > > > > >> >> thank u ! > > > > > >> >> > > > > > >> >> пт, 31 Мар 2017 г., 21:06 Denis Magda <[hidden email]>: > > > > > >> >> > > > > > >> >> Here is a good blog post about 1phase commit impl in Ignite > and > > > its > > > > > >> >> advantages: > > > > > >> >> > > > > > >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast- > > > > > >> transactions-for.html > > > > > >> >> < > > > > > >> >> http://gridgain.blogspot.com/2014/09/one-phase-commit-fast- > > > > > >> transactions-for.html > > > > > >> >>> > > > > > >> >> > > > > > >> >> Took a reference to it from there: > > > > > >> >> > > > > > >> >> https://apacheignite.readme.io/docs/transactions#section- > > > > > >> two-phase-commit-2pc > > > > > >> >> < > > > > > >> >> https://apacheignite.readme.io/docs/transactions#section- > > > > > >> two-phase-commit-2pc > > > > > >> >>> > > > > > >> >> > > > > > >> >> — > > > > > >> >> Denis > > > > > >> >> > > > > > >> >>> On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan < > > > > > >> [hidden email]> > > > > > >> >> wrote: > > > > > >> >>> > > > > > >> >>> On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < > > > > > >> >> [hidden email] > > > > > >> >>>> wrote: > > > > > >> >>> > > > > > >> >>>> Igniters! What is the point of one phase optimisation? > > > > > >> >>>> > > > > > >> >>> > > > > > >> >>> Performance > > > > > >> >> > > > > > >> >> -- > > > > > >> >> > > > > > >> >> *Best Regards,* > > > > > >> >> > > > > > >> >> *Kuznetsov Aleksey* > > > > > >> >> > > > > > >> > -- > > > > > >> > > > > > > >> > *Best Regards,* > > > > > >> > > > > > > >> > *Kuznetsov Aleksey* > > > > > >> > > > > > > -- > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > > -- > > > > > > > > > > *Best Regards,* > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > -- > > > > > > > > *Best Regards,* > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > -- > > > > *Best Regards,* > > > > *Kuznetsov Aleksey* > > > *Best Regards,* *Kuznetsov Aleksey* |
|
|
This optimization does not work when near cache is enabled because we need
the same ordering on near nodes. You should see the expected number of messages with near cache disabled. 2017-04-05 15:09 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email]>: > yes > > ср, 5 апр. 2017 г. в 15:07, Alexey Goncharuk <[hidden email]>: > > > Do you have a near cache enabled? > > > > 2017-04-05 15:00 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email]>: > > > > > The test shows as follows: > > > assertMessageCount(GridNearTxPrepareRequest.class, 1); > > > assertMessageCount(GridDhtTxPrepareRequest.class, 1); > > > assertMessageCount(GridDhtTxPrepareResponse.class, 1); > > > assertMessageCount(GridNearTxPrepareResponse.class, 1); > > > assertMessageCount(GridNearTxFinishRequest.class, 1); > > > assertMessageCount(GridDhtTxFinishRequest.class, 0); > > > assertMessageCount(GridNearTxFinishResponse.class, 1); > > > > > > ср, 5 апр. 2017 г. в 14:53, Alexey Goncharuk < > [hidden email] > > >: > > > > > > > Aleksey, > > > > > > > > Can you elaborate on which of the extra messages you observe? > > > > > > > > --AG > > > > > > > > 2017-04-04 14:17 GMT+03:00 ALEKSEY KUZNETSOV < > [hidden email] > > >: > > > > > > > > > any thoughts on one phase commit realization ? > > > > > > > > > > пн, 3 апр. 2017 г. в 19:35, ALEKSEY KUZNETSOV < > > > [hidden email] > > > > >: > > > > > > > > > > > I've attached test that prints messages exchange . Which shows us > > > that > > > > > > there are more messages then you declared in article. Perhaps, > > > > > > implementation has changed. > > > > > > I created it on base of IgniteOnePhaseCommitNearSelfTest > > > > > > > > > > > > пн, 3 апр. 2017 г. в 19:03, Dmitriy Setrakyan < > > [hidden email] > > > >: > > > > > > > > > > > > Aleksey, > > > > > > > > > > > > The blog describes the 1-phase commit at a high level, but I am > > still > > > > > > curious about the differences you found. Can you share them here? > > > > > > > > > > > > D. > > > > > > > > > > > > On Mon, Apr 3, 2017 at 2:11 AM, ALEKSEY KUZNETSOV < > > > > > > [hidden email]> > > > > > > wrote: > > > > > > > > > > > > > Regarding IgniteOnePhaseCommitNearSelfTest test , ignite's one > > > phase > > > > > > > optimisation works not as you said. > > > > > > > I attached picture of message exchange. There are partial > prepare > > > > phase > > > > > > > exists, along with finish phase. > > > > > > > > > > > > > > > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 10:55, Christos Erotocritou < > > > > > [hidden email] > > > > > > >: > > > > > > > > > > > > > >> As far as I know a partition is always allocated to a specific > > > node > > > > > and > > > > > > >> does not span nodes. Ignite has default 1024 partitions on > start > > > > that > > > > > > are > > > > > > >> split equally across nodes. > > > > > > >> > > > > > > >> > On 3 Apr 2017, at 08:10, ALEKSEY KUZNETSOV < > > > > > [hidden email]> > > > > > > >> wrote: > > > > > > >> > > > > > > > >> > in ur blog u texted belonging to the same partition is > > nessesary > > > > > for 1 > > > > > > >> > phase commit. But its not guarantee belonging to the same > > node. > > > > > > >> Partition > > > > > > >> > may span many nodes > > > > > > >> > > > > > > > >> > вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV < > > > > > [hidden email] > > > > > > >: > > > > > > >> > > > > > > > >> >> thank u ! > > > > > > >> >> > > > > > > >> >> пт, 31 Мар 2017 г., 21:06 Denis Magda <[hidden email]>: > > > > > > >> >> > > > > > > >> >> Here is a good blog post about 1phase commit impl in Ignite > > and > > > > its > > > > > > >> >> advantages: > > > > > > >> >> > > > > > > >> >> http://gridgain.blogspot.com/ > 2014/09/one-phase-commit-fast- > > > > > > >> transactions-for.html > > > > > > >> >> < > > > > > > >> >> http://gridgain.blogspot.com/ > 2014/09/one-phase-commit-fast- > > > > > > >> transactions-for.html > > > > > > >> >>> > > > > > > >> >> > > > > > > >> >> Took a reference to it from there: > > > > > > >> >> > > > > > > >> >> https://apacheignite.readme.io/docs/transactions#section- > > > > > > >> two-phase-commit-2pc > > > > > > >> >> < > > > > > > >> >> https://apacheignite.readme.io/docs/transactions#section- > > > > > > >> two-phase-commit-2pc > > > > > > >> >>> > > > > > > >> >> > > > > > > >> >> — > > > > > > >> >> Denis > > > > > > >> >> > > > > > > >> >>> On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan < > > > > > > >> [hidden email]> > > > > > > >> >> wrote: > > > > > > >> >>> > > > > > > >> >>> On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < > > > > > > >> >> [hidden email] > > > > > > >> >>>> wrote: > > > > > > >> >>> > > > > > > >> >>>> Igniters! What is the point of one phase optimisation? > > > > > > >> >>>> > > > > > > >> >>> > > > > > > >> >>> Performance > > > > > > >> >> > > > > > > >> >> -- > > > > > > >> >> > > > > > > >> >> *Best Regards,* > > > > > > >> >> > > > > > > >> >> *Kuznetsov Aleksey* > > > > > > >> >> > > > > > > >> > -- > > > > > > >> > > > > > > > >> > *Best Regards,* > > > > > > >> > > > > > > > >> > *Kuznetsov Aleksey* > > > > > > >> > > > > > > > -- > > > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > > > > > -- > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > -- > > > > > > > > > > *Best Regards,* > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > -- > > > > > > *Best Regards,* > > > > > > *Kuznetsov Aleksey* > > > > > > -- > > *Best Regards,* > > *Kuznetsov Aleksey* > |
|
|
Thank you for help!
ср, 5 апр. 2017 г. в 15:14, Alexey Goncharuk <[hidden email]>: > This optimization does not work when near cache is enabled because we need > the same ordering on near nodes. You should see the expected number of > messages with near cache disabled. > > 2017-04-05 15:09 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email]>: > > > yes > > > > ср, 5 апр. 2017 г. в 15:07, Alexey Goncharuk <[hidden email] > >: > > > > > Do you have a near cache enabled? > > > > > > 2017-04-05 15:00 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email] > >: > > > > > > > The test shows as follows: > > > > assertMessageCount(GridNearTxPrepareRequest.class, 1); > > > > assertMessageCount(GridDhtTxPrepareRequest.class, 1); > > > > assertMessageCount(GridDhtTxPrepareResponse.class, 1); > > > > assertMessageCount(GridNearTxPrepareResponse.class, 1); > > > > assertMessageCount(GridNearTxFinishRequest.class, 1); > > > > assertMessageCount(GridDhtTxFinishRequest.class, 0); > > > > assertMessageCount(GridNearTxFinishResponse.class, 1); > > > > > > > > ср, 5 апр. 2017 г. в 14:53, Alexey Goncharuk < > > [hidden email] > > > >: > > > > > > > > > Aleksey, > > > > > > > > > > Can you elaborate on which of the extra messages you observe? > > > > > > > > > > --AG > > > > > > > > > > 2017-04-04 14:17 GMT+03:00 ALEKSEY KUZNETSOV < > > [hidden email] > > > >: > > > > > > > > > > > any thoughts on one phase commit realization ? > > > > > > > > > > > > пн, 3 апр. 2017 г. в 19:35, ALEKSEY KUZNETSOV < > > > > [hidden email] > > > > > >: > > > > > > > > > > > > > I've attached test that prints messages exchange . Which shows > us > > > > that > > > > > > > there are more messages then you declared in article. Perhaps, > > > > > > > implementation has changed. > > > > > > > I created it on base of IgniteOnePhaseCommitNearSelfTest > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 19:03, Dmitriy Setrakyan < > > > [hidden email] > > > > >: > > > > > > > > > > > > > > Aleksey, > > > > > > > > > > > > > > The blog describes the 1-phase commit at a high level, but I am > > > still > > > > > > > curious about the differences you found. Can you share them > here? > > > > > > > > > > > > > > D. > > > > > > > > > > > > > > On Mon, Apr 3, 2017 at 2:11 AM, ALEKSEY KUZNETSOV < > > > > > > > [hidden email]> > > > > > > > wrote: > > > > > > > > > > > > > > > Regarding IgniteOnePhaseCommitNearSelfTest test , ignite's > one > > > > phase > > > > > > > > optimisation works not as you said. > > > > > > > > I attached picture of message exchange. There are partial > > prepare > > > > > phase > > > > > > > > exists, along with finish phase. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 10:55, Christos Erotocritou < > > > > > > [hidden email] > > > > > > > >: > > > > > > > > > > > > > > > >> As far as I know a partition is always allocated to a > specific > > > > node > > > > > > and > > > > > > > >> does not span nodes. Ignite has default 1024 partitions on > > start > > > > > that > > > > > > > are > > > > > > > >> split equally across nodes. > > > > > > > >> > > > > > > > >> > On 3 Apr 2017, at 08:10, ALEKSEY KUZNETSOV < > > > > > > [hidden email]> > > > > > > > >> wrote: > > > > > > > >> > > > > > > > > >> > in ur blog u texted belonging to the same partition is > > > nessesary > > > > > > for 1 > > > > > > > >> > phase commit. But its not guarantee belonging to the same > > > node. > > > > > > > >> Partition > > > > > > > >> > may span many nodes > > > > > > > >> > > > > > > > > >> > вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV < > > > > > > [hidden email] > > > > > > > >: > > > > > > > >> > > > > > > > > >> >> thank u ! > > > > > > > >> >> > > > > > > > >> >> пт, 31 Мар 2017 г., 21:06 Denis Magda <[hidden email] > >: > > > > > > > >> >> > > > > > > > >> >> Here is a good blog post about 1phase commit impl in > Ignite > > > and > > > > > its > > > > > > > >> >> advantages: > > > > > > > >> >> > > > > > > > >> >> http://gridgain.blogspot.com/ > > 2014/09/one-phase-commit-fast- > > > > > > > >> transactions-for.html > > > > > > > >> >> < > > > > > > > >> >> http://gridgain.blogspot.com/ > > 2014/09/one-phase-commit-fast- > > > > > > > >> transactions-for.html > > > > > > > >> >>> > > > > > > > >> >> > > > > > > > >> >> Took a reference to it from there: > > > > > > > >> >> > > > > > > > >> >> > https://apacheignite.readme.io/docs/transactions#section- > > > > > > > >> two-phase-commit-2pc > > > > > > > >> >> < > > > > > > > >> >> > https://apacheignite.readme.io/docs/transactions#section- > > > > > > > >> two-phase-commit-2pc > > > > > > > >> >>> > > > > > > > >> >> > > > > > > > >> >> — > > > > > > > >> >> Denis > > > > > > > >> >> > > > > > > > >> >>> On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan < > > > > > > > >> [hidden email]> > > > > > > > >> >> wrote: > > > > > > > >> >>> > > > > > > > >> >>> On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < > > > > > > > >> >> [hidden email] > > > > > > > >> >>>> wrote: > > > > > > > >> >>> > > > > > > > >> >>>> Igniters! What is the point of one phase optimisation? > > > > > > > >> >>>> > > > > > > > >> >>> > > > > > > > >> >>> Performance > > > > > > > >> >> > > > > > > > >> >> -- > > > > > > > >> >> > > > > > > > >> >> *Best Regards,* > > > > > > > >> >> > > > > > > > >> >> *Kuznetsov Aleksey* > > > > > > > >> >> > > > > > > > >> > -- > > > > > > > >> > > > > > > > > >> > *Best Regards,* > > > > > > > >> > > > > > > > > >> > *Kuznetsov Aleksey* > > > > > > > >> > > > > > > > > -- > > > > > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > > > > > > > > -- > > > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > -- > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > -- > > > > > > > > *Best Regards,* > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > -- > > > > *Best Regards,* > > > > *Kuznetsov Aleksey* > > > *Best Regards,* *Kuznetsov Aleksey* |
|
|
Maybe it will be useful to update the documentation?
2017-04-05 15:15 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email]>: > Thank you for help! > > ср, 5 апр. 2017 г. в 15:14, Alexey Goncharuk <[hidden email]>: > > > This optimization does not work when near cache is enabled because we > need > > the same ordering on near nodes. You should see the expected number of > > messages with near cache disabled. > > > > 2017-04-05 15:09 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email]>: > > > > > yes > > > > > > ср, 5 апр. 2017 г. в 15:07, Alexey Goncharuk < > [hidden email] > > >: > > > > > > > Do you have a near cache enabled? > > > > > > > > 2017-04-05 15:00 GMT+03:00 ALEKSEY KUZNETSOV < > [hidden email] > > >: > > > > > > > > > The test shows as follows: > > > > > assertMessageCount(GridNearTxPrepareRequest.class, 1); > > > > > assertMessageCount(GridDhtTxPrepareRequest.class, 1); > > > > > assertMessageCount(GridDhtTxPrepareResponse.class, 1); > > > > > assertMessageCount(GridNearTxPrepareResponse.class, > 1); > > > > > assertMessageCount(GridNearTxFinishRequest.class, 1); > > > > > assertMessageCount(GridDhtTxFinishRequest.class, 0); > > > > > assertMessageCount(GridNearTxFinishResponse.class, 1); > > > > > > > > > > ср, 5 апр. 2017 г. в 14:53, Alexey Goncharuk < > > > [hidden email] > > > > >: > > > > > > > > > > > Aleksey, > > > > > > > > > > > > Can you elaborate on which of the extra messages you observe? > > > > > > > > > > > > --AG > > > > > > > > > > > > 2017-04-04 14:17 GMT+03:00 ALEKSEY KUZNETSOV < > > > [hidden email] > > > > >: > > > > > > > > > > > > > any thoughts on one phase commit realization ? > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 19:35, ALEKSEY KUZNETSOV < > > > > > [hidden email] > > > > > > >: > > > > > > > > > > > > > > > I've attached test that prints messages exchange . Which > shows > > us > > > > > that > > > > > > > > there are more messages then you declared in article. > Perhaps, > > > > > > > > implementation has changed. > > > > > > > > I created it on base of IgniteOnePhaseCommitNearSelfTest > > > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 19:03, Dmitriy Setrakyan < > > > > [hidden email] > > > > > >: > > > > > > > > > > > > > > > > Aleksey, > > > > > > > > > > > > > > > > The blog describes the 1-phase commit at a high level, but I > am > > > > still > > > > > > > > curious about the differences you found. Can you share them > > here? > > > > > > > > > > > > > > > > D. > > > > > > > > > > > > > > > > On Mon, Apr 3, 2017 at 2:11 AM, ALEKSEY KUZNETSOV < > > > > > > > > [hidden email]> > > > > > > > > wrote: > > > > > > > > > > > > > > > > > Regarding IgniteOnePhaseCommitNearSelfTest test , ignite's > > one > > > > > phase > > > > > > > > > optimisation works not as you said. > > > > > > > > > I attached picture of message exchange. There are partial > > > prepare > > > > > > phase > > > > > > > > > exists, along with finish phase. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 10:55, Christos Erotocritou < > > > > > > > [hidden email] > > > > > > > > >: > > > > > > > > > > > > > > > > > >> As far as I know a partition is always allocated to a > > specific > > > > > node > > > > > > > and > > > > > > > > >> does not span nodes. Ignite has default 1024 partitions on > > > start > > > > > > that > > > > > > > > are > > > > > > > > >> split equally across nodes. > > > > > > > > >> > > > > > > > > >> > On 3 Apr 2017, at 08:10, ALEKSEY KUZNETSOV < > > > > > > > [hidden email]> > > > > > > > > >> wrote: > > > > > > > > >> > > > > > > > > > >> > in ur blog u texted belonging to the same partition is > > > > nessesary > > > > > > > for 1 > > > > > > > > >> > phase commit. But its not guarantee belonging to the > same > > > > node. > > > > > > > > >> Partition > > > > > > > > >> > may span many nodes > > > > > > > > >> > > > > > > > > > >> > вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV < > > > > > > > [hidden email] > > > > > > > > >: > > > > > > > > >> > > > > > > > > > >> >> thank u ! > > > > > > > > >> >> > > > > > > > > >> >> пт, 31 Мар 2017 г., 21:06 Denis Magda < > [hidden email] > > >: > > > > > > > > >> >> > > > > > > > > >> >> Here is a good blog post about 1phase commit impl in > > Ignite > > > > and > > > > > > its > > > > > > > > >> >> advantages: > > > > > > > > >> >> > > > > > > > > >> >> http://gridgain.blogspot.com/ > > > 2014/09/one-phase-commit-fast- > > > > > > > > >> transactions-for.html > > > > > > > > >> >> < > > > > > > > > >> >> http://gridgain.blogspot.com/ > > > 2014/09/one-phase-commit-fast- > > > > > > > > >> transactions-for.html > > > > > > > > >> >>> > > > > > > > > >> >> > > > > > > > > >> >> Took a reference to it from there: > > > > > > > > >> >> > > > > > > > > >> >> > > https://apacheignite.readme.io/docs/transactions#section- > > > > > > > > >> two-phase-commit-2pc > > > > > > > > >> >> < > > > > > > > > >> >> > > https://apacheignite.readme.io/docs/transactions#section- > > > > > > > > >> two-phase-commit-2pc > > > > > > > > >> >>> > > > > > > > > >> >> > > > > > > > > >> >> — > > > > > > > > >> >> Denis > > > > > > > > >> >> > > > > > > > > >> >>> On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan < > > > > > > > > >> [hidden email]> > > > > > > > > >> >> wrote: > > > > > > > > >> >>> > > > > > > > > >> >>> On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < > > > > > > > > >> >> [hidden email] > > > > > > > > >> >>>> wrote: > > > > > > > > >> >>> > > > > > > > > >> >>>> Igniters! What is the point of one phase > optimisation? > > > > > > > > >> >>>> > > > > > > > > >> >>> > > > > > > > > >> >>> Performance > > > > > > > > >> >> > > > > > > > > >> >> -- > > > > > > > > >> >> > > > > > > > > >> >> *Best Regards,* > > > > > > > > >> >> > > > > > > > > >> >> *Kuznetsov Aleksey* > > > > > > > > >> >> > > > > > > > > >> > -- > > > > > > > > >> > > > > > > > > > >> > *Best Regards,* > > > > > > > > >> > > > > > > > > > >> > *Kuznetsov Aleksey* > > > > > > > > >> > > > > > > > > > -- > > > > > > > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > > > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > -- > > > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > > > > -- > > > > > > > > > > *Best Regards,* > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > -- > > > > > > *Best Regards,* > > > > > > *Kuznetsov Aleksey* > > > > > > -- > > *Best Regards,* > > *Kuznetsov Aleksey* > -- Best Regards, Anton Churaev |
|
|
would like to add 1 phase optimisation isn't gonna work if write through is
enabled. ср, 5 апр. 2017 г. в 15:23, Антон Чураев <[hidden email]>: > Maybe it will be useful to update the documentation? > > 2017-04-05 15:15 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email]>: > > > Thank you for help! > > > > ср, 5 апр. 2017 г. в 15:14, Alexey Goncharuk <[hidden email] > >: > > > > > This optimization does not work when near cache is enabled because we > > need > > > the same ordering on near nodes. You should see the expected number of > > > messages with near cache disabled. > > > > > > 2017-04-05 15:09 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email] > >: > > > > > > > yes > > > > > > > > ср, 5 апр. 2017 г. в 15:07, Alexey Goncharuk < > > [hidden email] > > > >: > > > > > > > > > Do you have a near cache enabled? > > > > > > > > > > 2017-04-05 15:00 GMT+03:00 ALEKSEY KUZNETSOV < > > [hidden email] > > > >: > > > > > > > > > > > The test shows as follows: > > > > > > assertMessageCount(GridNearTxPrepareRequest.class, > 1); > > > > > > assertMessageCount(GridDhtTxPrepareRequest.class, 1); > > > > > > assertMessageCount(GridDhtTxPrepareResponse.class, > 1); > > > > > > assertMessageCount(GridNearTxPrepareResponse.class, > > 1); > > > > > > assertMessageCount(GridNearTxFinishRequest.class, 1); > > > > > > assertMessageCount(GridDhtTxFinishRequest.class, 0); > > > > > > assertMessageCount(GridNearTxFinishResponse.class, > 1); > > > > > > > > > > > > ср, 5 апр. 2017 г. в 14:53, Alexey Goncharuk < > > > > [hidden email] > > > > > >: > > > > > > > > > > > > > Aleksey, > > > > > > > > > > > > > > Can you elaborate on which of the extra messages you observe? > > > > > > > > > > > > > > --AG > > > > > > > > > > > > > > 2017-04-04 14:17 GMT+03:00 ALEKSEY KUZNETSOV < > > > > [hidden email] > > > > > >: > > > > > > > > > > > > > > > any thoughts on one phase commit realization ? > > > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 19:35, ALEKSEY KUZNETSOV < > > > > > > [hidden email] > > > > > > > >: > > > > > > > > > > > > > > > > > I've attached test that prints messages exchange . Which > > shows > > > us > > > > > > that > > > > > > > > > there are more messages then you declared in article. > > Perhaps, > > > > > > > > > implementation has changed. > > > > > > > > > I created it on base of IgniteOnePhaseCommitNearSelfTest > > > > > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 19:03, Dmitriy Setrakyan < > > > > > [hidden email] > > > > > > >: > > > > > > > > > > > > > > > > > > Aleksey, > > > > > > > > > > > > > > > > > > The blog describes the 1-phase commit at a high level, but > I > > am > > > > > still > > > > > > > > > curious about the differences you found. Can you share them > > > here? > > > > > > > > > > > > > > > > > > D. > > > > > > > > > > > > > > > > > > On Mon, Apr 3, 2017 at 2:11 AM, ALEKSEY KUZNETSOV < > > > > > > > > > [hidden email]> > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > Regarding IgniteOnePhaseCommitNearSelfTest test , > ignite's > > > one > > > > > > phase > > > > > > > > > > optimisation works not as you said. > > > > > > > > > > I attached picture of message exchange. There are partial > > > > prepare > > > > > > > phase > > > > > > > > > > exists, along with finish phase. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 10:55, Christos Erotocritou < > > > > > > > > [hidden email] > > > > > > > > > >: > > > > > > > > > > > > > > > > > > > >> As far as I know a partition is always allocated to a > > > specific > > > > > > node > > > > > > > > and > > > > > > > > > >> does not span nodes. Ignite has default 1024 partitions > on > > > > start > > > > > > > that > > > > > > > > > are > > > > > > > > > >> split equally across nodes. > > > > > > > > > >> > > > > > > > > > >> > On 3 Apr 2017, at 08:10, ALEKSEY KUZNETSOV < > > > > > > > > [hidden email]> > > > > > > > > > >> wrote: > > > > > > > > > >> > > > > > > > > > > >> > in ur blog u texted belonging to the same partition is > > > > > nessesary > > > > > > > > for 1 > > > > > > > > > >> > phase commit. But its not guarantee belonging to the > > same > > > > > node. > > > > > > > > > >> Partition > > > > > > > > > >> > may span many nodes > > > > > > > > > >> > > > > > > > > > > >> > вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV < > > > > > > > > [hidden email] > > > > > > > > > >: > > > > > > > > > >> > > > > > > > > > > >> >> thank u ! > > > > > > > > > >> >> > > > > > > > > > >> >> пт, 31 Мар 2017 г., 21:06 Denis Magda < > > [hidden email] > > > >: > > > > > > > > > >> >> > > > > > > > > > >> >> Here is a good blog post about 1phase commit impl in > > > Ignite > > > > > and > > > > > > > its > > > > > > > > > >> >> advantages: > > > > > > > > > >> >> > > > > > > > > > >> >> http://gridgain.blogspot.com/ > > > > 2014/09/one-phase-commit-fast- > > > > > > > > > >> transactions-for.html > > > > > > > > > >> >> < > > > > > > > > > >> >> http://gridgain.blogspot.com/ > > > > 2014/09/one-phase-commit-fast- > > > > > > > > > >> transactions-for.html > > > > > > > > > >> >>> > > > > > > > > > >> >> > > > > > > > > > >> >> Took a reference to it from there: > > > > > > > > > >> >> > > > > > > > > > >> >> > > > https://apacheignite.readme.io/docs/transactions#section- > > > > > > > > > >> two-phase-commit-2pc > > > > > > > > > >> >> < > > > > > > > > > >> >> > > > https://apacheignite.readme.io/docs/transactions#section- > > > > > > > > > >> two-phase-commit-2pc > > > > > > > > > >> >>> > > > > > > > > > >> >> > > > > > > > > > >> >> — > > > > > > > > > >> >> Denis > > > > > > > > > >> >> > > > > > > > > > >> >>> On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan < > > > > > > > > > >> [hidden email]> > > > > > > > > > >> >> wrote: > > > > > > > > > >> >>> > > > > > > > > > >> >>> On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < > > > > > > > > > >> >> [hidden email] > > > > > > > > > >> >>>> wrote: > > > > > > > > > >> >>> > > > > > > > > > >> >>>> Igniters! What is the point of one phase > > optimisation? > > > > > > > > > >> >>>> > > > > > > > > > >> >>> > > > > > > > > > >> >>> Performance > > > > > > > > > >> >> > > > > > > > > > >> >> -- > > > > > > > > > >> >> > > > > > > > > > >> >> *Best Regards,* > > > > > > > > > >> >> > > > > > > > > > >> >> *Kuznetsov Aleksey* > > > > > > > > > >> >> > > > > > > > > > >> > -- > > > > > > > > > >> > > > > > > > > > > >> > *Best Regards,* > > > > > > > > > >> > > > > > > > > > > >> > *Kuznetsov Aleksey* > > > > > > > > > >> > > > > > > > > > > -- > > > > > > > > > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > > > > > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > > > -- > > > > > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > > > > > > > -- > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > -- > > > > > > > > *Best Regards,* > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > -- > > > > *Best Regards,* > > > > *Kuznetsov Aleksey* > > > > > > -- > > Best Regards, Anton Churaev > *Best Regards,* *Kuznetsov Aleksey* |
|
|
In reply to this post by Alexey Goncharuk
Hi! Your talking about node ordering?
What is the point of the ordering? How is it implemented? Thanks for the answering in advance! ср, 5 апр. 2017 г. в 15:14, Alexey Goncharuk <[hidden email]>: > This optimization does not work when near cache is enabled because we need > the same ordering on near nodes. You should see the expected number of > messages with near cache disabled. > > 2017-04-05 15:09 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email]>: > > > yes > > > > ср, 5 апр. 2017 г. в 15:07, Alexey Goncharuk <[hidden email] > >: > > > > > Do you have a near cache enabled? > > > > > > 2017-04-05 15:00 GMT+03:00 ALEKSEY KUZNETSOV <[hidden email] > >: > > > > > > > The test shows as follows: > > > > assertMessageCount(GridNearTxPrepareRequest.class, 1); > > > > assertMessageCount(GridDhtTxPrepareRequest.class, 1); > > > > assertMessageCount(GridDhtTxPrepareResponse.class, 1); > > > > assertMessageCount(GridNearTxPrepareResponse.class, 1); > > > > assertMessageCount(GridNearTxFinishRequest.class, 1); > > > > assertMessageCount(GridDhtTxFinishRequest.class, 0); > > > > assertMessageCount(GridNearTxFinishResponse.class, 1); > > > > > > > > ср, 5 апр. 2017 г. в 14:53, Alexey Goncharuk < > > [hidden email] > > > >: > > > > > > > > > Aleksey, > > > > > > > > > > Can you elaborate on which of the extra messages you observe? > > > > > > > > > > --AG > > > > > > > > > > 2017-04-04 14:17 GMT+03:00 ALEKSEY KUZNETSOV < > > [hidden email] > > > >: > > > > > > > > > > > any thoughts on one phase commit realization ? > > > > > > > > > > > > пн, 3 апр. 2017 г. в 19:35, ALEKSEY KUZNETSOV < > > > > [hidden email] > > > > > >: > > > > > > > > > > > > > I've attached test that prints messages exchange . Which shows > us > > > > that > > > > > > > there are more messages then you declared in article. Perhaps, > > > > > > > implementation has changed. > > > > > > > I created it on base of IgniteOnePhaseCommitNearSelfTest > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 19:03, Dmitriy Setrakyan < > > > [hidden email] > > > > >: > > > > > > > > > > > > > > Aleksey, > > > > > > > > > > > > > > The blog describes the 1-phase commit at a high level, but I am > > > still > > > > > > > curious about the differences you found. Can you share them > here? > > > > > > > > > > > > > > D. > > > > > > > > > > > > > > On Mon, Apr 3, 2017 at 2:11 AM, ALEKSEY KUZNETSOV < > > > > > > > [hidden email]> > > > > > > > wrote: > > > > > > > > > > > > > > > Regarding IgniteOnePhaseCommitNearSelfTest test , ignite's > one > > > > phase > > > > > > > > optimisation works not as you said. > > > > > > > > I attached picture of message exchange. There are partial > > prepare > > > > > phase > > > > > > > > exists, along with finish phase. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > пн, 3 апр. 2017 г. в 10:55, Christos Erotocritou < > > > > > > [hidden email] > > > > > > > >: > > > > > > > > > > > > > > > >> As far as I know a partition is always allocated to a > specific > > > > node > > > > > > and > > > > > > > >> does not span nodes. Ignite has default 1024 partitions on > > start > > > > > that > > > > > > > are > > > > > > > >> split equally across nodes. > > > > > > > >> > > > > > > > >> > On 3 Apr 2017, at 08:10, ALEKSEY KUZNETSOV < > > > > > > [hidden email]> > > > > > > > >> wrote: > > > > > > > >> > > > > > > > > >> > in ur blog u texted belonging to the same partition is > > > nessesary > > > > > > for 1 > > > > > > > >> > phase commit. But its not guarantee belonging to the same > > > node. > > > > > > > >> Partition > > > > > > > >> > may span many nodes > > > > > > > >> > > > > > > > > >> > вс, 2 Апр 2017 г., 13:46 ALEKSEY KUZNETSOV < > > > > > > [hidden email] > > > > > > > >: > > > > > > > >> > > > > > > > > >> >> thank u ! > > > > > > > >> >> > > > > > > > >> >> пт, 31 Мар 2017 г., 21:06 Denis Magda <[hidden email] > >: > > > > > > > >> >> > > > > > > > >> >> Here is a good blog post about 1phase commit impl in > Ignite > > > and > > > > > its > > > > > > > >> >> advantages: > > > > > > > >> >> > > > > > > > >> >> http://gridgain.blogspot.com/ > > 2014/09/one-phase-commit-fast- > > > > > > > >> transactions-for.html > > > > > > > >> >> < > > > > > > > >> >> http://gridgain.blogspot.com/ > > 2014/09/one-phase-commit-fast- > > > > > > > >> transactions-for.html > > > > > > > >> >>> > > > > > > > >> >> > > > > > > > >> >> Took a reference to it from there: > > > > > > > >> >> > > > > > > > >> >> > https://apacheignite.readme.io/docs/transactions#section- > > > > > > > >> two-phase-commit-2pc > > > > > > > >> >> < > > > > > > > >> >> > https://apacheignite.readme.io/docs/transactions#section- > > > > > > > >> two-phase-commit-2pc > > > > > > > >> >>> > > > > > > > >> >> > > > > > > > >> >> — > > > > > > > >> >> Denis > > > > > > > >> >> > > > > > > > >> >>> On Mar 31, 2017, at 12:27 PM, Dmitriy Setrakyan < > > > > > > > >> [hidden email]> > > > > > > > >> >> wrote: > > > > > > > >> >>> > > > > > > > >> >>> On Fri, Mar 31, 2017 at 9:25 AM, ALEKSEY KUZNETSOV < > > > > > > > >> >> [hidden email] > > > > > > > >> >>>> wrote: > > > > > > > >> >>> > > > > > > > >> >>>> Igniters! What is the point of one phase optimisation? > > > > > > > >> >>>> > > > > > > > >> >>> > > > > > > > >> >>> Performance > > > > > > > >> >> > > > > > > > >> >> -- > > > > > > > >> >> > > > > > > > >> >> *Best Regards,* > > > > > > > >> >> > > > > > > > >> >> *Kuznetsov Aleksey* > > > > > > > >> >> > > > > > > > >> > -- > > > > > > > >> > > > > > > > > >> > *Best Regards,* > > > > > > > >> > > > > > > > > >> > *Kuznetsov Aleksey* > > > > > > > >> > > > > > > > > -- > > > > > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > > > > > > > > -- > > > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > -- > > > > > > > > > > > > *Best Regards,* > > > > > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > > > > > > > -- > > > > > > > > *Best Regards,* > > > > > > > > *Kuznetsov Aleksey* > > > > > > > > > -- > > > > *Best Regards,* > > > > *Kuznetsov Aleksey* > > > *Best Regards,* *Kuznetsov Aleksey* |
«
Return to Apache Ignite Developers - Legacy Mail Archive
|
1 view|%1 views
| Free forum by Nabble | Edit this page |