UNION or what else?

Question

Hi,
I have the next problem:
I have thousands of nodes, each node has different kind of statistical data stored in eight different tables (huge ones). The data in each table is stored based on node, day and hour. somethink like
node_id, day, hour_0, hour_1, hour_2 ... hour_23
I firstly download the info I need from the first table, the whole data of all days and all hours of one specific month for thousands of nodes is needed. If data in one hour is missed I need to go to look for that specific value of the same node, same day, same hour in the second table, if the second table does not have it I need to go to the look for it at the third table and so on until I reach the last table.
In order to improve performance, we decided to download the data massively (JDBC), that means that we downloaded all the info of all nodes for the thity days of an specific month of the first table at once, then we calculate what values are missed and we look for them in the second table, third .... eighth, all in massive downloads.
We noticed that it was a good idea to massively look for data in the first fetch,
select node_id, day, hour1, hour2 ... hour 23
from table1
where node_id in (thousands of nodes here)
and day between (first_day_month, last_day_month)
it drastically improved the response time, but for the next fetches (that are very specific) we use UNIONs like this
select val1 from table 2 where node=x1 and day=y1 and hour=z1
union
select val2 from table 2 where node=x2 and day=y2 and hour=z2
union
select valn from table 2 where node=xn and day=yn and hour=zn
We notice that using UNIONs does not make big difference than doing JDBC statements for each value separated, and our performance is far from the requirements.
Question: What would you do in order to improve performance in such a situation?
Thank you!!!

Answer 1

Not sure if I understood your requirement correctly. But maybe you can do it all in one select. Something long the lines of this
select t1.node_id,
          t1.day,
          t1.hour1,
          nvl(t1.hour2, (select min(val1) from table2 t2 where t2.node=t1.node_id and t2.day=t1.day and hour=2)) as hour2,
          nvl(t1.hour3, (select min(val1) from table2 t3 where t3.node=t1.node_id and t3.day=t1.day and hour=3)) as hour3,
          nvl(t1.hour23, (select min(val1) from table2 t23 where t23.node=t1.node_id and t23.day=t1.day and hour=23)) as hour23
from table1 t1
where t1.node_id in (thousands of nodes here)
and t1.day between (first_day_month, last_day_month)