Oracle Performance on Sun Server X4170 with Violin Memory Array Benchmark Report June 2012

Transcription

1 Oracle Performance on Sun Server X4170 with Violin Memory Array 3205 Benchmark Report June 2012

2 Contents 1 About Benchware 2 CPU Performance 3 Server Performance 4 Storage Performance 5 Database Performance 6 Conclusion copyright 2012 by benchware.ch slide 2

3 Benchware Ltd Services and Products Strong foundation in core technologies such as Oracle database systems, server and storage systems System Architecture, Component Evaluation, Reviews Performance Analysis & Optimization Benchmarking Database engineering copyright 2012 by benchware.ch slide 3

4 Benchware Ltd Value proposition Vendor-independent company - Benchware is completely committed to customers interests Holistic approach in designing, tuning and benchmarking Oracle systems Long experience track record - Responsible for system architecture of largest DWH and OLTP systems, mainly telecom and finance industry - Oracle since 1984 (Oracle Version 3) - Performance tuning and benchmarking since 1993 (Oracle Version 7) copyright 2012 by benchware.ch slide 4

5 Storage Network Networks (Application, Backup, Interconnect, ) System Management, Operations, Security, Resource Management Benchware Ltd Performance of complex Oracle platforms is not predictable Application Middleware Database System File System Volume Manager Networks (IP-based) Bandwidth, latency during remote database mirroring (sync, async) due to switches and sql*net and tcp/ip stack (frame size, etc.). Oracle Database Different versions, patches and options, about one hundred configuration parameters. Storage Network (FC-, IB- or IP-based) Bandwidth, latency during remote storage mirroring (sync, async) due to switches, hubs and distance. Volume & File Management Different volume managers (VxVM, ASM) and file systems (UFS, VxFS, ext3, JFS, ZFS, raw devices), different I/O methods (async, direct), a lot of configuration parameters (#LUNS, queue depth, max i/o unit), software striping and/or mirroring, multipathing. Operating System Virtualization Server Storage System Server & Operating System Different server systems, processors and CPU architectures, (x86, IA-64, UltraSparc, SPARC64, Power), #cores, multithreading, main memory, bus architecture. Different operating systems and patches, over one hundred configuration parameters, virtualization of resources. Storage System Different storage systems, storage tiers and storage technology: spindle count and System architects have a wide choice of components, speed, RAID management, cache management, server interface technology, technologies and configurations storage system options like remote copy, hardware striping and/or mirroring, virtualization of resources. copyright 2012 by benchware.ch slide 5

6 Storage Network DataGuard Network Fusion I/O Interconnect Benchware Ltd Complex architecture of Oracle platforms requires benchmarking Benchware Loader Benchware Monitor Benchware Performance Suite - Benchware Monitor - Benchware Loader Database System Performance measurement at the interface between application and Oracle database platform File System Volume Manager Operating System Virtualization Server Storage System Object of measurement: Oracle platform Key Performance Metrics can be used for SLA between IT operation and business Benchware uses Oracle database software to generate all kinds of loads for cpu, server, storage and database copyright 2012 by benchware.ch slide 6

7 Benchware Ltd Key performance metrics of Oracle platforms should be as easy as key performance metrics used in the automobile industry Source: copyright 2012 by benchware.ch slide 7

8 Benchware Ltd Library of Oracle benchmark tests - implemented in PL/SQL, Java and SQL CPU Performance CPU-bound PL/SQL operations OLTP systems DWH systems Proof of Efficiency Key Performance Metrics Unit pl/sql basic operations arithmetic mix, string mix multithreading virtualization speed throughput [s] [ops] pl/sql algorithms fibonacci, prime numbers Server Performance Server-bound SQL operations completely in SGA OLTP systems DWH systems Proof of Efficiency Key Performance Metrics Unit in-memory SQL full scan, index access pl/sql algorithms scalability cc-numa virtualization speed throughput [µs] [s] [bgps] [tps] [rps] [MBps] quicksort [s] seconds [ms] milli seconds (10-3 ) [µs] micro seconds (10-6 ) [ns] nano seconds (10-9 ) [bgps] [rps] [tps] [ops] buffer gets per second rows per second transactions per second operations per second [MBps] [GBps] [iops] [qpm] mega bytes per second giga bytes per second i/o operations per second queries per minute less important important very important copyright 2012 by benchware.ch slide 8

9 Benchware Ltd Library of Oracle benchmark tests - implemented in PL/SQL, Java and SQL Storage Performance I/O-bound Oracle operations OLTP systems DWH systems Proof of Efficiency Key Performance Metrics Unit sequential I/O 1 MByte, read and write random I/O 8 kbyte, read and write data integrity tiering, pooling striping virtualization replication throughput service time [ms] [MBps] [GBps] [iops] Database Performance Mixed resource usage: CPU, memory, storage OLTP systems DWH systems Proof of Efficiency Key Performance Metrics Unit data load uncompressed, compressed data scan data aggregation & reports scalability speed throughput service time [ms] [s] [rps] [tps] [qpm] [s] OLTP transactions insert, select, update [s] seconds [ms] milli seconds (10-3 ) [µs] micro seconds (10-6 ) [ns] nano seconds (10-9 ) [bgps] [rps] [tps] [ops] buffer gets per second rows per second transactions per second operations per second [MBps] [GBps] [iops] [qpm] mega bytes per second giga bytes per second i/o operations per second queries per minute less important important very important copyright 2012 by benchware.ch slide 9

10 Contents 1 About Benchware 2 CPU Performance 3 Server Performance 4 Storage Performance 5 Database Performance 6 Conclusion copyright 2012 by benchware.ch slide 10

11 Oracle Database Platform CPU CPU Type Frequency [GHz] Customer System Sun X4170 Violin VMA 3205 Intel Xeon E5649 #cores 6 Multithreading Server Customer System 2.53 GHz 2-fold Sun X4170 Violin VMA 3205 #sockets 2 #cores 12 #threads 24 CPU has huge impact on performance of many database operations - but also on Oracle license cost! copyright 2012 by benchware.ch slide 11

12 Throughput in [kops] CPU Performance PL/SQL string processing (data type VARCHAR2) 35'000 30'000 25'000 20'000 15'000 X4170 Customer System 10'000 5' Degree of parallelism (dop) one of the fastest available CPU (single thread speed) perfect scalability copyright 2012 by benchware.ch slide 12

13 X4170 CPU Performance PL/SQL string processing (data type VARCHAR2) Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E #N - describes the number of RAC nodes involved in this test #J - describes the number of jobs running the workload, jobs are round-robin distributed over all configured node in #N CPU [%] total CPU utilization copyright 2012 by benchware.ch slide 13

14 Throughput in [kops] CPU Performance PL/SQL integer processing (data type NUMBER) 9'000 8'000 7'000 6'000 5'000 4'000 X4170 Customer System 3'000 2'000 1' Degree of parallelism (dop) one of the fastest available CPU (single thread speed) perfect scalability copyright 2012 by benchware.ch slide 14

15 X4170 CPU Performance PL/SQL integer processing (data type NUMBER) Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E copyright 2012 by benchware.ch slide 15

16 Throughput in [kops] CPU Performance PL/SQL floating point processing (data type FLOAT) 140' ' '000 80'000 60'000 X4170 Customer System 40'000 20' Degree of parallelism (dop) one of the fastest available CPU (single thread speed) perfect scalability copyright 2012 by benchware.ch slide 16

17 X4170 CPU Performance PL/SQL floating point processing (data type FLOAT) Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E copyright 2012 by benchware.ch slide 17

18 Speed in [sec] CPU Performance PL/SQL algorithm (fibonacci, recursive) X Customer System 5 0 n=39 n=40 one of the fastest available CPU (single thread speed) record for n=39 is 12 sec record for n=40 is 19 sec copyright 2012 by benchware.ch slide 18

19 Benchmark Results Summary CPU Performance Metric Customer System Sun X4170 Violin VMA 3205 #cores 12 #threads 24 PL/SQL operations Metric Customer System String processing Speed (single thread) Throughput NUMBER processing Speed (singe thread) Throughput Floating point processing Speed (single thread) Throughput Algorithms Speed fibonacci recursive (n=39) Speed fibonacci recursive (n=40) [kops] [kops] [kops] [kops] [kops] [kops] [s] [s] Sun X4170 Violin VMA copyright 2012 by benchware.ch slide 19

20 Contents 1 About Benchware 2 CPU Performance 3 Server Performance 4 Storage Performance 5 Database Performance 6 Conclusion copyright 2012 by benchware.ch slide 20

21 Oracle Database Platform Server Server Customer System Sun X4170 Violin VMA 3205 #sockets 2 #cores 12 #threads (CPU_COUNT) 24 Main Memory [GByte] 96 Host-Bus-Adapter (type, quantity, throughput) 2 x 8 Gbit FC Operating System Solaris 11 Cluster #server - Most OLTP applications avoid I/O operations as much as possible and work predominately in RAM server performance is essential for these kind of OLTP applications! copyright 2012 by benchware.ch slide 21

22 Throughput in [krps] Server Performance In-memory SQL, full table scan 5'000 4'500 4'000 3'500 3'000 2'500 2'000 X4170, 8K Customer System X4170, 4K 1'500 1' Degree of parallelism (dop) perfect scalability no contention within Oracle copyright 2012 by benchware.ch slide 22

23 X4170 4K block size X4170 8K block size Server Performance In-memory SQL, full table scan Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E copyright 2012 by benchware.ch slide 23

24 Throughput in [tps] Server Performance In-memory SQL, primary key access 250'000 SQL statement service time < 5 ms 200' ' '000 X4170, 8K Customer System 50' Degree of parallelism (dop) perfect scalability up to tps avg service time of SQL statement < 5 ms no difference in performance between 8K block size and 4K block size copyright 2012 by benchware.ch slide 24

25 X4170 Server Performance In-memory SQL, primary key access Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E copyright 2012 by benchware.ch slide 25

26 Throughput in [krps] Server Performance In-memory SQL, secondary key access 6'000 5'000 SQL statement service time < 5 ms 4'000 3'000 X4170 Customer System 2'000 1' Degree of parallelism (dop) perfect scalability up to processed rows per second [rps] via index access no Oracle contention no difference in performance between 8K block size and 4K block size copyright 2012 by benchware.ch slide 26

27 X4170 Server Performance In-memory SQL, secondary key access Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E copyright 2012 by benchware.ch slide 27

28 Benchmark Results Summary Server Performance Server Metric Customer System Sun X4170 Violin VMA 3205 #cores 12 #threads 24 Main memory [GByte] 96 In-memory SQL operations Metric Customer System Sun X4170 Violin VMA 3205 Full table scan Throughput [rps] Random table access via primary key Throughput SQL service time Random table access via secondary key Throughput SQL service time [rps] [tps] [ms] [rps] [tps] [ms] < < 5 copyright 2012 by benchware.ch slide 28

29 Contents 1 About Benchware 2 CPU Performance 3 Server Performance 4 Storage Performance 5 Database Performance 6 Conclusion copyright 2012 by benchware.ch slide 29

30 Oracle Database Platform Storage System Storage System Customer System Sun X4170 Violin VMA 3205 Type 1 x VMA 3205 Front-end Controller (type, quantity, throughput) 2 x 8 Gbit FC Frontend Cache [GByte] - Back-end Controller (type, quantity, throughput) #HDD Disks - #SSD Disks - Single Layer Cell (SLC) NAND Flash Hardware RAID-Level RAID-3 (4+1) Hardware stripe unit size [KByte] 1 Thin provisioning, Pools - Usable disk capacity [TByte] - Usable flash capacity [TByte] ~ 5.25 copyright 2012 by benchware.ch slide 30

31 Oracle Database Platform Volume and File Management Volume Manager Customer System Sun X4170 Violin VMA 3205 #LUN s 7 Concatenated oder Striped striped Queue depth - Special parameters round-robin multipathing File System Customer System Sun X4170 Violin VMA 3205 Solaris UFS Block size [KByte] - Direct I/O - Concurrent I/O - Asynchronous I/O - Special parameters default configuration copyright 2012 by benchware.ch slide 31

32 Throughput in [MBps] Storage Performance Test Sequential read 1'400 1'200 7 ms 16 ms 33 ms Number at measuring point: Avg service time for 1 MByte I/O read request (128 x 8 kbyte) 1'000 3 ms 4 ms 8 ms 17 ms ms VMA 3205, 8K Customer System VMA 3205, 4K Degree of parallelism (dop) very high throughput even for low DOP copyright 2012 by benchware.ch slide 32

33 VMA K block size VMA K block size Storage Performance Test Sequential read Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E Avg I/O size about 1 MByte, equivalent to 128 x 8 KByte Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E Avg I/O size about 512 kbyte, equivalent to 128 x 4 KByte copyright 2012 by benchware.ch slide 33

34 Throughput in [MBps] Storage Performance Test Sequential write ms 3 ms 6 ms Number at measuring point: Avg service time for 1 MByte I/O write request (128 x 8 kbyte) ms VMA 3205 Customer System Degree of parallelism (dop) excellent throughput even for low DOP copyright 2012 by benchware.ch slide 34

35 VMA K block size VMA K block size Storage Performance Test Sequential write Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E Avg I/O size about 1 MByte Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E block size has almost no impact to performance of sequential write copyright 2012 by benchware.ch slide 35

36 Throughput in [iops] Storage Performance Test Random read 250' '000 1 ms 3 ms 4 ms Number at measuring point: Avg service time for random read request db_block_size = 8K (red) db_block_size = 4K (green) 150'000 < 500 µs VMA 3205, 8K 100'000 < 500 µs 1 ms 1 ms 3 ms 5 ms Customer System VMA 3205, 4K 50'000 0 < 500 µs < 500 µs < 500 µs < 500 µs < 500 µs Degree of parallelism (dop) High throughput with excellent service times Clear advantage of 4K database block size over 8K database block size copyright 2012 by benchware.ch slide 36

37 VMS K block size VMA K block size Storage Performance Test Random read Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E T E E Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E T E E copyright 2012 by benchware.ch slide 37

38 Throughput in [iops] Storage Performance Test Random write 120' '000 DBWR: < 500 µs LGWR: 2 ms Number at measuring point: Avg service time for DBWR and LGWR scattered write requests db_block_size = 8K (red) db_block_size = 4K (green) 80'000 60'000 DBWR: < 500 µs LGWR: 1 ms DBWR: < 500 µs LGWR: 5 ms VMA 3205, 8K Customer System VMA 3205, 4K 40'000 20'000 0 DBWR: < 500 µs LGWR: 1 ms Degree of parallelism (dop) Extraordinary throughput with excellent service times Clear advantage of 4K database block size over 8K database block size IOPS need about 400 (!) conventional 15k rpm disks copyright 2012 by benchware.ch slide 38

39 VMA K block size VMA K block size Storage Performance Test Random write Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E copyright 2012 by benchware.ch slide 39

40 Benchmark Results Summary Storage Performance Metric Customer System Sun X4170 Violin VMA K / 4K Sequential I/O Read Write [MBps] [MBps] / / 621 Random I/O read Throughput Service time [iops] [ms] / / 4 Random I/O write Throughput Service time, DBWR scattered write Service time, LGWR scattered write [iops] [µs] [ms] / < 500 / < / 5 Excellent throughput and service times for standard Solaris file system with default configuration copyright 2012 by benchware.ch slide 40

41 Contents 1 About Benchware 2 CPU Performance 3 Server Performance 4 Storage Performance 5 Database Performance 6 Conclusion copyright 2012 by benchware.ch slide 41

42 Oracle Database Platform Database System Installation Customer System Sun X4170 Violin VMA 3205 Oracle Edition (EE, SE) - EE Oracle Release Real Application Cluster - N Diagnostic Pack - Y DataGuard - N Flashback - N Configuration Customer System Sun X4170 Violin VMA 3205 SGA capacity [GByte] - 32 SGA keep cache [GByte] - 16 PGA capacity [GByte] - 8 Block size [KByte] - 8 / 4 NUMA support - disabled copyright 2012 by benchware.ch slide 42

43 Oracle Database Platform Benchmark Suite Benchmark Suite Customer System Sun X4170 Violin VMA 3205 Release - Benchware BPS 8.0 PL/SQL Code - native Benchmark Database Customer System Sun X4170 Violin VMA 3205 Benchmark Database size - V 1 TByte #rows small table #rows large table copyright 2012 by benchware.ch slide 43

44 Throughput in [krps] Database Performance Bulk load X4170, VMA 3205, 8K Customer System X4170, VMA 3205, 4K Degree of parallelism (dop) As expected, smaller database block size is less efficient for bulk load REDO log switch contention für DOP > 4 copyright 2012 by benchware.ch slide 44

45 VMA K block size VMA K block size Database Performance Bulk load Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E copyright 2012 by benchware.ch slide 45

46 Benchmark Results Summary Database Performance: bulk load Metric Customer System Sun X4170 Violin VMA 3205 Data load bulk, un-compressed Throughput best case (DOP = 1) Throughput max. Data load bulk, compressed Throughput best case (DOP = 1) Throughput max. Data aggregation Create b-tree index, speed Create b-tree index, throughput [rps] [MBps] [rps] [MBps] [rps] [MBps] [rps] [MBps] [s] [rps] 8K / 4 K / / / / @1x10 9 rows / / - Max REDO rate during bulk load [MBps] / - copyright 2012 by benchware.ch slide 46

47 Throughput in [tps] Database Performance OLTP select transaction, primary key access, 32 rows per tx 2'000 1'800 1'600 1ms 1 ms Number at measuring point: avg service time for single block random read db_block_size = 8K (red) db_block_size = 4K (green) 1'400 < 500 µs 1'200 1' < 500 µs 1 ms 1 ms 2 ms X4170, VMS3205 8K Customer System X4170, VMA 3205, 4K < 500 µs < 500 µs < 500 µs < 500 µs < 500 µs < 500 µs < 500 µs < 500 µs < 500 µs Degree of parallelism (dop) Excellent transaction rate for small system avg SQL service time < 0.06 s at DOP 128 for 4K block size avg SQL service time < 0.10 s at DOP 128 for 8k block size copyright 2012 by benchware.ch slide 47

48 X4170 VMA 3205, 4K X4170 VMA 3205, 8K Database Performance OLTP select transaction, primary key access, 32 rows per tx Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E T E E Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E T E E Due to block cleanout there is a mixed random I/O picture with both: read and write copyright 2012 by benchware.ch slide 48

49 Throughput in [rps] Database Performance OLTP insert transaction, 2 rows per tx 35'000 30'000 DBWR: < 500 µs LGWR: 1 ms Number at measuring point: avg service time for DBWR and LGWR scattered write requests 25'000 20'000 15'000 DBWR: < 500 µs LGWR: 2 ms X4170, VMA 3205, 8K Customer System X4170, VMA 3205, 4K 10'000 5' Degree of parallelism (dop) Excellent transaction rate for small system avg SQL service time < s at DOP 128 for 4K block size avg SQL service time < s at DOP 128 for 8k block size Extreme transaction rate with > commit per second copyright 2012 by benchware.ch slide 49

50 X4170 VMA 3205, 4K X4170 VMA 3205, 8K Database Performance OLTP insert transaction, 2 rows per tx Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E copyright 2012 by benchware.ch slide 50

51 Throughput in [tps] Database Performance OLTP update transaction, primary key access, 32 rows per tx 1'600 1'400 1'200 User: 2 ms DBWR: < 500 µs LGWR: 2 ms Numbers at measuring point: avg service time for single block I/O user process and scattered write requests of DBWR and LGWR processes 1' User: DBWR: LGWR: 3 ms 1 ms 3 ms X4170, VMA 3205, 8K Customer System X4170, VMA 3205, 4K Degree of parallelism (dop) avg SQL service time < 0.09s at DOP 128 for 4K block size avg SQL service time < 0.14s at DOP 128 for 8k block size copyright 2012 by benchware.ch slide 51

52 X4170 VMA 3205, 4K X4170 VMA 3205, 8K Database Performance OLTP update transaction, primary key access, 32 rows per tx Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E T E E Rows/sec Ops/sec CPU read write Total read write Total REDO Time Run Tst Code #N #J #T [rps] [ops] [%] [iops] [iops] [iops] [MBps] [MBps] [MBps] [MBps] [sec] T E E T E E T E E T E E T E E T E E T E E T E E copyright 2012 by benchware.ch slide 52

53 Benchmark Results Summary Database Performance: OLTP transactions Metric Customer System Sun X4170 Violin VMA 3205 OLTP select tx, 32 rows per tx Throughput SQL service time OLTP insert tx, 2 rows per tx Throughpu t SQL service time OLTP update tx, 32 rows/tx Throughput SQL service time [tps] [rps] [s] [tps] [rps] [s] [tps] [rps] [s] 8K / 4K / / / / / < / < / / / 0.09 copyright 2012 by benchware.ch slide 53

54 Contents 1 About Benchware 2 CPU Performance 3 Server Performance 4 Storage Performance 5 Database Performance 6 Conclusion copyright 2012 by benchware.ch slide 54

55 Conclusion Sun X4170 with Violin VMA 3205 All prices are list prices (spring 2012) Violin VMA VIMM s x 64 GByte 5.25 TB usable RAID-3 ~ 150k USD Oracle Standard Edition ~ 35k USD single instance ~ 70k USD RAC cluster Oracle Enterprise Edition Without any option ~ 285k USD single instance Sun X4170 M2 2 sockets, 2.53 GHz 12 cores, 24 threads 96 GB RAM 2 x 8 Gb FC HBA Solaris 11 ~ 10k USD single server ~ 20k USD RAC cluster 0 Server Storage Database License Database License copyright 2012 by benchware.ch slide 55

56 Conclusion Sun X4170 with Violin VMA 3205 Best-of-breed platform - Fast x86 CPU - Cost-effective 2 socket server - Enterprise class operating system Solaris - Moderate database license cost with Oracle Standard Edition; even in a high available Real Application Cluster configuration - Extraordinary I/O throughput and excellent low latency with Violin flash storage copyright 2012 by benchware.ch slide 56

57 Conclusion Sun X4170 with Violin VMA 3205 Superior storage performance - For all access pattern: random read/write, sequential read/write - Sustained I/O performance for all database objects, not only cached objects Bottom line - Excellent database server for OLTP environments - Outstanding key performance metrics - Outstanding price performance ratios copyright 2012 by benchware.ch slide 57

58 swiss precision in performance measurement

59 Further Reading [1] Consider Oracle Standard Edition to Reduce Database Management System Costs; Gartner Research Note G , Donald Feinberg, Jane B. Disbrow, March 2010 [2] Oracle Standard Edition als Alternative zur Enterprise Edition; DOAG Presentation 2011 (only in German); Johannes Ahrends, copyright 2012 by benchware.ch slide 59