Re: buffer advisor
- From: Justin Mungal <justin@xxxxxxx>
- To: Ls Cheng <exriscer@xxxxxxxxx>
- Date: Fri, 5 Sep 2014 03:18:24 -0500
64MB... fun... I take it there are performance issues?
I work mostly on Linux x64 systems... I use AMM, but sometimes we see lots
and lots of paging and resize operations on busier systems with larger
amounts of memory. On these systems we generally switch to ASMM and
HugePages, and this resolves the issue.
On Fri, Sep 5, 2014 at 2:40 AM, Ls Cheng <exriscer@xxxxxxxxx> wrote:
> Hi Chris
>
> That is why I am still questioning advisr's effciency and asking if people
> has got experiences using it to size buffer cache. If we look at ASMM or
> AMM which I suppose they use the advisors internally (or no?) the advisor
> doesnt seem useful at all.
>
> I just got a couple of AWR from a customer who is using memory_target, set
> to 3GB and AMM set shared pool to 2GB, PGA to 900MB and buffer cache to
> 64MB (!?#!).
>
>
> Thanks
>
>
>
> On Fri, Sep 5, 2014 at 12:50 AM, Chris Taylor <
> christopherdtaylor1994@xxxxxxxxx> wrote:
>
>> Seth,
>>
>> Really? I have run into a few situations where the advisor undersizes
>> the buffer cache significantly in favor of the shared pool because of the
>> workload of the application.
>> I've got a db right _now_ that has a 128MB buffer cache and a 20GB shared
>> pool that AMM resized because of the workload :)
>>
>> Obviously, the solution to this is to set floor (minimum) values for
>> shared_pool_size and db_cache_size but it still amazes me that ASMM/AMM
>> will significantly undersize the buffer cache when the workload uses a lot
>> of SQL that isn't reuseable.
>>
>> And I clearly recognize that the workload is suboptimal (lots of SQL with
>> literals and a few other things) that favor a large shared pool, and my
>> only point is that it isn't uncommon for the automatic memory resizing to
>> size the buffer cache to an absurd size :)
>>
>> Chris
>>
>>
>> On Thu, Sep 4, 2014 at 4:56 PM, Seth Miller <sethmiller.sm@xxxxxxxxx>
>> wrote:
>>
>>> Ls,
>>>
>>> I have found with very few exceptions that ASMM (SGA_TARGET) is very
>>> good at sizing the buffer cache. Have you tried this?
>>>
>>> Seth Miller
>>> On Sep 4, 2014 3:38 PM, "Ls Cheng" <exriscer@xxxxxxxxx> wrote:
>>>
>>>> Hi all
>>>>
>>>> Has anyone used buffer cache advisory in 10g or 11g to size a
>>>> production buffer cache? If so how good is the advisor recommending the
>>>> cache size? Did the recommended cache size meet the ohysical reads
>>>> reduction goal?
>>>>
>>>> TIA
>>>>
>>>>
>>>>
>>
>
Other related posts:
