New behavior cause data corruption in v11
|
View:
New views
20 Messages
—
Rating Filter:
Alert me
|
| < Prev | 1 - 2 - 3 - 4 | Next > |
 |
New behavior cause data corruption in v11
by Alan Chan
::
Rate this Message:
We all know that data corruption happens pretty often in v11. Tracking
down the reproducible steps is one thing and finding its cause is another. For the past week, we have been tracking hard on a reproducible data corruption and eventually found its cause. 4D now supports mutlipe CPUs on Trigger. Different invoking table maintain its own context in server. So far, it works great until same invoking table query a record that has the same Record Number of a freshly deleted record. In such case, you will get wrong query result. Picture the followings. 1) Create a new recod with unique primary key in One Table A. 2) Create a new record in Many Table B. 3) Delete the newly created record in Many Table B 4) Delete the newly created record in On Table A 5) Create a new record with new unique primary key in One Table A. Mark down the record number (this new record will be given the same record number as the deleted record in step 4) by 4D and this is normal). 6) Create a new record in Many Table B. This will cause the problem. In the trigger on step 6), Query([Table A];[Table A]PrimaryKey=[Table B]ForeignKey) will get the record deleted in step 4. C_string(20;$OldPrimary;$NewPrimary;$NewForeignKey) C_boolean($matchPrimaryKey) $OldPrimary:=[Table A]PrimaryKey `in debugger in trigger, you would see record content of old deteled record of Table A. It has the same record number of the newly created record but still have the old PrimaryKey $NewForeignKey:=Table B]ForeignKey Query([Table A];[Table A]PrimaryKey=[Table B]ForeignKey) `this query doesn't work if((records in selection([Table A])=1) $matchPrimarykey:=Table A]PrimaryKe=Table B]ForeignKey `this would return false although it should be true (pls note that the record is not corrupted at this stage Save record([Table A]) `Saving this record would corrupt the data and the saved record will have the old primarykey instead of new primary key. end if Please note that it would happen only 1) the invoking table is the same. In this case, it's Table B. 2) the invoking table did call Table A in trigger (Table A content would remain in trigger memory) 3) the newly create record of Table A called in the Table B trigger happens to have same record number of deteled record and already remain in the trigger memory. Alan Chan ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: New behavior cause data corruption in v11
by Chip Scheide
::
Rate this Message:
is all this happening inside a transaction??
it is the only way that it makes sense that a deleted record is found by a query On Wed, 11 Nov 2009 06:13:47 +0800, Alan Chan wrote: > We all know that data corruption happens pretty often in v11. Tracking > down the reproducible steps is one thing and finding its cause is another. > For the past week, we have been tracking hard on a reproducible data > corruption and eventually found its cause. > > 4D now supports mutlipe CPUs on Trigger. Different invoking table maintain > its own context in server. So far, it works great until same invoking > table query a record that has the same Record Number of a freshly deleted > record. In such case, you will get wrong query result. Picture the > followings. > > 1) Create a new recod with unique primary key in One Table A. > 2) Create a new record in Many Table B. > 3) Delete the newly created record in Many Table B > 4) Delete the newly created record in On Table A > 5) Create a new record with new unique primary key in One Table A. Mark > down the record number (this new record will be given the same record > number as the deleted record in step 4) by 4D and this is normal). > 6) Create a new record in Many Table B. This will cause the problem. > > In the trigger on step 6), Query([Table A];[Table A]PrimaryKey=[Table > B]ForeignKey) will get the record deleted in step 4. > > > C_string(20;$OldPrimary;$NewPrimary;$NewForeignKey) > C_boolean($matchPrimaryKey) > > $OldPrimary:=[Table A]PrimaryKey `in debugger in trigger, you would see > record content of old deteled record of Table A. It has the same record > number of the newly created record but still have the old PrimaryKey > $NewForeignKey:=Table B]ForeignKey > Query([Table A];[Table A]PrimaryKey=[Table B]ForeignKey) `this query > doesn't work > if((records in selection([Table A])=1) > $matchPrimarykey:=Table A]PrimaryKe=Table B]ForeignKey `this would return > false although it should be true (pls note that the record is not > corrupted at this stage > Save record([Table A]) `Saving this record would corrupt the data and the > saved record will have the old primarykey instead of new primary key. > end if > > Please note that it would happen only > > 1) the invoking table is the same. In this case, it's Table B. > 2) the invoking table did call Table A in trigger (Table A content would > remain in trigger memory) > 3) the newly create record of Table A called in the Table B trigger > happens to have same record number of deteled record and already remain in > the trigger memory. > > Alan Chan > > ********************************************************************** > Get the speed and power of 4D v11 SQL > before upgrade prices increase - http://www.4d.com > > > 4D Internet Users Group (4D iNUG) > FAQ: http://lists.4d.com/faqnug.html > Archive: http://lists.4D.com/archives.html > Options: https://lists.4d.com/mailman/options/4d_tech > Unsub: mailto:4D_Tech-Unsubscribe@... > ********************************************************************** > Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
|
|
|
|
|
|
Re: New behavior cause data corruption in v11
by Peter Jakobsson-2
::
Rate this Message:
On 10 Nov 2009, at 23:33, Chip Scheide wrote: > it is the only way that it makes sense that a deleted record is found > by a query In that case, the query result and the record properties (when accessed by the language) should still agree, but in Alan's account they don't appear to. if((records in selection([Table A])=1) <-- this was returning true, which is consistent with the query finding the new record ... but . . . $matchPrimarykey:=Table A]PrimaryKe=Table B]ForeignKey <-- this was returning false which is consistent with the language accessing the old record Maybe the query is finding the record but not loading it and the old one remains in cache. Just my interpretation - Thomas says he can't reproduce it so maybe it's not what it seems. Also, Alan, what do you mean by . . . > Query([Table A];[Table A]PrimaryKey=[Table B]ForeignKey) `this query > doesn't work . . . i.e. [A] - does the query not return any records when it should or [B] - it returns [records in selection([Table A])=1)] but it's the wrong record (this was my interpretation of your description) Regards Peter ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
|
|
|
Re: New behavior cause data corruption in v11
by Alan Chan
::
Rate this Message:
Yes, it is inside transaction and trigger.
Please note that the deleted record is in a comleted transaction and was found in another transaction inside the trigger. Alan Chan 4D iNug Technical <4d_tech@...> writes: >is all this happening inside a transaction?? > >it is the only way that it makes sense that a deleted record is found >by a query > > >On Wed, 11 Nov 2009 06:13:47 +0800, Alan Chan wrote: >> We all know that data corruption happens pretty often in v11. Tracking >> down the reproducible steps is one thing and finding its cause is >another. >> For the past week, we have been tracking hard on a reproducible data >> corruption and eventually found its cause. >> >> 4D now supports mutlipe CPUs on Trigger. Different invoking table >maintain >> its own context in server. So far, it works great until same invoking >> table query a record that has the same Record Number of a freshly >deleted >> record. In such case, you will get wrong query result. Picture the >> followings. >> >> 1) Create a new recod with unique primary key in One Table A. >> 2) Create a new record in Many Table B. >> 3) Delete the newly created record in Many Table B >> 4) Delete the newly created record in On Table A >> 5) Create a new record with new unique primary key in One Table A. Mark >> down the record number (this new record will be given the same record >> number as the deleted record in step 4) by 4D and this is normal). >> 6) Create a new record in Many Table B. This will cause the problem. >> >> In the trigger on step 6), Query([Table A];[Table A]PrimaryKey=[Table >> B]ForeignKey) will get the record deleted in step 4. >> >> >> C_string(20;$OldPrimary;$NewPrimary;$NewForeignKey) >> C_boolean($matchPrimaryKey) >> >> $OldPrimary:=[Table A]PrimaryKey `in debugger in trigger, you would see >> record content of old deteled record of Table A. It has the same record >> number of the newly created record but still have the old PrimaryKey >> $NewForeignKey:=Table B]ForeignKey >> Query([Table A];[Table A]PrimaryKey=[Table B]ForeignKey) `this query >> doesn't work >> if((records in selection([Table A])=1) >> $matchPrimarykey:=Table A]PrimaryKe=Table B]ForeignKey `this would >return >> false although it should be true (pls note that the record is not >> corrupted at this stage >> Save record([Table A]) `Saving this record would corrupt the data and >the >> saved record will have the old primarykey instead of new primary key. >> end if >> >> Please note that it would happen only >> >> 1) the invoking table is the same. In this case, it's Table B. >> 2) the invoking table did call Table A in trigger (Table A content would >> remain in trigger memory) >> 3) the newly create record of Table A called in the Table B trigger >> happens to have same record number of deteled record and already remain >in >> the trigger memory. >> >> Alan Chan >> >> ********************************************************************** >> Get the speed and power of 4D v11 SQL >> before upgrade prices increase - http://www.4d.com >> >> >> 4D Internet Users Group (4D iNUG) >> FAQ: http://lists.4d.com/faqnug.html >> Archive: http://lists.4D.com/archives.html >> Options: https://lists.4d.com/mailman/options/4d_tech >> Unsub: mailto:4D_Tech-Unsubscribe@... >> ********************************************************************** >> >********************************************************************** >Get the speed and power of 4D v11 SQL >before upgrade prices increase - http://www.4d.com > > >4D Internet Users Group (4D iNUG) >FAQ: http://lists.4d.com/faqnug.html >Archive: http://lists.4D.com/archives.html >Options: https://lists.4d.com/mailman/options/4d_tech >Unsub: mailto:4D_Tech-Unsubscribe@... >********************************************************************** ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: New behavior cause data corruption in v11
by Alan Chan
::
Rate this Message:
4D iNug Technical <4d_tech@...> writes:
>Many to one relation from fkey to prikey, Automatic relation turned on. Like many old timer 4D developers, I never turned on Automatic relation on data entry except in some special reporting. Alan Chan ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: New behavior cause data corruption in v11
by Alan Chan
::
Rate this Message:
4D iNug Technical <4d_tech@...> writes:
>Hi Alan, > >A couple oversights in my last post: > >1. I meant to let you know that I have the "On Saving New Record" trigger >on >for Table_1. Exactly as I did. > >2. The other kind of corruption you mean (saving an old prikey instead of >the newly created prikey) is also not occurring. Because the QUERY works >correctly the right prikey value is saved, so there's none of the non-MSC >find-able corruption. If your query works, then no corruption should be happened. The most important part is the "Record Number". Query will always work if the newly created record has different record number from the deleted record. As you know, 4D will use old Record Number. The same record number is the key. That's why we spent more than a week of 3 developers' time to track this down. Same record number aside, the deleted record must be already in the memory of the trigger. That means, the invoking table should have access the "pre-deleted record" in the trigger. Then this pre-deleted record would remain in its trigger memory even after it's completed. Once the record is deleted and new record created, use the previous invoking table to access the new record while the old deleted record was still in Trigger's memory, if you put trace in beginning of the trigger, you will see the old deleted record was still there with old Primiary key but same record as the new one. Do query at this time in the trigger, the query won't work. Please note that if the invoking table access other primary record before they access the newly created record, the problem won't happen. Alan Chan > > ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: New behavior cause data corruption in v11
by Alan Chan
::
Rate this Message:
Jesse,
It's interesting. As far as I know, triggers from different tables could be executed across CPUs. This information was obtained from 4D official. Please correct me, if all triggers are still running on one single CPU and only SQL access more than one CPU like the initial stage of v11, converting to v11 might not be top priority for us. It also meant that having dual quad core or even more powerful machine won't do any good unless we switch everything to SQL which is unlikely. I heard that 4D v11.4 has changed this from the first version of v11, but again, different 4D official would provide different information. Alan Chan 4D iNug Technical <4d_tech@...> writes: >Just to clarify, this is not a correct statement. Triggers run in the >cooperative process. All 4D code runs in the cooperative thread >regardless >of where it is run (project method, trigger, form method, ...). > >At this year's Summit, Josh Fletcher covered the topic of threads and the >new threading model in his preclass "4D Server v11 SQL: Understanding >Performance". So if you attended the Summit this year, I would recommend >looking at his session notes, there is a lot of great detail in there. ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: New behavior cause data corruption in v11
by David Dancy
::
Rate this Message:
Alan,
I think I've got this straight in my head. Let me see if I can get it out in words: At this point (AFAIK) triggers only run on one CPU/core on the server. The only change to the triggers mechanism in v11 from v2004 is that a trigger now no longer blocks access to the whole database. It only blocks access to the table it's running in, so you can have multiple triggers running at the same time, provided that they are for different tables. The triggers are all 4D language, which means that they execute in the cooperative thread maintained by the server on the client's behalf. Each client has a cooperative thread and a pre-emptive thread. The cooperative thread runs inside the 4D process, which means it's tied to one CPU/core. The OS may shift it to a different CPU/core, but the point is it's only running on one CPU/core at a time, and all the cooperative threads together run in the same CPU/core because they're all in the same process. The pre-emptive thread (I believe) can attach to any CPU/core. But since all triggers run in the cooperative thread for their client, although you may have multiple triggers going at once, they are all on the same CPU/core. If a trigger issues a query, that is handled by the server's DB engine, using as many threads as required. Since the DB engine threads are all pre-emptive, they can attach to any CPU/core as the OS sees fit. I think the trigger's cooperative thread will block until the results come back from the DB engine. Corrections/clarifications welcome. HTH David Dancy Sydney, Australia 2009/11/11 Alan Chan <alan@...>: > Jesse, > > It's interesting. As far as I know, triggers from different tables could > be executed across CPUs. This information was obtained from 4D official. > Please correct me, if all triggers are still running on one single CPU and > only SQL access more than one CPU like the initial stage of v11, > converting to v11 might not be top priority for us. It also meant that > having dual quad core or even more powerful machine won't do any good > unless we switch everything to SQL which is unlikely. > > I heard that 4D v11.4 has changed this from the first version of v11, but > again, different 4D official would provide different information. > > Alan Chan > > 4D iNug Technical <4d_tech@...> writes: >>Just to clarify, this is not a correct statement. Triggers run in the >>cooperative process. All 4D code runs in the cooperative thread >>regardless >>of where it is run (project method, trigger, form method, ...). >> >>At this year's Summit, Josh Fletcher covered the topic of threads and the >>new threading model in his preclass "4D Server v11 SQL: Understanding >>Performance". So if you attended the Summit this year, I would recommend >>looking at his session notes, there is a lot of great detail in there. > Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: New behavior cause data corruption in v11
by Alan Chan
::
Rate this Message:
Hi Thomas,
To reproduce the issue and minimize complication, you might want to disable automatic relation and use simple query (not even using Relate One). Also use simple Transaction and Trigger (no cascade trigger/multi-level transaction). It means every step is done in one single transaction/trigger and completed. >1. create some records in Table_1 >2. create some matching records in Table_2 (1 many record per one record) >3. delete one of the early records in Table_2 Did you delete exactly the record you just created in step 3)? Deleting other many records that ralate to other One table would break the test. > >4. delete the corresponding "one" record in Table_1 >5. create a new record in Table_1 >6. create a new record in Table_2 On step 6, if you debug in trigger, you will see deleted record of Table_1 on debug window. This deleted record should contain old primarykey and same Record Number (you could test its record number in the debug window) before you do the query. I hope this would help. Alan Chan ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: New behavior cause data corruption in v11
by Peter Schumacher-2
::
Rate this Message:
I'm missing here one sentence....
Am 10.11.2009 um 17:16 schrieb Jesse Pina: > At this year's Summit, Josh Fletcher covered the topic of threads > and the > new threading model in his preclass "4D Server v11 SQL: Understanding > Performance". So if you attended the Summit this year, I would > recommend > looking at his session notes, there is a lot of great detail in there. Greetings, [4D-Consulting.com]eK, Wiesbaden (Germany) Peter Schumacher -------------------------------------------------------- Web: http://www.4D-Consulting.com/ FreeCall: 0800-434 636 7 Tel.: +49-611-9406.850 - Fax: +49-611-9406.744 4D-Consulting.com eK - Scharnhorststr. 36 - 65195 Wiesbaden Member of the German Developer Network http://www.die4dwerkstatt.de ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: Does v11 support multiple CPUs for converted application?
by Jeffrey Kain
::
Rate this Message:
4D has always been really clear that anything running 4D code would not be
preemptively threaded. Originally, only queries were going to be preemptively threaded, but this changed as we got close to the release of 4D Server to include all database operations that do not involve the direct execution of 4D code. So, triggers, stored procedures, and code running in the web server can't run on multiple CPUs. All other database operations that are made from clients or other outside connections can all be preemptively threaded. And yes, it makes a really big difference in performance in the real world. It's not imaginary and it's not a trick. Of course, if your system is 100% trigger/web server/stored procedure based, you won't get this benefit, but the engine itself is also a lot faster than v2004. JEff On 11/10/09 9:35 PM, "Alan Chan" <alan@...> wrote: > Thanks for the detailed explanation. According to your understanding, how > 4D did its demo on 24 cores and make a big deal out of its scalability. ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Does v11 support multiple CPUs for converted application?
by Alan Chan
::
Rate this Message:
Hi David,
Thanks for the detailed explanation. According to your understanding, how 4D did its demo on 24 cores and make a big deal out of its scalability. While some did challange how 4d did its 500 users demo, 4D insists that they simulate real world practices like entering new orders and creates reporting. Are they all done in SQL and not 4D languages like trigger/transaction? How could real world data entry without trigger/transaction for any mission critical application. If all new entries are done in Trigger, 500 users are doing entries on "one" cpu??? I think they could take lunch instead of coffee break after hitting the enter key for a single order entry:-) I don't think that 4D assume their developers rewritting their mature applications in SQL yet without considering other SQL engines. Any clarification would be greatly appreciated. Alan Chan 4D iNug Technical <4d_tech@...> writes: >Alan, > >I think I've got this straight in my head. Let me see if I can get it >out in words: > >At this point (AFAIK) triggers only run on one CPU/core on the server. > >The only change to the triggers mechanism in v11 from v2004 is that a >trigger now no longer blocks access to the whole database. It only >blocks access to the table it's running in, so you can have multiple >triggers running at the same time, provided that they are for >different tables. > >The triggers are all 4D language, which means that they execute in the >cooperative thread maintained by the server on the client's behalf. >Each client has a cooperative thread and a pre-emptive thread. The >cooperative thread runs inside the 4D process, which means it's tied >to one CPU/core. The OS may shift it to a different CPU/core, but the >point is it's only running on one CPU/core at a time, and all the >cooperative threads together run in the same CPU/core because they're >all in the same process. The pre-emptive thread (I believe) can attach >to any CPU/core. But since all triggers run in the cooperative thread >for their client, although you may have multiple triggers going at >once, they are all on the same CPU/core. > >If a trigger issues a query, that is handled by the server's DB >engine, using as many threads as required. Since the DB engine threads >are all pre-emptive, they can attach to any CPU/core as the OS sees >fit. I think the trigger's cooperative thread will block until the >results come back from the DB engine. > >Corrections/clarifications welcome. > >HTH > >David Dancy >Sydney, Australia > > > >2009/11/11 Alan Chan <alan@...>: >> Jesse, >> >> It's interesting. As far as I know, triggers from different tables could >> be executed across CPUs. This information was obtained from 4D official. >> Please correct me, if all triggers are still running on one single CPU >and >> only SQL access more than one CPU like the initial stage of v11, >> converting to v11 might not be top priority for us. It also meant that >> having dual quad core or even more powerful machine won't do any good >> unless we switch everything to SQL which is unlikely. >> >> I heard that 4D v11.4 has changed this from the first version of v11, >but >> again, different 4D official would provide different information. >> >> Alan Chan >> >> 4D iNug Technical <4d_tech@...> writes: >>>Just to clarify, this is not a correct statement. Triggers run in the >>>cooperative process. All 4D code runs in the cooperative thread >>>regardless >>>of where it is run (project method, trigger, form method, ...). >>> >>>At this year's Summit, Josh Fletcher covered the topic of threads and >the >>>new threading model in his preclass "4D Server v11 SQL: Understanding >>>Performance". So if you attended the Summit this year, I would >recommend >>>looking at his session notes, there is a lot of great detail in there. >> >********************************************************************** >Get the speed and power of 4D v11 SQL >before upgrade prices increase - http://www.4d.com > > >4D Internet Users Group (4D iNUG) >FAQ: http://lists.4d.com/faqnug.html >Archive: http://lists.4D.com/archives.html >Options: https://lists.4d.com/mailman/options/4d_tech >Unsub: mailto:4D_Tech-Unsubscribe@... >********************************************************************** > ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: Does v11 support multiple CPUs for converted application?
by MIYAKO
::
Rate this Message:
Hello Alan,
the article mentions that the 500 demo was an upgraded application, with no change to the code (i.e. no SQL). http://www.4d.com/jp/support/resources/benchmarks.html the developer who agreed to have his application used for that demo identified himself at Josh's session during the Summit pre class, so obviously its not like we made the whole thing all up. but more to the point, the question of SQL vs QUERY is not really an issue over performance, it has more to do with syntax constraint and the necessity of a current selection. miyako On 2009/11/11, at 11:35, Alan Chan wrote: > Hi David, > > Thanks for the detailed explanation. According to your > understanding, how > 4D did its demo on 24 cores and make a big deal out of its > scalability. > While some did challange how 4d did its 500 users demo, 4D insists > that > they simulate real world practices like entering new orders and > creates > reporting. Are they all done in SQL and not 4D languages like > trigger/transaction? How could real world data entry without > trigger/transaction for any mission critical application. If all new > entries are done in Trigger, 500 users are doing entries on "one" > cpu??? I > think they could take lunch instead of coffee break after hitting the > enter key for a single order entry:-) I don't think that 4D assume > their > developers rewritting their mature applications in SQL yet without > considering other SQL engines. > > Any clarification would be greatly appreciated. > > Alan Chan Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: New behavior cause data corruption in v11
by Alan Chan
::
Rate this Message:
One more thing, it must be client/server. No such problem would be
happening in local mode. For a clean test, don't let our clients to login. In the beginning, we found randome data corruption and difficult to track down with multiple test engineers accessing the server. We tracked this down until I assigned only one single engineer to test and he knew exactly what he did and in what steps. Alan Chan 4D iNug Technical <4d_tech@...> writes: >Hi Thomas, > >To reproduce the issue and minimize complication, you might want to >disable automatic relation and use simple query (not even using Relate >One). Also use simple Transaction and Trigger (no cascade >trigger/multi-level transaction). It means every step is done in one >single transaction/trigger and completed. > >>1. create some records in Table_1 >>2. create some matching records in Table_2 (1 many record per one record) >>3. delete one of the early records in Table_2 >Did you delete exactly the record you just created in step 3)? Deleting >other many records that ralate to other One table would break the test. >> >>4. delete the corresponding "one" record in Table_1 >>5. create a new record in Table_1 >>6. create a new record in Table_2 >On step 6, if you debug in trigger, you will see deleted record of Table_1 >on debug window. This deleted record should contain old primarykey and >same Record Number (you could test its record number in the debug window) >before you do the query. > >I hope this would help. > >Alan Chan > >********************************************************************** >Get the speed and power of 4D v11 SQL >before upgrade prices increase - http://www.4d.com > > >4D Internet Users Group (4D iNUG) >FAQ: http://lists.4d.com/faqnug.html >Archive: http://lists.4D.com/archives.html >Options: https://lists.4d.com/mailman/options/4d_tech >Unsub: mailto:4D_Tech-Unsubscribe@... >********************************************************************** ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: Does v11 support multiple CPUs for converted application?
by Jeffrey Kain
::
Rate this Message:
On 11/10/09 10:21 PM, "Alan Chan" <alan@...> wrote:
> We never used Web server (it's interesting that you think we are using web > server anyway), I'm not sure why you would think that. I have no idea what your app does or doesn't do. I just was stating that 4D code running on the server, regardless of the context, doesn't get preemptively threaded by 4D. > This is the statement that confuse me. Since our discussion has already > covered 4D codes, SQL and Query, what would this other operations from > client be? I assume SAVE RECORD, SELECTION TO ARRAY is 4D code. SELECTION TO ARRAY and SAVE RECORD (without a trigger) do not require 4D code to be executed on the server. I'm not sure I follow your logic. Jeff ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: Does v11 support multiple CPUs for converted application?
by Alan Chan
::
Rate this Message:
We never used Web server (it's interesting that you think we are using web
server anyway), our application are strict Client/Server mode that users do data entry and query reports like "what 4D claimed in their 500 users demo". You're correct that 4D has been really "clear" that 4D code would not be preemptively and only SQL and Query would be benefitual to multiple CPUs. However, like you said, 4D changes a bit that claimed all other operations (?) would be ultilizing multiple CPUs except 4D codes. This is the statement that confuse me. Since our discussion has already covered 4D codes, SQL and Query, what would this other operations from client be? I assume SAVE RECORD, SELECTION TO ARRAY is 4D code. I sincerely want to know what codes did this 500 users demo used without 4D code??? Alan Chan 4D iNug Technical <4d_tech@...> writes: >4D has always been really clear that anything running 4D code would not be >preemptively threaded. Originally, only queries were going to be >preemptively threaded, but this changed as we got close to the release of >4D >Server to include all database operations that do not involve the direct >execution of 4D code. > >So, triggers, stored procedures, and code running in the web server can't >run on multiple CPUs. All other database operations that are made from >clients or other outside connections can all be preemptively threaded. And >yes, it makes a really big difference in performance in the real world. >It's >not imaginary and it's not a trick. > >Of course, if your system is 100% trigger/web server/stored procedure >based, >you won't get this benefit, but the engine itself is also a lot faster >than >v2004. > >JEff > >On 11/10/09 9:35 PM, "Alan Chan" <alan@...> wrote: > >> Thanks for the detailed explanation. According to your understanding, >how >> 4D did its demo on 24 cores and make a big deal out of its scalability. > >********************************************************************** >Get the speed and power of 4D v11 SQL >before upgrade prices increase - http://www.4d.com > > >4D Internet Users Group (4D iNUG) >FAQ: http://lists.4d.com/faqnug.html >Archive: http://lists.4D.com/archives.html >Options: https://lists.4d.com/mailman/options/4d_tech >Unsub: mailto:4D_Tech-Unsubscribe@... >********************************************************************** ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
|
|
Re: Does v11 support multiple CPUs for converted application?
by Alan Chan
::
Rate this Message:
Hi Miyako,
>the developer who agreed to have his application used for that demo >identified himself at Josh's session during the Summit pre class, >so obviously its not like we made the whole thing all up. I've never have a single doubt if 4D made this up. It's because we truly believed what 4D claims - simulating 500 users of doing data entry and reporting without SQL involved. But now, I was told the same old version that 4D languages are in single CPU. Would this 500 user demo be involved SAVE RECORD for order entry and SELECTION TO ARRAY for reporting? If so, how could this be spread out 24 cores? That's why I was led to believe that all these could be supported for multiple CPUs. Doesn't it make sense? I sincerely want to know what codes did this 500 users demo used without 4D code. Then I might be learning from this non-4D codes concept to do order entries in our application and to take advantage of multiple CPUs. Alan Chan ********************************************************************** Get the speed and power of 4D v11 SQL before upgrade prices increase - http://www.4d.com 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4D.com/archives.html Options: https://lists.4d.com/mailman/options/4d_tech Unsub: mailto:4D_Tech-Unsubscribe@... ********************************************************************** |
| < Prev | 1 - 2 - 3 - 4 | Next > |
| Free embeddable forum powered by Nabble | Forum Help |
