On 8 Listopad 2011, 10:49, Ondrej Ivanič wrote:
> Hi,
>>> mostly heavy read
>>> workloads but OLTP performance is required (like run query over 100m+
>>> dataset in 15 sec)
>>
>> that isn't OLTP, its OLAP. Online Analytic Processing rather than
>> Online
>> Transaction Processing.... large complex reporting queries that have
>> to
>> aggregate many rows is classic OLAP.
>
> I didn't say OLTP. Our workload is "aggregations/drill downs/roll
> ups/... " (= OLAP) but we need OLTP like performance i.e. our users
> are not keen to wait more than several seconds for the result.
> Greenplum or Postgres + Fusion IO can deliver this performance for us.
Sure you did - you've stated that "mostly heavy read
workloads but OLTP performance is required (like run query over 100m+
dataset in 15 sec)." That clearly mentions OLTP ...
And OLTP has nothing to do with the amount of time the user is willing to
wait, it's rather about the amount of data the user read/modifies and the
number of tables. OLTP workload usually consists of transactions accessing
small amount of data, usually by primary key (get customer with id X,
update account with id Y, ...).
OLAP/DSS workload is somehow opposite - read large amounts of data,
perform complex analysis etc.
Tomas