[jira] [Updated] (OAK-5859) Analyse and reduce IO amplification by OS

2017-11-14 Thread JIRA

 [ 
https://issues.apache.org/jira/browse/OAK-5859?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Michael Dürig updated OAK-5859:
---
Labels: performance scalability  (was: perfomance scalability)

> Analyse and reduce IO amplification by OS
> -
>
> Key: OAK-5859
> URL: https://issues.apache.org/jira/browse/OAK-5859
> Project: Jackrabbit Oak
>  Issue Type: Task
>  Components: segment-tar
>Reporter: Michael Dürig
>  Labels: performance, scalability
>
> Certain operation system settings might result in too much data actually 
> being read from disk causing early setting on of thrashing. E.g. transparent 
> huge pages or too big read aheads might be contra productive in combination 
> with the TarMKs memory mapping model. 
> * Determine the ratio of data being read by the TarMK and actual data being 
> read from disk. Determine the impact of relevant OS parameters (e.g. 
> transparent huge pages) on this ratio.
> * Compare memory mapped mode with file IO mode with an accordingly increased 
> segment cache. This would move prediction of what is likely to be read next 
> from the OS layer into our segment cache eviction strategy. 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)


[jira] [Updated] (OAK-5859) Analyse and reduce IO amplification by OS

2017-11-14 Thread JIRA

 [ 
https://issues.apache.org/jira/browse/OAK-5859?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Michael Dürig updated OAK-5859:
---
Labels: perfomance scalability  (was: )

> Analyse and reduce IO amplification by OS
> -
>
> Key: OAK-5859
> URL: https://issues.apache.org/jira/browse/OAK-5859
> Project: Jackrabbit Oak
>  Issue Type: Task
>  Components: segment-tar
>Reporter: Michael Dürig
>  Labels: perfomance, scalability
>
> Certain operation system settings might result in too much data actually 
> being read from disk causing early setting on of thrashing. E.g. transparent 
> huge pages or too big read aheads might be contra productive in combination 
> with the TarMKs memory mapping model. 
> * Determine the ratio of data being read by the TarMK and actual data being 
> read from disk. Determine the impact of relevant OS parameters (e.g. 
> transparent huge pages) on this ratio.
> * Compare memory mapped mode with file IO mode with an accordingly increased 
> segment cache. This would move prediction of what is likely to be read next 
> from the OS layer into our segment cache eviction strategy. 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)


[jira] [Updated] (OAK-5859) Analyse and reduce IO amplification by OS

2017-11-06 Thread JIRA

 [ 
https://issues.apache.org/jira/browse/OAK-5859?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Michael Dürig updated OAK-5859:
---
Fix Version/s: (was: 1.8)

> Analyse and reduce IO amplification by OS
> -
>
> Key: OAK-5859
> URL: https://issues.apache.org/jira/browse/OAK-5859
> Project: Jackrabbit Oak
>  Issue Type: Task
>  Components: segment-tar
>Reporter: Michael Dürig
>
> Certain operation system settings might result in too much data actually 
> being read from disk causing early setting on of thrashing. E.g. transparent 
> huge pages or too big read aheads might be contra productive in combination 
> with the TarMKs memory mapping model. 
> * Determine the ratio of data being read by the TarMK and actual data being 
> read from disk. Determine the impact of relevant OS parameters (e.g. 
> transparent huge pages) on this ratio.
> * Compare memory mapped mode with file IO mode with an accordingly increased 
> segment cache. This would move prediction of what is likely to be read next 
> from the OS layer into our segment cache eviction strategy. 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)


[jira] [Updated] (OAK-5859) Analyse and reduce IO amplification by OS

2017-09-11 Thread JIRA

 [ 
https://issues.apache.org/jira/browse/OAK-5859?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Michael Dürig updated OAK-5859:
---
Fix Version/s: (was: 1.7.8)

> Analyse and reduce IO amplification by OS
> -
>
> Key: OAK-5859
> URL: https://issues.apache.org/jira/browse/OAK-5859
> Project: Jackrabbit Oak
>  Issue Type: Task
>  Components: segment-tar
>Reporter: Michael Dürig
> Fix For: 1.8
>
>
> Certain operation system settings might result in too much data actually 
> being read from disk causing early setting on of thrashing. E.g. transparent 
> huge pages or too big read aheads might be contra productive in combination 
> with the TarMKs memory mapping model. 
> * Determine the ratio of data being read by the TarMK and actual data being 
> read from disk. Determine the impact of relevant OS parameters (e.g. 
> transparent huge pages) on this ratio.
> * Compare memory mapped mode with file IO mode with an accordingly increased 
> segment cache. This would move prediction of what is likely to be read next 
> from the OS layer into our segment cache eviction strategy. 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)


[jira] [Updated] (OAK-5859) Analyse and reduce IO amplification by OS

2017-08-04 Thread JIRA

 [ 
https://issues.apache.org/jira/browse/OAK-5859?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Michael Dürig updated OAK-5859:
---
Fix Version/s: (was: 1.7.6)
   1.7.8
   1.8

> Analyse and reduce IO amplification by OS
> -
>
> Key: OAK-5859
> URL: https://issues.apache.org/jira/browse/OAK-5859
> Project: Jackrabbit Oak
>  Issue Type: Task
>  Components: segment-tar
>Reporter: Michael Dürig
> Fix For: 1.8, 1.7.8
>
>
> Certain operation system settings might result in too much data actually 
> being read from disk causing early setting on of thrashing. E.g. transparent 
> huge pages or too big read aheads might be contra productive in combination 
> with the TarMKs memory mapping model. 
> * Determine the ratio of data being read by the TarMK and actual data being 
> read from disk. Determine the impact of relevant OS parameters (e.g. 
> transparent huge pages) on this ratio.
> * Compare memory mapped mode with file IO mode with an accordingly increased 
> segment cache. This would move prediction of what is likely to be read next 
> from the OS layer into our segment cache eviction strategy. 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)


[jira] [Updated] (OAK-5859) Analyse and reduce IO amplification by OS

2017-07-05 Thread JIRA

 [ 
https://issues.apache.org/jira/browse/OAK-5859?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Michael Dürig updated OAK-5859:
---
Fix Version/s: (was: 1.7.7)
   1.7.6

> Analyse and reduce IO amplification by OS
> -
>
> Key: OAK-5859
> URL: https://issues.apache.org/jira/browse/OAK-5859
> Project: Jackrabbit Oak
>  Issue Type: Task
>  Components: segment-tar
>Reporter: Michael Dürig
> Fix For: 1.7.6
>
>
> Certain operation system settings might result in too much data actually 
> being read from disk causing early setting on of thrashing. E.g. transparent 
> huge pages or too big read aheads might be contra productive in combination 
> with the TarMKs memory mapping model. 
> * Determine the ratio of data being read by the TarMK and actual data being 
> read from disk. Determine the impact of relevant OS parameters (e.g. 
> transparent huge pages) on this ratio.
> * Compare memory mapped mode with file IO mode with an accordingly increased 
> segment cache. This would move prediction of what is likely to be read next 
> from the OS layer into our segment cache eviction strategy. 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)


[jira] [Updated] (OAK-5859) Analyse and reduce IO amplification by OS

2017-07-05 Thread JIRA

 [ 
https://issues.apache.org/jira/browse/OAK-5859?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Michael Dürig updated OAK-5859:
---
Fix Version/s: (was: 1.7.5)
   1.7.7

> Analyse and reduce IO amplification by OS
> -
>
> Key: OAK-5859
> URL: https://issues.apache.org/jira/browse/OAK-5859
> Project: Jackrabbit Oak
>  Issue Type: Task
>  Components: segment-tar
>Reporter: Michael Dürig
> Fix For: 1.7.7
>
>
> Certain operation system settings might result in too much data actually 
> being read from disk causing early setting on of thrashing. E.g. transparent 
> huge pages or too big read aheads might be contra productive in combination 
> with the TarMKs memory mapping model. 
> * Determine the ratio of data being read by the TarMK and actual data being 
> read from disk. Determine the impact of relevant OS parameters (e.g. 
> transparent huge pages) on this ratio.
> * Compare memory mapped mode with file IO mode with an accordingly increased 
> segment cache. This would move prediction of what is likely to be read next 
> from the OS layer into our segment cache eviction strategy. 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)


[jira] [Updated] (OAK-5859) Analyse and reduce IO amplification by OS

2017-06-12 Thread JIRA

 [ 
https://issues.apache.org/jira/browse/OAK-5859?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Michael Dürig updated OAK-5859:
---
Fix Version/s: (was: 1.7.2)
   1.7.5

> Analyse and reduce IO amplification by OS
> -
>
> Key: OAK-5859
> URL: https://issues.apache.org/jira/browse/OAK-5859
> Project: Jackrabbit Oak
>  Issue Type: Task
>  Components: segment-tar
>Reporter: Michael Dürig
> Fix For: 1.7.5
>
>
> Certain operation system settings might result in too much data actually 
> being read from disk causing early setting on of thrashing. E.g. transparent 
> huge pages or too big read aheads might be contra productive in combination 
> with the TarMKs memory mapping model. 
> * Determine the ratio of data being read by the TarMK and actual data being 
> read from disk. Determine the impact of relevant OS parameters (e.g. 
> transparent huge pages) on this ratio.
> * Compare memory mapped mode with file IO mode with an accordingly increased 
> segment cache. This would move prediction of what is likely to be read next 
> from the OS layer into our segment cache eviction strategy. 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)


[jira] [Updated] (OAK-5859) Analyse and reduce IO amplification by OS

2017-02-28 Thread JIRA

 [ 
https://issues.apache.org/jira/browse/OAK-5859?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Michael Dürig updated OAK-5859:
---
Fix Version/s: 1.7.2

> Analyse and reduce IO amplification by OS
> -
>
> Key: OAK-5859
> URL: https://issues.apache.org/jira/browse/OAK-5859
> Project: Jackrabbit Oak
>  Issue Type: Task
>  Components: segment-tar
>Reporter: Michael Dürig
> Fix For: 1.7.2
>
>
> Certain operation system settings might result in too much data actually 
> being read from disk causing early setting on of thrashing. E.g. transparent 
> huge pages or too big read aheads might be contra productive in combination 
> with the TarMKs memory mapping model. 
> * Determine the ratio of data being read by the TarMK and actual data being 
> read from disk. Determine the impact of relevant OS parameters (e.g. 
> transparent huge pages) on this ratio.
> * Compare memory mapped mode with file IO mode with an accordingly increased 
> segment cache. This would move prediction of what is likely to be read next 
> from the OS layer into our segment cache eviction strategy. 



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)