Das folgende SQL berechnet einige Key performance Indicators einer DB – auf Basis der AWR History Views. Speziell wird hier für Jeden Snapshot von AWR die zugehörige Elapsed Time, DB Time, CPU Time und die AAS = Average Active Sessions berechnet. AAS eignet sich sehr als erster Indikator die Performance der Datenbank für einzelne Snapshots schnell auszuschliessen und auch schnell Kandidaten zu finden, die auf einen Engpass innerhalb der DB hinweisen.
Die AAS berechnet sich als DB_TIME / Elapsed_Time.
Ist diese deutlich kleiner 1 (nahe 0), kann man einen Engpass auf DB – Ebene ausschliessen. Liegt diese im Bereich von 1 bis Anzahl Cores, können einzelne Statements oder Session Probleme mit der Performance haben. Die Gesamtperformance liegt aber noch im vertretbaren Bereich. Je näher die AAS sich dem Wert von NUM_CORES nähert oder diesen sogar überschreitet, um zu sicherer ist von einem Engpass auf DB Seite auszugehen.
Daneben werden noch eine Reihe andere KPI (Key performance Indikatoren) berechnet, um gleich ein besseren Überblick zu haben. So wird neben der DB – Time auch die CPU Time berechnet, um zu sehen, wieviel der Zeit in DB Time auf Warten oder auf Nutzung der CPU fällt. Je kleiner die CPU Time neben hoher DB Time, um so höher das Potential an Tuningmaßnahmen zur Steigerung der Gesamtperformance der DB.
Desweitern stellt die COMMIT Time als Durchschnitt der Wartezeit des Wait-Events „log file sync“ pro Snapshot ein Indikator für durchschnittliche Dauer eines COMMIT. Diese Zeit kann durch Trigger oder andere synchrone Operationen wie z.B. der Refresh einer Mat. View (bei REFRESH on COMMIT) deutlich mit den Anzahl der DML Operationen ansteigen.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 | WITH BASISDATA AS ( SELECT S.SNAP_ID AS SNAP_ID , S.DBID AS DBID , S.INSTANCE_NUMBER AS INST_ID , S.BEGIN_INTERVAL_TIME AS STARTTIME , S.END_INTERVAL_TIME AS ENDTIME , round(t.value/1000000/60, 2) AS TVALUE , CASE WHEN t.stat_name = 'DB time' THEN 'DB' WHEN t.stat_name = 'DB CPU' THEN 'CPU' WHEN t.stat_name = 'PL/SQL execution elapsed time' THEN 'PL/SQL' WHEN t.stat_name = 'RMAN cpu time (backup/restore)' THEN 'RMAN' WHEN t.stat_name = 'sql execute elapsed time' THEN 'SQL' END AS TTYPE , oc.value AS CORES , round( oi.value / (oi.value+ob.value) , 2) AS IDLE FROM dba_hist_osstat oc , dba_hist_osstat oi , dba_hist_osstat ob , (SELECT * FROM dba_hist_snapshot s WHERE S.BEGIN_INTERVAL_TIME >= trunc(SYSTIMESTAMP-7) ) s , dba_hist_sys_time_model t WHERE OC.STAT_NAME='NUM_CPUS' AND OC.DBID = S.DBID AND OC.SNAP_ID = S.SNAP_ID AND OC.INSTANCE_NUMBER = S.INSTANCE_NUMBER AND OI.STAT_NAME = 'IDLE_TIME' AND OI.DBID = S.DBID AND OI.SNAP_ID = S.SNAP_ID AND OI.INSTANCE_NUMBER = S.INSTANCE_NUMBER AND OB.STAT_NAME = 'BUSY_TIME' AND OB.DBID = S.DBID AND OB.SNAP_ID = S.SNAP_ID AND OB.INSTANCE_NUMBER = S.INSTANCE_NUMBER AND oi.value + ob.value > 0 AND T.DBID = S.DBID AND T.SNAP_ID = S.SNAP_ID AND T.INSTANCE_NUMBER = S.INSTANCE_NUMBER AND t.stat_name IN ('DB time', 'DB CPU' , 'PL/SQL execution elapsed time' , 'sql execute elapsed time' , 'RMAN cpu time (backup/restore)') ) , COMMITTM AS ( SELECT se.DBID AS DBID , se.SNAP_ID AS SNAP_ID , se.INSTANCE_NUMBER AS INST_ID , round( SUM(se.TIME_WAITED_MICRO)/1000/1000/60,2 ) AS COMMIT_TIME FROM dba_hist_system_event se WHERE se.EVENT_NAME = 'log file sync' GROUP BY se.DBID , se.SNAP_ID , se.INSTANCE_NUMBER ) , TXNUM AS ( SELECT ss.DBID AS DBID , ss.SNAP_ID AS SNAP_ID , ss.INSTANCE_NUMBER AS INST_ID , SUM(ss.VALUE) AS TXNUMS FROM dba_hist_sysstat ss WHERE ss.stat_name IN ('user commits', 'user rollbacks') GROUP BY ss.DBID , ss.SNAP_ID , ss.INSTANCE_NUMBER ) , WAITTM AS ( SELECT se.DBID AS DBID , se.SNAP_ID AS SNAP_ID , se.INSTANCE_NUMBER AS INST_ID , round( SUM(se.TIME_WAITED_MICRO)/1000/1000/60,2 ) AS WAIT_TIME FROM dba_hist_system_event se WHERE se.wait_class IN ('User I/O' , 'System I/O' , 'Commit' , 'Administrative') GROUP BY se.DBID , se.SNAP_ID , se.INSTANCE_NUMBER ) , ADDVAL AS ( SELECT st.DBID AS DBID , st.SNAP_ID AS SNAP_ID , st.INSTANCE_NUMBER AS INST_ID , round( avg( CASE WHEN st.metric_name LIKE 'Redo Generated Per Sec' THEN st.value END),2 ) AS REDO , round( avg( CASE WHEN st.metric_name LIKE 'Average Active Sessions' THEN st.value END),2 ) AS AAS , round( avg( CASE WHEN st.metric_name LIKE 'Logical Reads Per Txn' THEN st.value END),2 ) AS LOGREADS_TXN , round( avg( CASE WHEN st.metric_name LIKE 'Physical Reads Per Txn' THEN st.value END),2 ) AS READS_TXN , round( avg( CASE WHEN st.metric_name LIKE 'Physical Reads Per Sec' THEN st.value END),2 ) AS READS_SEC , round( avg( CASE WHEN st.metric_name LIKE 'Physical Writes Per Sec' THEN st.value END),2 ) AS WRITES_SEC , round( avg( CASE WHEN st.metric_name LIKE 'SQL Service Response Time' THEN st.value END),2 ) AS RESP_SQL_SERVTIME , round( avg( CASE WHEN st.metric_name LIKE 'Response Time Per Txn' THEN st.value END),2 ) AS RESP_TIME_TXN , round( avg( CASE WHEN st.metric_name LIKE 'Executions Per Sec' THEN st.value END),2 ) AS EXECS_PER_SEC , round( avg( CASE WHEN st.metric_name LIKE 'User Calls Per Sec' THEN st.value END),2 ) AS CALLS_PER_SEC FROM dba_hist_sysmetric_history st GROUP BY st.DBID , st.SNAP_ID , st.INSTANCE_NUMBER ) , DATA AS ( SELECT d1.INST_ID AS INST_ID , d1.STARTTIME AS STARTTIME , d1.ENDTIME AS ENDTIME , d1.TVALUE AS DB_TIME , d2.TVALUE AS CPU_TIME , d3.TVALUE AS PLSQL_TIME , d4.TVALUE AS RMAN_TIME , d5.TVALUE AS SQL_TIME , d1.cores AS CORES , d1.idle AS IDLE , w.WAIT_TIME AS WAIT_TIME , c.COMMIT_TIME AS COMMIT_TIME , t.TXNUMS AS TX_NUMS , a.REDO AS REDO , a.AAS AS AAS , a.LOGREADS_TXN AS LREADS_TXN , a.READS_TXN AS READS_TXN , a.READS_SEC AS READS_SEC , a.WRITES_SEC AS WRITES_SEC , a.RESP_SQL_SERVTIME AS RESP_SQL_SERVTIME , a.RESP_TIME_TXN AS RESP_TIME_TXN , a.EXECS_PER_SEC AS EXECS_PER_SEC , a.CALLS_PER_SEC AS CALLS_PER_SEC FROM BASISDATA d1 , BASISDATA d2 , BASISDATA d3 , BASISDATA d4 , BASISDATA d5 , WAITTM w , COMMITTM c , TXNUM t , ADDVAL a WHERE (d1.DBID = d2.DBID) AND (d1.SNAP_ID = d2.SNAP_ID) AND (d1.INST_ID = d2.INST_ID) AND (d1.STARTTIME = d2.STARTTIME) AND (d1.DBID = d3.DBID) AND (d1.SNAP_ID = d3.SNAP_ID) AND (d1.INST_ID = d3.INST_ID) AND (d1.STARTTIME = d3.STARTTIME) AND (d1.DBID = d4.DBID) AND (d1.SNAP_ID = d4.SNAP_ID) AND (d1.INST_ID = d4.INST_ID) AND (d1.STARTTIME = d4.STARTTIME) AND (d1.DBID = d5.DBID) AND (d1.SNAP_ID = d5.SNAP_ID) AND (d1.INST_ID = d5.INST_ID) AND (d1.STARTTIME = d5.STARTTIME) AND (d1.DBID = w.DBID) AND (d1.SNAP_ID = w.SNAP_ID) AND (d1.INST_ID = w.INST_ID) AND (d1.DBID = c.DBID) AND (d1.SNAP_ID = c.SNAP_ID) AND (d1.INST_ID = c.INST_ID) AND (d1.DBID = t.DBID) AND (d1.SNAP_ID = t.SNAP_ID) AND (d1.INST_ID = t.INST_ID) AND (d1.DBID = a.DBID) AND (d1.SNAP_ID = a.SNAP_ID) AND (d1.INST_ID = a.INST_ID) AND (d1.TTYPE = 'DB') AND (d2.TTYPE = 'CPU') AND (d3.TTYPE = 'PL/SQL') AND (d4.TTYPE = 'RMAN') AND (d5.TTYPE = 'SQL') ) , DIFF AS ( SELECT inst_id , starttime , CAST(starttime AS DATE) AS sdate , endtime , CAST(endtime AS DATE) AS edate , trunc( CAST(endtime AS DATE)-CAST(starttime AS DATE)) * 24 * 60 + trunc( MOD( (CAST(endtime AS DATE)-CAST(starttime AS DATE))*24, 24 ) ) * 60 + trunc( MOD( (CAST(endtime AS DATE)-CAST(starttime AS DATE))*24*60, 60 ) ) AS DURATION , cores , IDLE AS IDLE , db_time AS db_time_2 , Lag(db_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS db_time_1 , db_time - Lag(db_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS DBTIME , cpu_time AS cpu_time_2 , Lag(cpu_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS cpu_time_1 , cpu_time - Lag(cpu_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS CPUTIME , plsql_time AS plsql_time_2 , Lag(plsql_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS plsql_time_1 , plsql_time - Lag(plsql_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS PLSQLTIME , rman_time AS rman_time_2 , Lag(rman_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS rman_time_1 , rman_time - Lag(rman_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS RMANTIME , sql_time AS sql_time_2 , Lag(sql_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS sql_time_1 , sql_time - Lag(sql_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS SQLTIME , wait_time AS wait_time_2 , Lag(wait_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS wait_time_1 , wait_time - Lag(wait_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS WAITTIME , commit_time AS commit_time_2 , Lag(commit_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS commit_time_1 , commit_time - Lag(commit_time, 1, 0) OVER (ORDER BY inst_id, starttime) AS COMMITTIME , tx_nums AS txnum_2 , Lag(tx_nums, 1, 0) OVER (ORDER BY inst_id, starttime) AS txnum_1 , tx_nums - Lag(tx_nums, 1, 0) OVER (ORDER BY inst_id, starttime) AS TXNUMs , REDO AS REDO , AAS AS AAS , LREADS_TXN AS LREADS_TXN , READS_TXN AS READS_TXN , READS_SEC AS READS_SEC , WRITES_SEC AS WRITES_SEC , RESP_SQL_SERVTIME AS RESP_SQL_SERVTIME , RESP_TIME_TXN AS RESP_TIME_TXN , EXECS_PER_SEC AS EXECS_PER_SEC , CALLS_PER_SEC AS CALLS_PER_SEC FROM DATA ), VAL AS ( SELECT inst_id , starttime , sdate , endtime , DURATION , IDLE , DBTIME , CPUTIME , PLSQLTIME , RMANTIME , SQLTIME , WAITTIME , COMMITTIME , TXNUMS , round ( DBTIME / Duration, 2) AS AAS2 , cores , REDO AS REDO , AAS AS AAS , LREADS_TXN AS LREADS_TXN , READS_TXN AS READS_TXN , READS_SEC AS READS_SEC , WRITES_SEC AS WRITES_SEC , round(RESP_SQL_SERVTIME / 100,2) AS RESP_SQL_SERVTIME , round(RESP_TIME_TXN / 100,2) AS RESP_TIME_TXN , EXECS_PER_SEC AS EXECS_PER_SEC , CALLS_PER_SEC AS CALLS_PER_SEC FROM DIFF WHERE db_time_1 > 0 AND db_time_2 - db_time_1 > 0 AND cpu_time_1 > 0 AND cpu_time_2 - cpu_time_1 > 0 ) SELECT inst_id , sdate AS "Startdate" , duration AS "Ela(min)" , Cores , idle , AAS AS AAS , CASE WHEN aas > 0.5 THEN '1 - GREEN' WHEN aas >= 0.5 AND aas <= 1 THEN '2 - green' WHEN aas > 1 AND aas < cores/2 THEN '3 - yellow' WHEN aas >= cores/2 AND aas <= cores THEN '4 - YELLOW' ELSE '6 - RED' END AS "PerfInd" , dbtime AS "DB(min)" , cputime AS "CPU(min)" , committime AS "COMMIT(min)" , TXNUMS AS "Num TX" , plsqltime AS "PL/SQL(min)" , sqltime AS "SQL(min)" , rmantime AS "RMAN(min)" , waittime AS "WAIT(min)" , REDO AS "REDO (B/s)" , LREADS_TXN AS "LOG.READS/Tx" , READS_TXN AS "PHY.READS/Tx" , READS_SEC AS "PHY.READS/s" , WRITES_SEC AS "PHY.WRITES/s" , RESP_SQL_SERVTIME AS "SQL RESP/Call(s)" , RESP_TIME_TXN AS "RESPTIME/Tx(s)" , EXECS_PER_SEC AS "EXECS(s)" , CALLS_PER_SEC AS "CALLS(s)" FROM VAL ORDER BY 2 DESC,1 ; |
Thanks but this is what we are trying to understand
we want to find out the concurrent sessions for a snap id (avg sessions)
Since we have a value of 2.4 so still its tough to see what is 2.4 value refering to
Total sessions – 1300 so at that particular snap do we have 2.4% of 1300 or something else – if you can confirm that would be great.
Vinary,
as far as I understood you want to identify those sessions produced the hight load during the snap interval – right?
If this is your intention,
– AAS is only an indicator if there might be some sessions with an performance impact
– to identify the session you should inspect the ASH (Active Session History)
Please look at https://oracle-base.com/articles/10g/active-session-history for simple queries you can/should extend
HTH
regards
Rainer
What is AAS here i have a value of 2.4 does that mean 2.4% of total sessions , how do i interprest the concurrent sessions at that snap time
Hi Vinary,
as described in Section two, AAS = DB_TIME / Elapsed_Time.
If the Value is euql or larger than 1 you can / will have some session having issues where the performance. AAS should be low, e.g less than 0.2
HTH
regards
Rainer