On 5.9.2013 07:29, Alvaro Herrera wrote:
> Satoshi Nagayasu wrote:
>
>> But, for now, I think we should have a real index for the
>> statistics data because we already have several index storages, and
>> it will allow us to minimize read/write operations.
>>
>> BTW, what kind of index would be preferred for this purpose? btree
>> or hash?
>
> I find it hard to get excited about using the AM interface for this
> purpose. To me it makes a lot more sense to have separate, much
> simpler code. We don't need any transactionality, user defined
> types, user defined operators, or anything like that.
+1 to these concerns
And I think using regular tables might actually cause more harm than
benefits. For example let's say we have a large database with many
objects (which is the aim of this thread) with high activity - sessions
accessing objects, i.e. updating many "rows" in the stats tables.
Now, the stats table is likely to bloat thanks of the MVCC
copy-on-update. Not a good think, and it might easily happen the price
for maintenance of the table will be much higher than what we saved.
There are probably other similar scenarios.
Tomas