IDS Forum

Art's Fear the Panther....

Posted By: Fernando Nunes
Date: Tuesday, 19 October 2010, at 6:23 p.m.

You probably noticed that Art Kagel posted a new entry in his blog:

http://informix-myview.blogspot.com/2010/10/fear-panther.html

At the end he mentions that IIUG is preparing the vNext survey (!). The
world never stops!

He wrote:

"The IIUG will soon be posting its new Features Survey which we will had to
IBM to help them design the next release of Informix Dynamic Server. If you
don't think that you have a voice in how IBM determines features for
Informix, guess again. Every one of the features mentioned here was
requested by users like you/me/us either directly or through an expressed
need that had to be filled. Several are features that were asked for and
received high marks on the previous Features Surveys. So, when you see the
Insider or email announcement of the Survey, fill it out."

And mentioned some features he would like to see next. I commented and
Kernoal Stephens followed me. While we wait for the survey I thought it
would be nice to start thinking about some possible features.
The one me and Kernoal discussed was:

- Possibility of making upgrades without restoring secondaries.

Kernoal mentioned that he has RSS on the other side of the globe, so a total
re-initialization is not good.
In my perspective, even for "next rack" secondaries, this is terrible. Many
times the secondaries are used for ETL or DSS queries. And to restore them
can sometimes take hours (ifxclone will be a great help by the way).
So, would it be so difficult to implement this? I'm assuming that the
conversion actions (inplace migration) are logged. If this is correct, I
don't see why the secondaries could not "replay the upgrade".
The oninit would have to check that it's running a different version from
the reserved pages... Then it would need to verify that the logs it got or
that are on the primary start with a new log record that says "UPGRADE TO
VERSION x". Check that "x" is the version it is running (oninit), and if ok
then apply the logs... If something doesn't match say "must run with version
x. Please restart with correct version".
Would this be too hard to implement? From a "blind man" point of view this
looks much easier that the new rolling upgrade (cdr start sec2er) which by
the way I tested today (very small instance) and I must say it's amazing!

Other features to consider:
- expand the concept of FULL_DISK_INIT to other situations (restores, chunk
creation etc.)
- capture and remotely apply workload
- HDR between platforms (I know Madison doesn't like this one. I would love
if on oninit -i we could define the "system page size" to 4K for example,
and if we could send the logs and other stuff in "network format" so that we
got the same thing for the same page on the remote site). This would be a
killer for moving to other platforms and new hardware... and for the grid...
- ability to create a clone (standalone) from a secondary server using
ifxclone even if that forced the secondary to be a "lag" server
- ability to "rename chunks" in ifxclone. A MUST have I would say for
creating test environments and I suppose not very hard to do
- More dynamic parameters
- lock limit per session (I'd kill for this one)
- a FORCE_DDL for everything...
- fine grain auditing for other menemonics besides RDRW,UPRW,DLRW and INRW
(like EXSP)
- Session ID in the audit logs. PID is nearly useless in many situations
like java applications. And I need to be able to match audit logs and
logical logs
- more info for ACTB (access table) specifying if it was SELECTED, INSERTED,
UPDATED or DELETED and the result (OK, NOK)
- Explain in any tool (server side view for last or N last statements)
- in memory database (it's not the same as forcing the tables in cache...)
- Materialized views
- XPS "slices"
- result set caching (we already keep the number of updates/deletes/inserts
so for a frequent query at least on a single table, why not keep the result
if none of these change?)
Start by looking at the statement cache... see the more used queries... if
single table and "non variant" (i.e.: no "datetime" etc. data)...
- PL/SQL datablade? :)
- TSQL datablade? :)
- Purescale :P

Ok... enough. Now is the time to tame the latest beast, but someone out
there should be thinking about this. Some of the suggestions are easy I
believe, others not so, and others are just for laugh.
For now, a great thanks to the development team who was able to put a great
set of features into Panther. But as usual, we need to keep improving!

Regards.
--
Fernando Nunes
Portugal

http://informix-technology.blogspot.com
My email works... but I don't check it frequently...

--00163662e56633019d0492ffbfdb

Messages In This Thread

• Art's Fear the Panther....
  Fernando Nunes -- Tuesday, 19 October 2010, at 6:23 p.m.
  • Re: Art's Fear the Panther....
    Madison Pruet -- Wednesday, 20 October 2010, at 12:24 p.m.
    • Re: Art's Fear the Panther....
      Fernando Nunes -- Wednesday, 20 October 2010, at 1:14 p.m.
      • Re: Art's Fear the Panther....
        Madison Pruet -- Wednesday, 20 October 2010, at 1:37 p.m.
    • Re: Art's Fear the Panther....
      Art Kagel -- Wednesday, 20 October 2010, at 3:47 p.m.

IDS Forum is maintained by Administrator with WebBBS 5.12.