|
Hi Arie,
Best regards to you too for the new year.
Here is where I believe we stand with all of the problems at V7.0
release (if it isn't already shipping, should be soon)
1. Performance problem : too much memory ,cpu and i/o - Notes : 602,952,953.
>>> re: 602 - We've cut back on the amount of stuff we load when you do a show
>>> table table_name
>>> re: 952 - Some of the HPV stuff has been reworked, I'm not sure if the
>>> problem referenced in this note specifically is one of the things that got
>>> addressed.
>>> re: 953 - We still require dispatch for remote, and dispatch still creates
>>> the new processes. The unavailable partition problem we could not fix in
>>> the V7 timeframe. It isn't in the first tier of stuff we want to do for V8
>>> either (native connect and nt have that distinction).
2. DBI doesn't recognize that an unavailable node becomes avaialable
Note : 938.
>>> We looked at it but couldn't do anything in the V7 timeframe. Ditto for
>>> V8.
3. There is a problem define a very large Partition View - Note : 944.
>>> As far as we know, this is a dispatch problem which was fixed.
4. Performance problem in Attach to many remote nodes - Note : 558.
>>> Code is in V7 but didn't get into dispatch for Rdb7. The dispatch folks
>>> said this was checked in for V8, so you may see it soon.
5. Partiotioned View over multiple columns not supported - Note : 845.
>>> Not implemented.
6. DBI is not doing partition elimination at run-time according to the value
of the host program variable - Note: 603.
>>> Implemented and in the next version.
7. All aggregate processing is done in DBI (when all data isn't coming
from the same database - Note : 478.
>>> Done in some cases now (for V7). There can't be a lot other stuff in the
>>> statements.
8. DBI is not supporting the ability to call stored procedures outside
of DBI (for example : calling ORACLE Stored procedures) - Note : 72.
>>> On our wish list.
|