1、Putting It All Together: Trends in Business Intelligence,Claudia Imhoff, PhD Intelligent Solutions, Inc. CImhoffIntelS Blog: http:/www.b-eye- Imhoff,President and Founder Intelligent Solutions, Inc. A thought leader, visionary, and practitioner in the rapidly growing fields of business intelligence
2、 and customer focused-strategy Claudia Imhoff, Ph.D., is an internationally recognized expert on analytical CRM, business intelligence, and the infrastructure to support these initiatives the Corporate Information Factory (CIF). Dr. Imhoff has co-authored five highly-regarded and popular books on th
3、ese subjects and writes monthly columns (totaling more than 100) for technical and business magazines.,Email: Phone: 303-444-6650,Putting It All Together,Going Beyond Traditional BI Operational BI Takes the Stage Data Warehouse Appliances and Analytic Databases Making Life Simpler BI Software as a
4、Service Feeling SaaS-y? Open Source BI Free Software Anyone?,3,4,The Three Levels of Business Intelligence,Strategic BI timeframe months,Tactical BI timeframe days or weeks,Operational BI timeframe is intra-day,The Three Levels of Business Intelligence,Paradigm Shift,5,6,What is Operational BI*,A se
5、t of services, applications and technologies for monitoring, reporting on, analyzing and managing the business performance of an organizations daily business operations,*From research study. “Embedded BI”, written by Colin White and Judy Davis, www.B-EYE-R,7,Operational BI Answers to Day-to-Day Busi
6、ness Questions,Information on demandReal-time + historical dataAccess to SAP, Siebel, Oracle and BI results,New Data Needs,pickedpackedshippedinvoiced,What is my customers order status? What can I offer based on customers life-time value?,Can I afford to make this move at current margin rates?,What
7、is my current inventory level world wide? Is it sufficient to meet demands?,What is my production yield right now? Am I at par with acceptable standards?,8,Real Time Decision-Making*,Operational BI optimizes time latency between when a business event occurs and when an appropriate action is taken Th
8、e goal to “right-size” the decision-making cycle Compressing time lag between knowing what is happening and taking action based on that knowledge Real-time must consider potential trade-off between time-to-action and business value of actions,* From “Right-Time Business Intelligence: Optimizing the
9、Business Decision Cycle” By Judy Davis, www. B-EYE-N Research paper,Impact on BI Environment,History of BI Extract usable information from operational systems Users, technologies, processes, procedures all independent of operations Now what? Impact on BI environment is significant Increase in number
10、 of users, volume of data, and faster performance Operational BI MUST be integrated into the operational environment Requires understanding of operational systems, processes, procedures, workflows, personnel,9,10,Impact on BI Environment,Numbers of users increase significantly Traditional BI rarely
11、supported a few hundred, maybe a thousand or so users Opening BI up to operational personnel means ramping up into tens of thousands of users These users have very different interface requirements Means BI implementers must rethink how BI is delivered to business users Means tighter and faster conne
12、ctivity of enterprise decision support environment to rest of the company.,11,Impact on BI Environment,Volumes of data increase substantially Detailed intraday snapshots of data are loaded or trickle-fed into data warehouses Tens of terabytes to hundreds of terabytes are not unusual storage requirem
13、ents for operational BI Scalability now a mandatory requirement in any BI technology Whether in processing and integration of data, storage of massive volumes, or retrieval of query responses,12,Impact on BI Environment,Faster performance Query performance must mimic or emulate response times in ope
14、rational systems Sub-second to just a few seconds to return data from a query. Ability to prioritize queries not only according to their importance but also their response requirements is mandatory success criterion This last feature has stumped many BI implementers and BI vendors Must have ability
15、to handle mixed work load gracefully and simultaneously,Getting Started Assess Reality,First step perform honest assessment of existing data delivery capabilities available technologies, maturity of the BI architecture, existing personnel, etc. Combine these with solid understanding of business requ
16、irements for operational BI data Important to understand which weaknesses discovered in assessment will be exaggerated as you speed up the enterprise,13,Operational BI Requirements,Continuous availability of operational data and BI results Current information from operational systems Integrated with
17、 BI data on demand Minimal impact on operational systems performance Presented in a proactive manner Make decisions act on information presented Easy to understand and use Dynamic modeling Ability to change business rules on the fly Show different set of metrics depending on situation,14,15,Picking
18、a Project,Look for workflow activities that have significant impact on costs or revenues Bottlenecks today that can be made more efficient through use of operational BI Dont make big changes to operational processes Just speed up or make more efficient processes you already have in place You will ha
19、ve to retrain personnel and retool SOPs Project managers may not realize operational BI application has ramifications beyond projects immediate boundaries,Putting It All Together,Going Beyond Traditional BI Operational BI Takes the Stage Data Warehouse Appliances and Analytic Databases Making Life S
20、impler BI Software as a Service Feeling SaaS-y? Open Source BI Free Software Anyone?,16,Data Warehouse Appliances,BI and data warehousing technologies continue to evolve and innovate Produce more efficient & cost effective ways to deliver BI Latest innovations are DW and BI appliances Definition of
21、an appliance* One purpose One package One installation One vendor,17,* From the BEYE-R paper titled “Data Warehouse Appliances: Evolution or Revolution?” by Colin White, and Richard Hackathorn,Data Warehouse Appliances,All-in-one box that provides a hardware server preconfigured with all software co
22、mponents Designed for a specific purpose supporting data warehouse processing Offers ease of use, simplicity, and compatibility tested, ordered and delivered as a single system Simple to understand even though mechanism may be complex Low cost in terms of TCO High performance in achieving its purpos
23、e Single point of service provided by single vendor,18,Data Warehouse Appliances,Cost effective solution TCO of a data warehouse appliance is lower because cost of hardware and software is cheaper Also because simplicity and ease of reduces installation, administration and support cots Improved usab
24、ility of a data warehouse appliance means projects can be developed and deployed faster Includes popular BI capabilities Interactive dashboards, analysis, reports, alerting, and data integration,19,20,Sweet Spot for Data Warehouse Appliances,Mega- to Gigabytes,Multiple Terabytes,Mixed Purpose,Any da
25、tabase vendor,Data Warehouse Appliances,Specialized Databases (e.g., Teradata, IBM),Focused Purpose,Data Warehouse Appliances,Pros Immediate visibility & interaction into business performance Non-disruptive to existing infrastructure Faster deployment Low maintenance black box,Cons Still some opposi
26、tion to use of appliances by IT departments Loss of “control” over moving parts DW and BI appliance scalability Customization to fit each companys needs,21,Sample Data Warehouse Appliance Vendors,Netezza Teradata DATAllegro (now Microsoft) Sun + Green Plum Sun + Vertica Sun + ParAccel Sun + Kognitio
27、 IBM InfoSphere Warehouse,22,Role of Appliances in BI SaaS,Many data warehouse appliance and BI SaaS vendors are forming partnerships Gives SaaS vendors scalability, reliability, performance Gives appliance vendors applications, new markets, greater exposure Gives customers more confidence that solu
28、tion is on solid technological footing Performance Support for multi-tenancy Scalability Applications,23,Analytic Databases,Many are Massive Parallel Processing (MPP) Can use commodity hardware Many have column-based data organization Limit I/O by putting similar data together reduces reads to only
29、columns needed for query Single data type per column allows for significant compression Data compression Compression can be optimized for particular data types CPU is not the bottleneck, only I/O is,24,Analytic Databases,Built-in intelligence Allows decompression of only data that must be for query
30、resolution and ignore all others Is major factor in overall improved performance Load times remain constant regardless of table size Should also have query times that remain constant regardless of table size Bottom line technology must be seamlessly scalable,25,Analytic Databases,Many new vendors on
31、 the market (sample): Green Plum Vertica (Michael Stonebraker*) ParAccel (Barry Zane*) Dataupia (Foster Hinshaw*) InfoBright (Warsaw University) Aster Data (Stanford University) illuminate (Former Synerra Systems founders) One well-established vendor: Sybase IQ since 1993 Most are column based, MPP,
32、 shared nothing architectures (not all though),26,* Ingres and Illustra founder * Netezza founders,Analytic Databases Really Fast: TPC-H 1 TB,27,Analytic Databases Really Fast and Really Inexpensive,28,Graph compliments of Jos van Dongen, Tholis Consulting, NL. Numbers are estimated.,Analytic Databa
33、ses,Pros Excellent performance Very cost-effective Low maintenance Partnering with hardware vendors (DW appliance),Cons Many are small companies May not handle mixed work load well New (unknown) technology for IT,29,Putting It All Together,Going Beyond Traditional BI Operational BI Takes the Stage D
34、ata Warehouse Appliances and Analytic Databases Making Life Simpler BI Software as a Service Feeling SaaS-y? Open Source BI Free Software Anyone?,30,BI Delivery Models,There are two BI delivery models today On-premises traditional model Software as a Service (SaaS),31,On-premises Traditional Model,I
35、nternal IT is responsible for entire environment from first project Find excess capacity on machines Upgrade memory on existing machine for usage Leverage installed end user access tools Buy smaller platforms that can scale Migrate to bigger box when necessary Use smaller box for data mart(s) Look i
36、nto data warehouse appliances for very large, focused BI analytics,32,33,Software as a Service (SaaS),Characteristics* Secure, flexible, and efficient business processes & workflows Service level agreements Value-added business services such as analytics & best practices Extensive use of service-ori
37、ented architecture (SOA) to enable scaling, configurability, and integration Subscription monitoring & usage-based billing,* From , “Get Ready for SaaS 2.0” by Bill McNee, Saugatuck Technology, May 8. 2006,Advantages for SaaS Vendors,Vendors support only one platform and one version of the applicati
38、on No need to support multiple operating systems, platforms, and older versions of the software Decreases development costs significantly SaaS gives vendor great visibility into how their customers are actually using their software See every move, every feature, every function used by customers Give
39、s vendor great intelligence on how to build a better product based on actual usage,34,Advantages for SaaS Vendors,SaaS model gives vendor a predictable cash flow Subscription model is reliable for cash flow estimation Improves start-up estimations and growth track Vendors dont get trapped in “featur
40、e bloat” No need to keep adding feature after feature to get customers to buy new versions Create only features that are needed based on actual customer usage,35,Disadvantages for SaaS Vendors,SaaS produces lower revenues at first than traditional vendor models Must attain critical mass of subscribi
41、ng customers Vendor must have enough funding to tide them over More time is needed to ramp up to mature status Higher customer set up costsTraditional vendor model send customer a CD SaaS vendors must allocate space, set up customer support, etc. SaaS vendor becomes IT support for their customers (h
42、igher costs for customer service?),36,Disadvantages to SaaS Vendors,Customers still need ability to integrate SaaS application data with other enterprise data Need mechanism to export data out of SaaS environment Who supplies integration of SaaS data with customers other data? If customer is not SOA
43、-compliant yet, what does this mean to SaaS model?,37,Reasons for Adoption: Ease of Deployment,This is the SaaS models greatest advantage No installation of hardware No installation of software No administration of new versions of either No need for IT expertise in the tool or application Set up con
44、sists of getting a login and password set up for the business users,38,Reasons for Adoption: More Flexibility for Evolving Needs,Perhaps You can certainly change SaaS vendors quite easily If you are unhappy with one vendor, changing to another one is about as easy as getting a new login and password
45、 You can influence the direction and R & D of the current SaaS vendor You can easily add or subtract users You can easily add or subtract functionality It may not be as easy to customize the SaaS offering to your specific needs,39,Reasons for Adoption: Not Locked into Long Licenses,True Great advant
46、age in BI world where technology is moving very fast Can switch from one SaaS vendor to another But watch for cancellation fees And make sure you know what the subscription fee is based on Reduction or addition of users may be cross price break threshold S model is typical,40,Considerations for BI S
47、aaS,SaaS good at supporting particular types of users Highly mobile work force Field sales personnel Product support specialists at customer sites Telecommuters Highly geographically disbursed workforce International enterprises Non-office workers (virtual offices) Customer or partners worldwide Mus
48、t include support for various mobile devices Phones, mobile PCs, handheld devices, PDAs, etc.,41,42,Considerations for BI SaaS,Ensuring quality of delivered environment Correct mappings, verified data lineage, transformations Sufficient data quality processing Data represented in analytic engine cor
49、rectly Appropriate presentation of information, e.g., personalized dashboards Scalability of environment Data volumes small beginnings to 100s of terabytes? From a few users to 1000s,43,Considerations for BI SaaS,Performance From simple to complex queries Response times operational to strategic BI Getting right data to right people at right time Open Architecture Compliance with best practices? Non-proprietary infrastructures? Integration with existing infrastructure?,
