Database performance tuning

Question

i have a very large database consisting of around 300,000,000 records with 38 columns in each record.
i need help in tuning the database so that query results don't take more than a few seconds.
the data is the log of events occuring in the SMSC server.
there are about 10,000,000 events daily.
Search criteria is usually time range and the mobile numbers.
Please advise on the sql and table structure to be used for best performance.presently using composite partitioning with range partitioning based on time( one partition for each day). query with 1 day range takes about 70-80 seconds.
i'm using java servlets and jdbc for accessing the database.
there's no problem in inserting data.

Answer 1

Hi
Thanks for your replies.
The table structure is as follows :
LOGGING_TIME DATE
LOG_TYPE NUMBER
SUC_INDICATOR NUMBER
ORIG_IW_TYPE NUMBER
ORIG_TYPE NUMBER
ORIG_ADDR VARCHAR2(21)
ORIG_ADDR_LEN NUMBER
DEST_IW_TYPE NUMBER
DEST_TYPE NUMBER
DEST_ADDR VARCHAR2(21)
DEST_ADDR_LEN NUMBER
SMS_CENTRE VARCHAR2(21)
INCOMING_TIME DATE
TIME DATE
ERROR_CAUSE NUMBER
ERROR_ORIGINATOR NUMBER
NO_OF_ATTEMPTS NUMBER
TARIFF_CLASS NUMBER
MSG_LEN NUMBER
PID NUMBER
SR_REQUEST NUMBER
DEFERRED_DEL NUMBER
SERV_DESC NUMBER
REF_NR NUMBER
MAX_NR NUMBER
SEQ_NR NUMBER
SP_MSG_IND NUMBER
DCS NUMBER
ACCESS_METHOD NUMBER
PRIORITY NUMBER
SENDER_CHG_TYPE NUMBER
RECIPIENT_CHG_TYPE NUMBER
SENDER_PREPAID_STATUS NUMBER
RECIPIENT_PREPAID_STATUS NUMBER
CHARGED_PARTY NUMBER
VMSC VARCHAR2(21)
ORIG_IMSI VARCHAR2(21)
CONSO_MSG VARCHAR2(10)
I have used composite partitioning with range partition based on logging_time and one day for each partition.there are 32 subpartitions.
I have created a local partitioned index on columns logging_time,orig_addr and dest_addr respectively as they are the most commonly used in queries.
query : select /*+ parallel_index(event_logs, event_logs_ind2)*/ * from event_logs where (logging_time>=?) and (logging_time<?) and (orig_addr like ?) and (log_type like ?) and (dest_addr like ?) order by logging_time
execution plan is as follows:
select statement
partition range(iterator)
     sort(order by)
     partition hash(all)
          table access(by local index rowid) of "smsevent.event_logs"
          index(range scan) of "smsevent.event_logs_ind"(non-unique)
I tried to make this more readable but the white spaces are getting removed from the posted message.
looking forward to a quick response