On Mon, 20 Feb 2017 14:36:06 +0000, Vernooij, Kees wrote:

>> Huge directory provides really
>> poor performance.
>
>Unless cached by LLA or another directory caching product.

I'm not sure that helps much.

The poor performance that Radoslaw mentioned is more on directory updates. 
If you create a member (insert a directory entry), the last entry will probably 
be pushed off the end of the directory entry block. That entry will be moved 
to the next block, and probably push the last entry off that block. This 
process will continue to the end of the directory.

Similarly, if a member is deleted, the space that remains at the end of the 
block 
will probably be big enough to hold the first entry on the next track. If it 
is, that 
entry will be moved, leaving room at the end of that block. Again, if the first 
entry on the block after that one will fit, and it probably will, it is also 
moved. 
Again, this process continues to the end of the directory.

The directory is always maintained in member name order, and all of the 
directory 
blocks except the last one are always kept as full as possible.

For directory searches, SEARCH KEY HIGH OR EQUAL is used, allowing the 
correct block to be read with a minimum of CPU processing and data transfer.

-- 
Tom Marchant

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to