Performance Baseline of Exadata X4-2

Size: px
Start display at page:

Download "Performance Baseline of Exadata X4-2"

Transcription

1 Performance Baseline of Exadata X4-2 Part IV: Database Load (DBL) Performance Technical Presentation June 2014

2 Contents 1 Introduction to Data Load (DBL) Performance Tests 2 System Configuration 3 Database Load Benchmark Results Conventional Load 4 Database Load Benchmark Results Bulk Load 5 Reviewing Database Load Benchmark Results copyright 2014 by benchware.ch slide 2

3 Introduction to Database Load (DBL) Performance Tests Why measure database load performance? What is measured? How is database load performance measured? Overview Benchware DBL performance test for Oracle Monitoring Benchware DBL performance tests Look at for detailed information copyright 2014 by benchware.ch slide 3

4 Contents 1 Introduction to Data Load (DBL) Performance Tests 2 System Configuration 3 Database Load Benchmark Results Conventional Load 4 Database Load Benchmark Results Bulk Load 5 Reviewing Database Load Benchmark Results copyright 2014 by benchware.ch slide 4

5 System Configuration Server configuration for load generation Server X2-2 X3-2 X4-2 CPU Type X5675 E E V2 #sockets #cores #threads Memory capacity [GByte] Performance Numbers from other Benchmarks SPECint_base_rate_2006 (throughput) Cluster #server within cluster Software Operating System Oracle Lx Oracle Lx Oracle Lx Oracle Database System Benchware Performance Suite copyright 2014 by benchware.ch slide 5

6 System Configuration Oracle Capacities and Licensing on Exadata Storage Systems Storage Capacities, raw, 1 Exadata Storage Server X2-2 X3-2 X4-2 Flash in [GByte] High Performance HDD in [TByte] 7.2@15K 7.2@15K 14.4@10K High Capacity HDD in [TByte] 36@7.2K 36@7.2K 48@7.2K Ratio of Flash capacity in percentage of HDD capacity 1% 4.5% 6.7% Exadata has: No flash storage, but flash cache Auto tiering technology Service time is not guaranteed Oracle Storage Server License X2-2 X3-2 X4-2 Oracle license cost (list price 13th of February 2014) Eighth Rack Quarter Rack Half Rack Software license cost in [USD] per 1 TByte Capacity copyright 2014 by benchware.ch slide 6

7 Contents 1 Introduction to Data Load (DBL) Performance Tests 2 System Configuration 3 Database Load Benchmark Results Conventional Load 4 Database Load Benchmark Results Bulk Load 5 Reviewing Database Load Benchmark Results copyright 2014 by benchware.ch slide 7

8 Load rate in [rps] Database Load Benchmark Results Oracle conventional load: 1 user process, different transaction size LGWR commit stress test X Transaction size in rows per transaction [rpt] copyright 2014 by benchware.ch slide 8

9 Exadata X4-2 FR HC Database Load Benchmark Results Oracle conventional load: 1 user process, different transaction size TX CPU Throughput Throughput SQL service Physical Physical Physical REDO REDO REDO REDO REDO Elap size busy rows/sec txn/sec time write write write size writes svt sync sync svt time Run Tst Code #N #J #T [rpt] [%] [rps] [tps] [s] [iops] [dbps] [MBps] [MBps] [iops] [ms] writes [us] [s] DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E Legend: Run benchmark run id #N number of RAC nodes [rps] rows per second [iops] i/o operations per second [s] time in seconds Tst benchmark test id #J number of load generators (jobs) [tps] transactions per second [dbps] database blocks per second [ms] time in milli seconds Code benchmark test code #T number of threads (PX) [ops] operations per second [MBps] mega byte per second [μs] time in micro seconds One user process is capable to fire insert transactions (commits) per second! copyright 2014 by benchware.ch slide 9

10 Load rate in [tps] Database Load Benchmark Results Oracle conventional load: 1 DB server, 2 rows per transaction LGWR commit stress test: Exadata X4-2 does not achieve the X3-2 performance Outlier for X4-2 at 16 load generators X Number of load generators copyright 2014 by benchware.ch slide 10

11 Load rate in [tps] Database Load Benchmark Results Oracle conventional load: Cluster, 2 rows per transaction X node 2 nodes 4 nodes 8 nodes copyright 2014 by benchware.ch slide 11

12 Exadata X4-2 FR HC Database Load Benchmark Results Oracle conventional load: Cluster, 2 rows per transaction TX CPU Throughput Throughput SQL service Physical Physical Physical REDO REDO REDO REDO REDO Elap size busy rows/sec txn/sec time write write write size writes svt sync sync svt time Run Tst Code #N #J #T [rpt] [%] [rps] [tps] [s] [iops] [dbps] [MBps] [MBps] [iops] [ms] writes [us] [s] DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E ###### Legend: Run benchmark run id #N number of RAC nodes [rps] rows per second [iops] i/o operations per second [s] time in seconds Tst benchmark test id #J number of load generators (jobs) [tps] transactions per second [dbps] database blocks per second [ms] time in milli seconds Code benchmark test code #T number of threads (PX) [ops] operations per second [MBps] mega byte per second [μs] time in micro seconds copyright 2014 by benchware.ch slide 12

13 Contents 1 Introduction to Data Load (DBL) Performance Tests 2 System Configuration 3 Database Load Benchmark Results Conventional Load 4 Database Load Benchmark Results Bulk Load 5 Reviewing Database Load Benchmark Results copyright 2014 by benchware.ch slide 13

14 Load rate in [rps] Database Load Benchmark Results Oracle bulk load: 1 DB server Million Rows per second are equal to 1 TByte per second usable table data X4-2 X Number of load generators X4-2 Outlier at 16 load generators copyright 2014 by benchware.ch slide 14

15 Load rate in [rps] Database Load Benchmark Results Oracle bulk load: cluster X4-2 X node 2 nodes 4 nodes 8 nodes copyright 2014 by benchware.ch slide 15

16 Exadata X4-2 FR HC Database Load Benchmark Results Oracle bulk load: cluster TX CPU Throughput Throughput SQL service Physical Physical Physical REDO REDO REDO REDO REDO Elap size busy rows/sec txn/sec time write write write size writes svt sync sync svt time Run Tst Code #N #J #T [rpt] [%] [rps] [tps] [s] [iops] [dbps] [MBps] [MBps] [iops] [ms] writes [us] [s] DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E DBL E E E copyright 2014 by benchware.ch slide 16

17 Contents 1 Introduction to Database Load (DBL) Performance Tests 2 System Configuration 3 Database Benchmark Results Conventional Load 4 Database Benchmark Results Bulk Load 5 Reviewing Database Load Benchmark Results copyright 2014 by benchware.ch slide 17

18 Reviewing Database Load Benchmark Results Summary Buffer Cache OLTP Insert, PL/SQL optimized Throughput, 1 process Maximum throughput, 1 DB Server Maximum throughput, Cluster Bulk load, un-compressed Maximum throughput, 1 process Maximum throughput, 1 DB Server Maximum throughput, Cluster Metric [tps] [µs] [tps] [µs] [tps] [µs] [rps] [rps] [rps] X4-2 FR copyright 2014 by benchware.ch slide 18

19 swiss precision in performance measurement