One big advantage is if you use memory store for high cardinality where one
or more of the columns stored in memory are numeric, the other columns are
the superset of predicates, and a lot of computation and aggregates are done
on the numeric with respect to row selection from the predicate columns.
Then you can unleash the full value of the vector processing. I really don't
know whether any of the other products out there compete at that level if
your "use case" fits this model.
Nice thread, and Andy Rivenes could certainly help someone evaluate whether
their "use case" hit the sweetest spot for DataBase In Memory (DBIM) or
whether it will merely speed things up a lot.
The notes from JL and Chris are also particularly interesting.
mwf
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx]
On Behalf Of Mladen Gogala
Sent: Wednesday, November 01, 2017 12:36 PM
To: oracle-l@xxxxxxxxxxxxx
Subject: Re: Anyone Using In-Memory?
Hi Scott,
I haven't yet seen a customer using it. I've played with it a bit, using
demo schemas and it seems to work well. However, I haven't done any
extensive testing of a real world application. The word "played" is the
operative word here. My impression is that in-memory caching makes the most
sense for low cardinality columns, i.e. the columns with small number of
distinct values like the gender column.
The problem is the price and the fact that the same capability is available
form DB2 10+, SQL Server 2016 or MariaDB at much lower price. This
capability is great if you're building a data warehouse. The first database
which came up with that technology was DB2 with its "Blu Acceleration",
almost a full year before Oracle. DB2 on Linux is now on version 11, the
technology has matured and is now very reliable.
Regards
On 10/30/2017 01:49 PM, Scott Canaan wrote:
I've been asked to look into the In-Memory option. One of our customers
attended a webinar for a BI tool (Looker) that uses a columnar database for
quick analytics of data. Instead of trying to move data from Oracle to
another database, we'd prefer to try it using In-Memory. Has anyone played
with this? Is it all it's advertised? Is it difficult to use?
We are on Oracle 12.1.0.2.
Thank you,
Scott Canaan '88 (srcdco@xxxxxxx)
(585) 475-7886 - work (585) 339-8659 - cell
"Life is like a sewer, what you get out of it depends on what you put into
it." - Tom Lehrer
--
Mladen Gogala
Oracle DBA
Tel: (347) 321-1217
