Re: [Hardhats-members] BIG NEWS re HealtheVet- St. Petersburg Tim es

Thu, 14 Apr 2005 23:17:37 -0700

OK, true Binary Trees are a heirarchical record structure which is ordered by some criteria, records after this one sort to the right and lesser records sort to the left.  Each new record walks the chain of records until it finds the niche where the new record belongs.  A pointer is captured, a new record is created and pieced into the chain.  OK, that is the traditional Binary Tree.
 
In most MUMPS implementations, we have a global directory block, which knows where all of the globals (persistent symbols) begin (their head nodes).  Each time a global diretory block gets full, it is extended into a new global directoy extension block.  Great, we have a quick string of names of the globals and their pointers to their head blocks.  Each global has one head block.  When it is first created with its first data block, there are two blocks involved, the head block and one data block.  As more data is added and arranged, the data block fills up.  Soon it fills up and we need to add a new data block.  Half of theone block is moved to the new data block and there is a new type of block added, a pointer block to now point to the two data blocks.  The data blocks are sequentially linked with at least a right pointer to the next data block.
 
Now as the database grows with more data blocks added, pointer blocks are added to the intermediate pointer structure between the head block and the data blocks.  The general rule of thumb is that the fan-out rate of the pointer blocks to data blocks is about 1:50.  This relationship is retained as there are more pointer blocks are created.  The structure grows as the data is added.  So the heirarchy is exceedingly wide and there are a minimum number of pointer levels between the pointer blocks.  I see that Maury has added some external references which might help.  I have to cut it here for tonight.
 
    Best wishes;  Chris
----- Original Message -----
From: Ron Ponto
Sent: Thursday, April 14, 2005 9:29 PM
Subject: Re: [Hardhats-members] BIG NEWS re HealtheVet- St. Petersburg Tim es

For us new Mumpies could you explain what B-trees are and what is a pointer split? 
 
-------Original Message-------
 
Date: 04/14/05 23:23:45
Subject: Re: [Hardhats-members] BIG NEWS re HealtheVet- St. Petersburg Tim es
 
Actually there is.
 
   Most MUMPS implementations are multi-way B-trees.  Real true B-trees
suffer badly in real-life, nearly every action is a pointer split.  The
performance of MUMPS sparse arrays is in the shared buffer pools and broad
pointer structures that get you to any data location in a 1,000,000 element
database in 7 physical disk accesses or less.
 
  Best wishes;  Chris
 
----- Original Message -----
From: "Thurman Pedigo" <[EMAIL PROTECTED]>
Sent: Thursday, April 14, 2005 7:26 PM
Subject: RE: [Hardhats-members] BIG NEWS re HealtheVet- St. Petersburg Tim
es
 
 
> Back to Donald E. Knuth eh? One would think that by now... Guess there
isn't
> much new under the Sun. ..tx/t
>
> > -----Original Message-----
> > From: [EMAIL PROTECTED] [mailto:hardhats-
> > [EMAIL PROTECTED]] On Behalf Of Greg Woodhouse
> > Sent: Thursday, April 14, 2005 6:14 PM
> > Subject: Re: [Hardhats-members] BIG NEWS re HealtheVet- St. Petersburg
Tim
> > es
> >
> > Of course, when you really get down to basics, it's all B-trees,
> > whether you're talking about MUMPS or your favorite RDBMS.
> > s-members
>
>
>
> -------------------------------------------------------
> SF email is sponsored by - The IT Product Guide
> Read honest & candid reviews on hundreds of IT Products from real users.
> Discover which products truly live up to the hype. Start reading now.
> _______________________________________________
> Hardhats-members mailing list
>
>
 
 
 
 
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
_______________________________________________
Hardhats-members mailing list
 

Reply via email to