Re: OMLETv4 The Ultimate Visual Real Time Oracle Monitoring Tool

amjadd_at_uop.edu.jo (omlet v4) wrote in
news:604b7892.0406010424.7fff3994_at_posting.google.com:

> Martin Burbridge <pobox002_at_bebub.com> wrote in message
> news:<Xns94F99E03A3AC3pobox002bebubcom_at_204.127.199.17>...
>> amjadd_at_uop.edu.jo (omlet v4) wrote in
>> news:604b7892.0405290545.1b3fb9df_at_posting.google.com:
>>
>> > Ed Stevens <nospam_at_noway.nohow> wrote in message
>> > news:<4269b012v3vns4nputqbo4g5sm8pskrhnl_at_4ax.com>...
>> >> On 26 May 2004 03:19:27 -0700, amjadd_at_uop.edu.jo (omlet v4) wrote:
>> >>
>> <snip>
>> >>
>> >> 4) "Set the following three init.ora parameters to: cursor_sharing
>> >> = force . . . " HONK! Wrong answer! If I do that, my acess
>> >> plans change, quite possibly for the worse.
>> >>
>> >> I don't consider myself an expert in this field, so if even I
>> >> could find problems without even opening the product . . .
>> >
>> > Ed,
>> >
>> > I doubt you know anything about cursor_sharing or access plans.
>> >
>>
>> And you would? I know exactly what cursor sharing force does. It
>> turns all literals in your select statement into bind valiables
>> before they are parsed, optimized and executed. And that is all
>>
>> So a report to find everyone in your department that made above
>> 100,000 in sales last month which might have been
>>
>> select ename from emp, sales
>> where deptno = 30
>> and qty > 100000
>> and emp.empid = sales.empid
>>
>> becomes something like
>>
>> select ename from emp, sales
>> where deptno = :bind_1
>> and qty > :bind_2
>> and emp.empid = sales.empid
>>
>
> Here is a guy that cannot write SQL arguing about access plans?! Our
> expert here skipped the "last month" predicate.
>

I also skipped a lot of things like all the columns you might want to see in an actual report. Its a code snippet to illustrate a point. The last month predicate is immaterial to the effects of cursor sharing so I left it out. If I wanted to be cute I would say sales is a view that only returns last months sales as that is what this application focuses on.

I can write SQL perfectly well enough to know this is valid and will parse and run, given the presence of the dependent objects, not bad for pseudo code.

>
> Have you heard anything about "bind variable peeking"?! Have you heard
> about oracle 9i?! Do you think Oracle engineers are as stupid as you
> not to consider values of bind_1 and bind_2?!
>

Yes I have. I can also imagine that peeking at bind variables that needn't be variables can be less efficient than not having to bother, because the constant value is already a literal. Maybe you can present a test case showing that bind variable peeking for a fixed value is more efficient than just parsing a literal, but I would doubt it.

>
> Do you know what an optimizer is? you don't!
>

Do you always answer the questions you ask yourself?

>
> Please limit your stupidity exposure to subjects you understand well.
> OMLET DOES NOT require this parameter set to FORCE at the instance
> level.
>

This is from the link in your original post

> 5- Set the following three init.ora parameters to:
>
>
> cursor_sharing = force
>

While this will set the parameter at the instance level, I'll admit it doesn't actually say its mandatory, it just tells you to do it.

I am just trying to point out to anyone reading that it is a crazy idea and should be a sign that either the app is technically suspect or the person giving the advice is.

>
> In fact, you can use similar or exact if you like, it is a matter of
> overhead.
>

You keep saying this, if it doesn't matter, why do you specify force in your installation guide? Of the hundreds of parameters available, of the three you decide to mention, the value for one is now redundant?

>
> The point raised is USE of scripts that repeatedly bombard v$ views
> and x$
> tables with queries that would translate into hard parses.
>

The point is that the only parses cursor sharing force avoids are those that are needlesly caused by apps with bugs.

>
> With OMLET there is no authority! no "ask tom"!? simply ask OMLET.
> OMLET's authority is the Oracle layers actual code.
>

Well that's not surprising as ask tom is a wealth of valuable technical information, and I can believe that you get nothing like that with your product. From others posts on the matter it would appear the application's reliance on discredited ratios means the product is mostly pointless. I'm just concerned that it might be dangerous also.

-- 
Martin Burbridge
add one to pobox002 for email