Dell EqualLogic Best Practices Series Performance Baseline for Deploying ® Oracle 11g Release 2 Based Decision Support Systems using Dell™ EqualLogic™ PS6110XV Arrays A Supplemental Test Report Base White paper: Sizing and Best Practices for Deploying Oracle 11g Release 2 Based Decision Support Systems using Dell EqualLogic PS6110XV Arrays A Dell Technical White paper http://en.community.dell.com/dell-groups/dtcmedia/m/mediagallery/20059328/download.
THIS TEST REPORT IS FOR INFORMATIONAL PURPOSES ONLY, AND MAY CONTAIN TYPOGRAPHICAL ERRORS AND TECHNICAL INACCURACIES. THE CONTENT IS PROVIDED AS IS, WITHOUT EXPRESS OR IMPLIED WARRANTIES OF ANY KIND. © 2012 Dell Inc. All rights reserved. Reproduction of this material in any manner whatsoever without the express written permission of Dell Inc. is strictly forbidden. For more information, contact Dell.
Table of Contents 1 Introduction ............................................................................................................................................................................ 2 1.1 2 Audience ................................................................................................................................................................................ 2 DSS I/O simulation ........................................................................................
Acknowledgements This whitepaper was produced by the PG Storage Infrastructure and Solutions of Dell Inc.
1 Introduction This paper provides a summary of results for tests executed using EqualLogic PS6110XV arrays. These tests were set up the same as and run in addition to the set of test scenarios described in the Dell white paper titled, Sizing and Best Practices for Deploying Oracle 11g Release 2 Based Decision Support Systems with Dell EqualLogic 10GbE iSCSI SAN. All tests executed as a part of that paper were executed using PS6010XV arrays.
2 DSS I/O simulation 2.1 Test methodology A subset of I/O simulation tests were executed with large sized I/O block requests simulating a DSS database I/O to understand the read and write I/O throughput rate offered by the EqualLogic PS6110XV storage arrays. 2.2 Test scenarios The main objectives of these tests were to determine the difference in performance of PS6110XV arrays compared to the older generation PS6010XV arrays.
The following chart compares the throughput produced by PS60110XV and PS6110XV arrays in RAID 10 configuration. Figure 1 ORION I/O throughput in RAID10 configuration: PS6010XV versus PS6110XV PS6110XV arrays provided better throughput than the PS6010XV arrays as shown in the above chart. The behavior can primarily be attributed to the increased number of disks hosting active data on PS6110XV arrays.
2.3.2 PS6010XV and PS6110XV – RAID 50 The four ORION I/O workload scenarios described earlier were executed on a PS6110XV array and the results were compared with a PS6010XV array. The storage arrays were configured with RAID 50 for all scenarios. The following chart compares the throughput produced by PS60110XV and PS6110XV arrays in RAID 50 configuration.
2.3.3 PS6110XV – RAID 10 versus RAID 50 The four ORION workloads described earlier were executed on a PS6110XV array in RAID 10 and in RAID 50. The I/O throughout reported by ORION for both configurations are summarized in the chart below. Figure 3 ORION I/O throughput for PS6110XV: RAID 10 versus RAID 50 The RAID 50 configuration provided a slightly better throughput than the RAID 10 configuration.
2.3.4 PS6110XV scaling studies The four ORION workloads described in section 2.2 were executed on PS6110XV arrays using RAID 50. The storage pool was configured with one array initially and then scaled to two arrays in the same storage pool. Test volumes were recreated and mounted on a server for each scaled configuration. The I/O throughput reported by ORION across the scaled configurations is shown in Figure 4.
concat option. Executing the ORION tool with this option is like running any other I/O workload simulator like VDBench or IOMeter. The following chart summarizes the performance scalability results using the ORION tool concat option. Figure 5 ORION I/O throughput on PS6110XV Scalability: Using ORION concat option The following table shows the percentage of increase in throughput for different I/O workloads when ORION was executed using the default concat option.
3 DSS application simulation 3.1 Simulation tool and workload definition Additional DSS tests were conducted to understand the throughput behavior of the PS6110XV arrays when the Oracle Real Application Cluster (RAC) database executed DSS queries. The following additional DSS query simulation tests (TPC-H), using an Oracle database, were executed on the storage configurations listed below.
As seen in the bar graph in Figure 6, both the average and maximum throughput scaled linearly with the number of arrays. The throughput from PS6110XV arrays was significantly higher compared to the PS6010XV results. The behavior can primarily be attributed to the increased number of disks hosting active data on PS6110XV arrays. The linear scaling was possible because of EqualLogic’s scale-out architecture; an array adds not only disks but also controllers and NIC ports.
Figure 7 Storage I/O throughput – I/O (ORION) and application simulation The following table shows the percentage of increase in throughput on the PS6110XV array configuration when DSS simulation tests were executed using the actual Oracle database as compared to I/O simulation. I/O Workload TPCH Workload using 5 concurrent streams. 1 x PS6010XV RAID 50 1 x PS6110XV RAID 50 % Increase 446.40 672.14 50.56 I/O Workload TPCH Workload using 5 concurrent streams.
The I/O throughput across storage array members, as reported by EqualLogic SAN HQ in the two-array configuration, is shown in Figure 8. Figure 8 Storage I/O throughput – SAN HeadQuarters The initial ramp up in Figure 8 happened during the start of the test when all five users login and start submitting queries. As the users login, the database server commences query processing and generates increasing I/O requests resulting in a ramp up of I/O throughput.
3.2.1.3 Query Completion Time The end-to-end completion time for all 22 queries submitted by the five simulated users is shown in Figure 10 below. The duration is between the time when the first query was submitted to the database and when the last query was completed. This graph also displays a comparison between PS6010XV and PS6110XV arrays with respect to query completion time.
4 Conclusion EqualLogic 10 GbE iSCSI PS Series arrays provide high levels of I/O throughput warranted by DSS. By adding arrays, PS Series scale not only capacity but also I/O throughput as well. This is due to the scale-out architecture where all array resources, including controllers and NICs, are being scaled proportionately. Optimal operation of DSS was achieved when the applicable best practices laid out in the reference white paper are adhered to.
THIS TEST REPORT IS FOR INFORMATIONAL PURPOSES ONLY, AND MAY CONTAIN TYPOGRAPHICAL ERRORS AND TECHNICAL INACCURACIES. THE CONTENT IS PROVIDED AS IS, WITHOUT EXPRESS OR IMPLIED WARRANTIES OF ANY KIND.