Home Page for the TeradataForum

Archives of the TeradataForum

Message Posted: Fri, 06 Aug 2004 @ 11:56:14 GMT

  <Prev Next>   <<First <Prev

Subj:   Re: Cylinder Read vs MAXBLOCKSIZE
From:   Michael Larkins

Hi All:

This is a interesting topic with some interesting observations being made so far.

At the same time, we should keep in mind that as the blocksize shrinks, the number of blocks (to store the same amount of data) increases. As the number of blocks increase, so do the number of block headers and trailers. Therefore, the amount of disk space required to store the same data increases too. One Teradata user that I worked with doubled their blocksize and swore up and down that some of the data had to be missing because the PERM usage decreased so much. So, space is the first consideration.

Also, the number of block reads performed by the disk will increase and therefore the total number of physical reads will increase on the disk. Remember, a disk can only perform one read at a time; second consideration.

Like a disk, a bus can only transfer one piece of data at a time. So, the number of transfer requests will also increase on every SCSI bus; the third consideration.

To access the data, it must be in a buffer in memory on a node. Therefore, the system will have to manage more buffers. This will drive up the UNIX and CPU utilization on a node. If a system is already pushing the limits in these areas, the performance will decrease; yet another area for consideration.

Now, with this being said, please realize that some of this is conjecture on my part. I have not played with this new setting and can only draw on the normal physics of computing based on a past 30 yrs of experince with computers. I simply raise these issues as a basis to prompt more discussion along this line of thinking.

Best regards and have a blessed day,

Michael Larkins
Certified Teradata Master

  <Prev Next>   <<First <Prev
  Top Home Privacy Feedback  
Copyright for the TeradataForum (TDATA-L), Manta BlueSky    
Copyright 2016 - All Rights Reserved    
Last Modified: 15 Jun 2023