Scalability and High Performance with MicroStrategy 10

Scalability and High Performance with MicroStrategy 10 Enterprise Analytics and Mobility at Scale.

Copyright Information All Contents Copyright 2017 MicroStrategy Incorporated. All Rights Reserved. Trademark Information The following are either trademarks or registered trademarks of MicroStrategy Incorporated or its affiliates in the United States and certain other countries: MicroStrategy, MicroStrategy 10, MicroStrategy 10 Secure Enterprise, MicroStrategy 9, MicroStrategy 9s, MicroStrategy Analytics, MicroStrategy Analytics Platform, MicroStrategy Desktop, MicroStrategy Operations Manager, MicroStrategy Analytics Enterprise, MicroStrategy Evaluation Edition, MicroStrategy Secure Enterprise, MicroStrategy Web, MicroStrategy Mobile, MicroStrategy Server, MicroStrategy Parallel Relational In-Memory Engine (MicroStrategy PRIME), MicroStrategy MultiSource, MicroStrategy OLAP Services, MicroStrategy Intelligence Server, MicroStrategy Intelligence Server Universal, MicroStrategy Distribution Services, MicroStrategy Report Services, MicroStrategy Transaction Services, MicroStrategy Visual Insight, MicroStrategy Web Reporter, MicroStrategy Web Analyst, MicroStrategy Office, MicroStrategy Data Mining Services, MicroStrategy Narrowcast Server, MicroStrategy Health Center, MicroStrategy Analyst, MicroStrategy Developer, MicroStrategy Web Professional, MicroStrategy Architect, MicroStrategy SDK, MicroStrategy Command Manager, MicroStrategy Enterprise Manager, MicroStrategy Object Manager, MicroStrategy Integrity Manager, MicroStrategy System Manager, MicroStrategy Analytics App, MicroStrategy Mobile App, MicroStrategy Tech Support App, MicroStrategy Mobile App Platform, MicroStrategy Cloud, MicroStrategy R Integration, Dossier, Usher, MicroStrategy Usher, Usher Badge, Usher Security, Usher Security Server, Usher Mobile, Usher Analytics, Usher Network Manager, Usher Professional, MicroStrategy Services, MicroStrategy Professional Services, MicroStrategy Consulting, MicroStrategy Customer Services, MicroStrategy Education, MicroStrategy University, MicroStrategy Managed Services, BI QuickStrike, Mobile QuickStrike, Transaction Services QuickStrike Perennial Education Pass, MicroStrategy Web Based Training (WBT), MicroStrategy World, Best in Business Intelligence, Pixel Perfect, Global Delivery Center, Direct Connect, Enterprise Grade Security For Every Business, Build Your Own Business Apps, Code-Free, Welcome to Ideal, The World s Most Comprehensive Analytics Platform, The World s Most Comprehensive Analytics Platform. Period. Other product and company names mentioned herein may be the trademarks of their respective owners. Specifications subject to change without notice. MicroStrategy is not responsible for errors or omissions. MicroStrategy makes no warranties or commitments concerning the availability of future products or versions that may be planned or under development. Patent Information This product is patented. One or more of the following patents may apply to the product sold herein: U.S. Patent Nos. 6,154,766, 6,173,310, 6,260,050, 6,263,051, 6,269,393, 6,279,033, 6,567,796, 6,587,547, 6,606,596, 6,658,093, 6,658,432, 6,662,195, 6,671,715, 6,691,100, 6,694,316, 6,697,808, 6,704,723, 6,741,980, 6,765,997, 6,768,788, 6,772,137, 6,788,768, 6,798,867, 6,801,910, 6,820,073, 6,829,334, 6,836,537, 6,850,603, 6,859,798, 6,873,693, 6,885,734, 6,940,953, 6,964,012, 6,977,992, 6,996,568, 6,996,569, 7,003,512, 7,010,518, 7,016,480, 7,020,251, 7,039,165, 7,082,422, 7,113,993, 7,127,403, 7,174,349, 7,181,417, 7,194,457, 7,197,461, 7,228,303, 7,260,577, 7,266,181, 7,272,212, 7,302,639, 7,324,942, 7,330,847, 7,340,040, 7,356,758, 7,356,840, 7,415,438, 7,428,302, 7,430,562, 7,440,898, 7,486,780, 7,509,671, 7,516,181, 7,559,048, 7,574,376, 7,617,201, 7,725,811, 7,801,967, 7,836,178, 7,861,161, 7,861,253, 7,881,443, 7,925,616, 7,945,584, 7,970,782, 8,005,870, 8,051,168, 8,051,369, 8,094,788, 8,130,918, 8,296,287, 8,321,411, 8,452,755, 8,521,733, 8,522,192, 8,577,902, 8,606,813, 8,607,138, 8,645,313, 8,761,659, 8,775,807, 8,782,083, 8,812,490, 8,832,588, 8,943,044, 8,943,187. 8,958,537, 8,966,597, 8,983,440, 8,984,274, 8,984,288, 8,995,628, 9,027,099, 9,027,105, 9,037, 577, 9,038,152, 9,076,006, 9,086,837, 9,116,954, 9,124,630, 9,154,303, 9,154,486, 9,160,727, 9,166,986, 9,171,073, 9,172,699, 9,173,101, 9,183, 317, 9,195,814, 9,208,213, 9,208,444, 9,262,481, 9,264,415, 9,264,480, 9,269,358, 9,275,127, 9,292,571, 9,300,646, 9,311,683 9,313,206, 9,330,174, 9,338,157, 9,361,392, 9,378,386, 9,386,416, 9,391,782, 9,397,838, 9,397,980, 9,405,804, 9,413,710, 9,413,794, 9,430,629, 9,432,808, 9,438,597, 9,444,805, 9,450,942, 9,450,958, 9,454,594, 9,507,755, 9,513,770, 9,516,018, 9,529,850, 9,563,761, 9,565,175, 9,608,970, 9,640,001, 9,646,165, 9,680,908, and 9,697,350. Other patent applications are pending.

TABLE OF CONTENTS OVERVIEW 1 CRITICAL MEASURES FOR PERFORMANCE 1 Scalability 1 Speed 1 PERFORMANCE BENCHMARK TESTS 2 Introduction 2 Test methodology 2 MicroStrategy Desktop 3 MICROSTRATEGY WEB 4 Assumptions 4 Specifications 4 Analytic complexity test 4 Results 5 IN-MEMORY ANALYTICS 6 Assumptions 7 Specifications 7 Analytic complexity test 7 Version 9.4 versus MicroStrategy 10: Test compares in-memory capabilities for Version 10, with and without partitioning 7 Results 7 In-memory cube publication 8 Specifications 8 MICROSTRATEGY MOBILE 9 Host machine 9 iserver 9 Mobile server 9 Assumptions for ipad/iphone 9 ipad results 9 iphone results 9 MICROSTRATEGY DESKTOP 10 Specifications 10 Assumptions 10 Results 10 INDEX 11 *Words denoted by an asterisk in this document can be found in the Index. These terms offer context to help you better understand the high-performance tests run by MicroStrategy for our various solutions within this white paper.

Scalability and High Performance with MicroStrategy 10 Enterprise Analytics and Mobility at Scale When it comes to business intelligence and analytics, high performance and scalability are no longer luxuries they are requirements for modern organizations. The obsession with performance is driven in large part by user expectations. Research has shown that most users typically only have an eight-second attention span, making it imperative for software to deliver the fastest possible response times and query performance. ** MicroStrategy 10 is an enterprise analytics and mobility platform that s architected from the ground up to deliver results with sub-second response times and to scale to hundreds of thousands of users without any restrictions on data size. The platform s unified architecture and industry-leading in-memory capabilities help MicroStrategy 10 offer best-in-class performance. Overview In today s increasingly data-driven world, more and more business users depend on analytics to do their jobs. As these users continue to adopt data discovery and self-service analytics, the BI teams that serve them need to deliver a platform that can meet their performance and scalability requirements. A truly enterprise-grade BI platform should deliver best-in-class response times even under the strain of large user or data volumes. This document evaluates the performance of MicroStrategy across all its product offerings. Note: tests have been performed with MicroStrategy 10.4 our latest platform release. Critical measures for performance Scalability Data scalability*: The capability of a system to manage and handle a growing volume of work, or its potential to be enlarged to accommodate that growth. User scalability*: The ability of a system to manage and handle a growing volume of users. Speed Response time*: The total amount of time it takes for a database query to load data as a dashboard or report. Throughput*: The maximum rate that something is processed. This can be affected by both processing power and user behavior. **https://advertising.microsoft.com/en/wwdocs/user/display/cl/researchreport/31966/en/microsoft-attention-spans-research-report.pdf microstrategy.com 1

Performance benchmark tests Introduction MicroStrategy is a flexible, robust platform designed to scale to meet the needs of any analytical deployment. Our long-standing customers have used the platform across a variety of deployment scenarios that vary in data volume, number of users, and data model complexity. MicroStrategy is highly scalable the same architecture can serve departments or SMBs and also meet the demands of larger enterprise-wide deployments. Organizations can start small with MicroStrategy and grow as their needs change, without any restriction or compromises. This document provides information to help you quantify and assess the performance and scalability of MicroStrategy 10. Benchmark tests were conducted across MicroStrategy Web, Mobile, and Desktop. These tests were designed to simulate real-world usage based on MicroStrategy customer implementations. Each section in this document is dedicated to exploring the assumptions and results of a specific MicroStrategy product or capability. Test methodology Test scripts based on Apache JMeter* simulate typical MicroStrategy usage scenarios e.g., multiple users log on, browse folders, run dashboards, and perform various manipulations. Simulated users are increased over time, and the number of jobs at each data point as well as the end-to-end response time for each user interaction is measured. Measurement is continued for one hour for each data point to ensure reliable and statistically significant results. MicroStrategy Web A browser-based interface for building and consuming analytical applications. MicroStrategy Desktop A self-service single user tool for data discovery. MicroStrategy Mobile A platform for deploying BI to tablets and smartphones and building custom mobile productivity applications. microstrategy.com 2

MicroStrategy Web and Mobile MicroStrategy Web and Mobile leverage the power of the MicroStrategy Server, which now offers greater integrated in-memory capabilities with version 10. The benchmark tests for MicroStrategy Web and Mobile have been designed to push the limits of the platform with higher data volumes and heavier user loads. The results of these tests showcase the scalability of the platform, and MicroStrategy s ability to maintain high throughput levels and low response times even under heavy strain. MicroStrategy Desktop MicroStrategy Desktop is a data discovery tool that individual users can download and use on their own personal PCs or Macs. The section on Desktop highlights the performance test results conducted for single users and studies how quickly users can connect to data sources and build dashboards. Client Web browser Mobile app Web server MicroStrategy Web Mobile server MicroStrategy Mobile BI server MicroStrategy Server Data assets microstrategy.com 3

MicroStrategy Web The MicroStrategy platform offers interfaces that provide a fast, intuitive experience for every user, from business analysts and IT developers, to project administrators. MicroStrategy Web delivers analytics through any modern web browser, allowing users to easily connect to the MicroStrategy Server and access a broad spectrum of analytical capabilities. This includes viewing and interacting with reports and dashboards, conducting ad hoc data discovery, performing in-memory cube publication, and setting up mass document distribution. To help quantify the scalability of MicroStrategy Web, our in-house performance tests simulated typical use cases and were conducted using a standard benchmark application. The simulated workload included a mix of report and dashboard executions, ad hoc data discovery interactions, cube publications, and history list manipulations with up to 125,000 users. Multiple (up to eight) server nodes* were clustered together as part of the standard benchmark test application to study the capacity of more than one MicroStrategy Server. The results of our benchmark testing demonstrate that MicroStrategy approaches linear scalability* under increasing user loads across multiple server nodes, and is able to maintain sub-second response times for all user actions. Testing saw roughly 25,000 users consistently performing one action every three minutes, and resulting in 7,500 reports executed every minute with sub-second response time. Assumptions Each Server node has a four CPU configuration. 20% of the total user* population was active during peak usage Active users performed an average of one action every three minutes Response times indicated in the table only measure the server response time and do not include client rendering time Specifications Client - Web browser Load: JMeter Server: Windows Server 2008 R2 Enterprise CPUs: 32 cores (Intel(R) Xeon(R) CPU E5-2695 v2 @2.40GHz Memory: 200GB Web server - MicroStrategy Web MicroStrategy: MicroStrategy 10.4 Op Sys: CentOS release 6.5 (Final) Server: apache-tomcat-7.0.68 CPUs: 48 cores Intel(R) Xeon (R) CPU E5-2695 v2 @ 2.40GHz Memory: 250G BI server - MicroStrategy Intelligence Server MicroStrategy: MicroStrategy 10.4 Op Sys: Cent OS release 6.5 (Final) CPUs: 4 cores Intel(R) Xeon(R) CPu E5-2695 v2 @2.40GHz Memory: 250G Analytic complexity test Report and document cache size ranged from 1.1 MB to 23 MB Cube size: 0.9 MB - 1.3 MB microstrategy.com 4

Asymmetric Configuration with a 0.9 average response time Nodes Throughput (rpm) Active users Total users 1 Node 1,052 3,412 17,059 4 Nodes 4,166 13,511 67,557 8 Nodes 7,272 25,061 125,303 Results Response time for all user actions were under one second. Nearly linear user and throughput scalability across multiple Server nodes. ~25,000 active users* (~125,000 total users) were able to execute dashboards and publish cubes without experiencing slowdowns in response time. Throughput across multiple server nodes in a cluster User scalability across multiple server nodes in a cluster 125,303 67,557 25,061 17,059 13,511 3,412 microstrategy.com 5

In-memory analytics MicroStrategy 10 delivers next-generation in-memory analytics fueled by powerful parallel processing* and partitioning capabilities. These breakthrough in-memory capabilities are tightly integrated with the MicroStrategy Server and help MicroStrategy deliver fast performance for both Web and Mobile users. Often referred to as PRIME (Parallel Relational In-Memory Engine), MicroStrategy s highly integrated in-memory architecture* delivers three key differentiators with respect to enterprise scalability and performance: Parallel connections to enterprise assets and data warehouses that lead to faster data fetch and cube publishing Partitioned cubes that improve memory management and deliver a highly flexible cube schema that avoids unnecessary pre-joins to deliver faster in-memory processing MicroStrategy supports the industry s largest in-memory cubes that can now exceed two billion rows In-house performance tests were conducted using a standard benchmark application that was designed to test the performance implications of in-memory cube functionality. They were completed in a controlled environment to highlight the differences in data size limits between MicroStrategy version 9.x and 10. The test was similar in design to the previous test for MicroStrategy Web, but with added complexity to the data model and larger data volumes. The added complexity assumed that in-memory features typically cache the data in the Server and deliver faster results. Complex analytical reporting tests were conducted against in-memory cubes that held 50 million rows and 45 columns. The test included nearly 60 reports running against the described cubes, ranging in size up to three million rows. microstrategy.com 6

Assumptions 20% of the total users population was active during peak usage. Active users performed an average of one action every three minutes. Partitioning was based on data in the business model. Specifications Client - web browser Load test tool: JMeter CPU speed: Intel Xeon X5560 @2.8GHz Memory: 16GB Test Environment Details Web server - MicroStrategy Web MicroStrategy: MicroStrategy 10 and 9.4 Op Sys: CentOS 5.10 Number of physical cores: 12 CPU speed: Intel Xeon E5-2695 @2.8GHz Memory: 61GB App server: Tomcat 6 BI server - MicroStrategy Intelligence Server MicroStrategy: MicroStrategy 10 and 9.4 Op Sys: RH Enterprise 6.3 Number of physical cores: 8 CPU speed: Intel Xeon X5560 @2.8GHz Memory: 144GB The cube returned data of 50 million rows and 45 columns. 57 view reports returned data that ranged from 1 to 2,901,706 rows. On average, each report contained 56,257 rows. Version 9.4 versus MicroStrategy 10: This test compares MicroStrategy 9.4 versus MicroStrategy 10, both with and without partitioning Response time (sec) 3.50 2.50 1.50 0.50 3.61 2.50 1.68 Throughput (rpm) 1K 100 10 1,400 344 393 394 Results MicroStrategy 10 s in-memory processing power performs 30% faster than version 9.4. MicroStrategy 10 supports more users than version 9.4 and delivers higher throughput Active users 1,000 600 1,116 1,275 1,278 with lower response times.* MicroStrategy 10 s in-memory cube partitioning 200 performs 50% better under high concurrency. Version 9.4.1 10 10 - Partitioned cube microstrategy.com 7

In-memory cube publication To benefit from MicroStrategy s in-memory capabilities and processing speed, data from the warehouse is published to the in-memory layer of the MicroStrategy Server. MicroStrategy 10 allows administrators to publish cubes with billions of rows, exceeding any limitations in previous versions. And because data can be loaded in-parallel from the warehouse using multiple threads, the speed at which cubes can be loaded dramatically increases with MicroStrategy 10. The following tests were designed to compare the time needed to publish and process large in-memory cubes between MicroStrategy 9.4 and MicroStrategy 10. Results show that MicroStrategy 10 can scale up to data sizes 80X greater than previous versions. The tests also highlight that the response time has significantly improved while processing data within these cubes in MicroStrategy 10. MicroStrategy 9.4 was unable to run cubes with over two billion rows regardless of the number of CPU cores. Version 10 eliminated that limitation and has been able to publish cubes in excess of two billion rows. RESULTS Create and publish a cube - data scalability* without restrictions 9.4.1: OLAP Services Data: 2.3B rows Failed due to 2 billion row limit VS MSTR 10: 8Thread Partitioning Data: 2.3B rows Cube size: 265GB Data analysis - 5x faster data interactions 9.4.1: OLAP Services Data: 100M rows Response time: 0:09:51 VS MSTR 10: 8Thread Partitioning Data: 100M rows Response time: 0:01:59 Specifications Client - web browser Number of physical cores: 2 CPU speed: Intel Xeon X5560 @2.8GHz Memory: 2GB Browser: Chrome and Firefox Web server - MicroStrategy Web MicroStrategy: MicroStrategy 10 and 9.4 Op Sys: RH Enterprise 6.3 Number of physical cores: 4 CPU speed: Intel Xeon X5560 @2.8GHz Memory: 15GB App server: IBM Websphere BI server - MicroStrategy Intelligence Server MicroStrategy: MicroStrategy 10 and 9.4 Op Sys: RH Enterprise 5.10 Number of physical cores: 8 CPU speed: Intel Xeon X5560 @2.8GHz Memory: 188GB microstrategy.com 8

MicroStrategy Mobile Host machine Number of physical cores: 24 CPU Speed: Intel(R) Xeon(R) 2 x E5-2695 v2@ 2.40GHz Number of processors: 28 iserver Number of virtual processors: 16 Memory: 96 GB Ops Sys: Windows Server 2012 R2 Mobile server Number of virtual processors: 24 Memory: 64 GB Ops Sys: CentOS release 5.10 Assumptions for ipad/iphone Concurrent users are people who interact with BI applications during peak hours. Active users make up 20% of the total named user count. Response times only measure the server response time and do not include client rendering time. Active users perform, on average, one action every three minutes. ipad Performance tests were conducted on an ipad app. The app included various mobile dashboard execution scenarios. The test application ranged from 2,000 to over 32,000 active users accessing in-memory datasets. The results of this test show that up to 30,000 active users were able to access over 10,000 report per minute (RPM) with an average response time not exceeding.2 seconds. ipad results Sub-second response times Test supported 29,140 active users (145,700 actual users) Version Optimal throughput (rpm) Response time (sec) Peak power rating of 540 kilocycles The test was nearing around 80% usage while the mobile server had around 10% CPU usage MicroStrategy 10 10,762 0.194 iphone Performance tests were conducted on an iphone. The app includes various mobile dashboard execution scenarios. The test application ranged from 2,000 to over 30,000 active users accessing in-memory datasets. The results of this test show that up to 30,000 active users were able to access nearly 10,000 report per minute (RPM) with an average response time of.268 seconds. iphone results Sub-second response times Test support 27,090 active users (135,450 actual users) Peak power rating of 510 kilocycles Version Optimal throughput (rpm) Response time (sec) When the test was nearing the peak throughput, the I-Server CPU stressed around 80% usage while the mobile server had around 10% CPU usage MicroStrategy 10 9,968 0.268 microstrategy.com 9

MicroStrategy Desktop MicroStrategy Desktop is a powerful self-service data discovery tool available on Mac or PC. With Desktop, users can connect to, blend, and prepare data from multiple sources to build dashboards for data exploration. Unlike MicroStrategy Web, Desktop does not require a connection to the MicroStrategy Server. However, if connected, users are able to upload reports and dashboards to the Server to share content and download reports and dashboards from the Server to work offline. Though the processing speed of an individual PC or Mac affects data discovery, performance or response time is a critical factor for business users who engage with data discovery tools. In-house performance tests were conducted to examine the response time for various actions typically performed by business users for ad hoc data discovery with Desktop. The test scenario was based on the dataset provided to MicroStrategy as a participant in Gartner s 2016 BI Bake Off. User actions included: importing data from Excel, preparing and blending data, visualizing and interacting with data with ad hoc data exploration, and saving and exporting dashboards. The results demonstrate that MicroStrategy Desktop allows users to access raw data and visualize information in under a minute. The table below highlights the details of the actions and the average duration for each action. Specifications OS: Windows 7 64-bit Memory: 16GB Processor: Intel Core i7-4770 CPU @3.4GHz Assumptions One Excel dataset that contains 7,800 rows and is 1.8MB. A second Excel file that contains 250 rows and is 4.8 MB. Both files were modified for this test scenario. Results Category Step details End to end time(s) Import data Import CSV file = 1.8 MB 0.75 Change attributes to metrics 0.59 Profile data 0.62 Rename columns 0.58 Clean and edit data Loading data for wrangling 1.71 Change data type 1.06 Remove characters NULL 2.12 Finish data wrangle 0.59 Add a new table-excel file = 4.8 MB 0.91 Select sheets to be imported to Desktop 0.56 Parse data 0.55 Blend data Rearrange parsed data 0.53 Click apply 0.26 Exclude columns from the data import process 0.57 Finish 3.18 Combine dashboards Import a.mstr dashboards = 1.2MB 1.04 Drag and drop objects 0.73 Create Change visualization 1.10 visualization Lasso select to keep only 1.56 and play Sort descending 0.71 Swap 0.88 Share Export as image 0.57 Save as.mstr file 0.60 TOTAL 21.77 Gartner BI Bake Off 2016 MicroStrategy Desktop can access, blend, and visualize data in a standard Excel file quickly and efficiently. Desktop allows users to create powerful dashboards in less than five minutes. The table on the left breaks down the end-to-end times for each step. microstrategy.com 10

Glossary These terms offer context to help you better understand the high performance benchmark tests discussed within this white paper. Active user These are users who are actively engaging with the platform rather than just leaving the application open. For example, they may execute dashboards or publish cubes. Parallel processing A method of powering data analytics that breaks up large data requests into more manageable sizes by distributing them to multiple processors that run simultaneously. Apache JMeter An open-source, pure Java application that tests functional behavior to measure the performance of static and dynamic resources web services (SOAP/REST), web dynamic languages (PHP, Java, ASP.NET, Files, etc.), Java Objects, Databases and Queries, FTP Servers, and more. MicroStrategy used it to simulate heavy server loads and test platform performance. Architecture A model that governs and defines how data is collected, stored, and managed in database systems. Data scalability The capability of a system to manage and handle a growing amount of work, or its potential to be enlarged to accommodate that growth. Linearity The ability of a platform to test consistently and deviate minimally from a straight line projection. Node A connection or redistribution point that creates, receives, or transmits information from one point and redirects it to another. Response Time The total amount of time it takes for a database query to load data as a dashboard or report. Reports per minute (RPM) The standard unit of measurement listed in our mobile and web test results. Throughput The maximum rate that something is processed. This can be affected by both processing power and user behavior. Total users In a typical environment, most individuals log into the application but do not run a dashboard, remaining passive users. It is common for an average of 10% of total users to be active at any given time. To increase the complexity of the test scenarios discussed in this document, we have assumed that 20% of the total user population are active users. User scalability The ability of a system to manage and handle a growing volume of users. microstrategy.com 11

1850 Towers Crescent Plaza Tysons Corner, VA 22182 Copyright 2017. All Rights Reserved. microstrategy.com COLL-1646 0717