Re: [PoC] pgstattuple2: block sampling to reduce physical read - Mailing list pgsql-hackers

From firoz e v
Subject Re: [PoC] pgstattuple2: block sampling to reduce physical read
Date
Msg-id 5D90A4A7A6AC31449EDDBE18CD1CD0506F14FB35@szxeml521-mbx.china.huawei.com
Whole thread
In response to [PoC] pgstattuple2: block sampling to reduce physical read  (Satoshi Nagayasu <snaga@uptime.jp>)
List pgsql-hackers

 

On 16/09/13 16:20, Satoshi Nagayasu wrote:

> Thanks for checking. Fixed to eliminate SnapshotNow.

 

Looking forward to get a new patch, incorporating the comments, that are already given in the following mails:

 

1. Jaime Casanova: "The name pgstattuple2, doesn't convince me... maybe you can use pgstattuple() if you use a second argument (percentage of the sample) to overload the function".

(http://www.postgresql.org/message-id/5265AD16.3090704@catalyst.net.nz)

 

The comment related to having an argument, to mention the sampling number, is also given by Greg Smith: “There should be an input parameter to the function for how much sampling to do”

(http://www.postgresql.org/message-id/51EE62D4.7020401@2ndQuadrant.com)

 

2. Yourself: "I think it could be improved by sorting sample block numbers before physical block reads in order to eliminate random access on the disk."

(http://www.postgresql.org/message-id/525779C5.2020608@uptime.jp) for which, Mark Kirkwood , has given a rough patch.

 

Regards,

Firoz EV

 

pgsql-hackers by date:

Previous
From: Noah Misch
Date:
Subject: Re: UNION ALL on partitioned tables won't use indices.
Next
From: Amit Kapila
Date:
Subject: Re: Heavily modified big table bloat even in auto vacuum is running