You are viewing a plain text version of this content. The canonical link for it is here.
Posted to derby-dev@db.apache.org by "Knut Anders Hatlen (JIRA)" <ji...@apache.org> on 2008/04/15 15:23:05 UTC

[jira] Created: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Implement more load types for org.apache.derbyTesting.perf.clients.Runner
-------------------------------------------------------------------------

                 Key: DERBY-3619
                 URL: https://issues.apache.org/jira/browse/DERBY-3619
             Project: Derby
          Issue Type: Test
          Components: Performance
            Reporter: Knut Anders Hatlen
            Assignee: Knut Anders Hatlen
            Priority: Minor


I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).

For the following types of load, we don't have test clients in derbyTesting.jar:

  1) tests for insert performance
  2) tests for delete performance
  3) single-record select + single-record update on secondary index
  4) single-record select + single-record update on non-indexed column
  5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)

(3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Updated: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Knut Anders Hatlen (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Knut Anders Hatlen updated DERBY-3619:
--------------------------------------

    Attachment: lob.diff

The attached patch makes it possible to run the single-record select tests with a blob or clob column instead of a varchar column. "-load sr_select" uses VARCHAR(100), "-load sr_select_blob" uses BLOB(100) and "-load sr_select_clob" uses CLOB(100).

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>         Attachments: lob.diff
>
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Knut Anders Hatlen (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12598574#action_12598574 ] 

Knut Anders Hatlen commented on DERBY-3619:
-------------------------------------------

Committed revision 658572.

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>         Attachments: lob.diff, no-pk.diff
>
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Knut Anders Hatlen (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12607889#action_12607889 ] 

Knut Anders Hatlen commented on DERBY-3619:
-------------------------------------------

Committed revision 671462

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>         Attachments: bank-clients.diff, create-accounts.diff, lob.diff, no-pk.diff
>
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Updated: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Dag H. Wanvik (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Dag H. Wanvik updated DERBY-3619:
---------------------------------

    Derby Categories: [Performance]

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>             Fix For: 10.5.1.1
>
>         Attachments: bank-clients.diff, create-accounts.diff, lob.diff, no-pk.diff, runner.diff
>
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Kristian Waagan (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12608763#action_12608763 ] 

Kristian Waagan commented on DERBY-3619:
----------------------------------------

For what it's worth, I have tested the bank transaction client by using it as a benchmark to test the JDBC client side statement cache.
No problems discovered.

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>         Attachments: bank-clients.diff, create-accounts.diff, lob.diff, no-pk.diff, runner.diff
>
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Updated: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Knut Anders Hatlen (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Knut Anders Hatlen updated DERBY-3619:
--------------------------------------

    Attachment: create-accounts.diff

Here's the first increment on a test that can replace the TPC-B like test. This patch (create-accounts.diff) only adds code that creates and populates tables that can be used in the test. Will write test clients that perform operations on the tables later. I've tried to make the schema and the contents of the tables follow the TPC-B specification, but it would be great if someone else could take a look and verify that it is implemented correctly.

The code is located in the class org.apache.derbyTesting.perf.clients.BankAccountFiller, which I committed to the repository with revision 667586.

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>         Attachments: create-accounts.diff, lob.diff, no-pk.diff
>
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Updated: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Knut Anders Hatlen (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Knut Anders Hatlen updated DERBY-3619:
--------------------------------------

    Attachment: no-pk.diff

The attached patch (no-pk.diff) adds some more variants of the sr_select and sr_update load types:
  - sr_update can now operate on a CLOB or a BLOB column instead of a VARCHAR column, just like sr_select
  - sr_select and sr_update can now be told to pick out the row by a non-indexed integer column or a column with a non-unique index instead of the primary key column

I updated the help text printed by perf.clients.Runner to explain how to modify the load. If you're running sr_select or sr_update and add "-load_opts blob,nonIndexed" to the command line, the load is changed so that it retrieves/updates blobs instead of varchars and selects rows by the non-indexed column instead of the primary key column.

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>         Attachments: lob.diff, no-pk.diff
>
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Updated: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Knut Anders Hatlen (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Knut Anders Hatlen updated DERBY-3619:
--------------------------------------

    Attachment: bank-clients.diff

The attached patch (bank-clients.diff) implements the clients that perform bank transactions. I haven't wired them into the Runner class yet, so they can't be run from command line, but I have tested them using the API in the o.a.derbyTesting.perf.clients package. I'll wire them into Runner in a follow-up. I have tried to make the transaction profile follow the TPC-B spec as closely as possible. Please let me know if you spot something that doesn't match the spec.

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>         Attachments: bank-clients.diff, create-accounts.diff, lob.diff, no-pk.diff
>
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Knut Anders Hatlen (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12590894#action_12590894 ] 

Knut Anders Hatlen commented on DERBY-3619:
-------------------------------------------

Committed lob.diff with revision 650112.

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>         Attachments: lob.diff
>
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Knut Anders Hatlen (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12589558#action_12589558 ] 

Knut Anders Hatlen commented on DERBY-3619:
-------------------------------------------

The single-record tests that are currently checked in fetch or update strings with 100 characters. We could easily add an option to replace the strings with Clobs or Blobs. I'll try to do that and also see if I can squeeze in an extra test in the nightly runs.

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Closed: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Knut Anders Hatlen (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Knut Anders Hatlen closed DERBY-3619.
-------------------------------------

       Resolution: Fixed
    Fix Version/s: 10.5.0.0

I'm not planning to add more test clients for now, so I'm closing this issue.

Most of Olav's nightly performance regression tests (http://home.online.no/~olmsan/derby/perf/) now use org.apache.derbyTesting.perf.clients.Runner. The only exceptions are the insert and delete tests, as they don't really fit into the model used by Runner. For instance, the delete tests quickly run out of rows to delete and must stop, whereas Runner expects the test to go on for a certain amount of time. Similar issues with insert when the table is full.

(It doesn't look like the graphs on Olav's site have been updated since January, but I know the tests are still running each night. Will look into that.)

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>             Fix For: 10.5.0.0
>
>         Attachments: bank-clients.diff, create-accounts.diff, lob.diff, no-pk.diff, runner.diff
>
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Knut Anders Hatlen (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12591636#action_12591636 ] 

Knut Anders Hatlen commented on DERBY-3619:
-------------------------------------------

Now the CLOB test is part of the nightly regression suite and the results are published at http://home.online.no/~olmsan/derby/perf/. The test doesn't show the above mentioned regression (DERBY-3312), probably because the nightly regression tests run the clients on a different machine than the server. In fact, the tests show a ~50X speedup for the client/server test from 10.2.2.0 to the current trunk when running a single-threaded test.

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>         Attachments: lob.diff
>
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Kathey Marsden (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12589128#action_12589128 ] 

Kathey Marsden commented on DERBY-3619:
---------------------------------------

It would be good to have a performance sanity check for lobs. We have been regressing. See:
https://issues.apache.org/jira/browse/DERBY-3312


> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Updated: (DERBY-3619) Implement more load types for org.apache.derbyTesting.perf.clients.Runner

Posted by "Knut Anders Hatlen (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/DERBY-3619?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Knut Anders Hatlen updated DERBY-3619:
--------------------------------------

    Attachment: runner.diff

The attached patch wires the bank transaction test into the Runner class. Committed revision 671861.

> Implement more load types for org.apache.derbyTesting.perf.clients.Runner
> -------------------------------------------------------------------------
>
>                 Key: DERBY-3619
>                 URL: https://issues.apache.org/jira/browse/DERBY-3619
>             Project: Derby
>          Issue Type: Test
>          Components: Performance
>            Reporter: Knut Anders Hatlen
>            Assignee: Knut Anders Hatlen
>            Priority: Minor
>         Attachments: bank-clients.diff, create-accounts.diff, lob.diff, no-pk.diff, runner.diff
>
>
> I'm planning to add some more load types to org.apache.derbyTesting.perf.clients.Runner so that we can run tests whose load is roughly equivalent to the tests run in Olav's nightly performance regression tests - http://home.online.no/~olmsan/derby/perf/. Having such tests available would make it easier to investigate performance regressions reported by those test runs. Currently, we have clients under o.a.dt.perf.clients that generate load that is similar to the single-record select and single-record update operations with primary key and the primary key join operations (in fact, the nightly regression tests have used the tests in derbyTesting.jar for these types of load for about a month now).
> For the following types of load, we don't have test clients in derbyTesting.jar:
>   1) tests for insert performance
>   2) tests for delete performance
>   3) single-record select + single-record update on secondary index
>   4) single-record select + single-record update on non-indexed column
>   5) TPC-B like transactions (should probably be called something else, but I think it's useful to have a test that uses the same schema and queries as described in the TPC-B spec)
> (3) and (4) are just variants over the existing single-record tests.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.