re:
http://www.garlic.com/~lynn/2018c.html#9 Graph database on z/OS?
http://www.garlic.com/~lynn/2018c.html#10 Graph database on z/OS?

some old "graph", much earlier I had been involved in original
sql/relational implementation, System/R ... some past posts
http://www.garlic.com/~lynn/submain.html#systemr

the "official" next generation DBMS was "EAGLE" ... and while the
corporation was focused on EAGLE, we were able to do technology
transfer to Endicott for release as SQL/DS. Then when EAGLE implodes,
there is request about how fast could System/R be ported to MVS
... eventually released as DB2, originally for decision support only.

About the same time, I was also sucked into helping with a different
kind of relational ... that physically instantiated every entity and
every relation ... a little more like IMS ... but w/o record pointers
... entities and relations were content addressable indexes. As a
result it could represent any kind of information structure
... including tables as well as graphs. IDEA was heavily influenced by
System/R in eliminating explicit record numbers with indexing under
the covers ... but also Sowa, who was at IBM STL at the time.
http://www.jfsowa.com/pubs/semnet.htm
topic drift, other Sowa reference (about IBM FS)
http://www.jfsowa.com/computer/memo125.htm
http://www.jfsowa.com/computer/

In some respect, System/R (RDBMS) was optimized for financial
transactions, tables with account number index and most everything
related to the account was physically in same record. IDEA could have
seperate record for every (indexed) entity and every (indexed) relation
(could be 5-10 times physical space of RDBMS tables).

Obvious doing financial transaction ran much faster on RDBMS (one record
with all information) than compared to dozen or more records for the
same information in IDEA.

However, for non-uniform information structure, IDEA could be several
times faster. A large VLSI chip design was loaded into DB2 and then
several traces were run to get best optimization. Then DB2 (hihgly
optimized) test took nearly 3hrs elapsed time to extract the full chip
design running on 3081 with 3380 disks. IDEA running with no
optimization on same 3081 and 3380s ... took less than 30mins to extract
same chip design (almost ten times faster).

IDEA also had query language that solved/addressed the SQL NULLs problem
(IDEA only has fields for things that exist, there is no direct concept
of "missing values") ... old archived post from DBMS theory discussion
http://www.garlic.com/~lynn/2003g.html#40 How to cope with missing values - 
NULLS?
http://www.garlic.com/~lynn/2003g.html#41 How to cope with missing values - 
NULLS?

We also did some work with NIH national library of medicine. They had
hired a company to load the NLM "index" information into RDBMS ... they
had spent 18months on "normalization" and could only do about 80% of the
data (the rest was loaded unnormalized with lots of duplicates).
Normalization/integrating new information was taking longer than real
time (four months of new medical knowledge was taking more than four
months).

I (one person) spent about three weeks doing the equivalent for IDEA.

some NLM refs:
https://www.nlm.nih.gov/databases/umls.html
https://www.nlm.nih.gov/research/umls/
https://www.nlm.nih.gov/research/umls/sourcereleasedocs/current/MSH/
https://www.nlm.nih.gov/mesh/intro_trees.html

trivia: at the time, NLM still had people that had originally done their
mainframe based online catalog in the 60s (BDAM with their own home
grown transaction system). This was same time that the univ library had
gotten an ONR grant to do online catalog, some of the money was used for
2321 datacell and it was also selected to be betatest site for original
CICS product ... and I got tasked with supporting/debugging the CICS
implementation (so we had lots of discussion about online catalogs and
IBM BDAM). past BDAM/CICS posts
http://www.garlic.com/~lynn/submain.html#bdam

-- 
virtualization experience starting Jan1968, online at home since Mar1970

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

Reply via email to