Informix IDS vs Oracle: A Competitive Comparison Goal of the Presentation • Present Architectural Differences between Informix and Oracle • Show Informix’s superiority in: – OLTP – Data Warehousing / Decision Support – Content Management and Web Applications • Show clear business benefits of technology differences – Performance & scalability – Extensibility – Lower cost of ownership Topics • • • • Architecture OLTP Data Warehousing Complex Data / Content Management • Summary At the Core: Informix Dynamic Scalable Architecture • Consistent code line across OWS, ODS, XPS, and IUS • Advanced functionality integrated at the core • Connectivity across the enterprise • Integrated Sytems Management and Replication Workgroups/ Desktops SMP MPP Cluster Massively Parallel Cluster Workgroup Server/ Workstation DynamicSMP Server Universal Workgroups Server Extended Parallel Server Informix Server Architecture • • • ... • Virtual Processors CPU CPU ... CPU ... CPU DB Buffer Cache Shared Data Shared Memory • • • • • Full parallel processing Efficient internal multithreading Open API for user defined functions and datatypes Configurable pool of virtual processors No context switching overhead Dynamic system administration Dynamic load balancing DB level prioritization, scheduling & locking Database migration Industry View of Informix Architecture “ The general industry consensus among users and developers is that Informix has the best architecture.” Patricia Seybold Group, 12/96 • Informix RDBMS was completely rearchitected in 1991 for parallel processing and extensibility • Informix DSA sets the industry standard for simplicity, elegance and flexibility Advantages of DSA’s Parallelism Oracle DBMS Application Application Query SQL Database Server Informix OnLine DSA Database Server Query SQL Task Task IDLE IDLE CPU CPU Subtask Subtask Subtask CPU CPU 30% CPU usage CPU CPU Near 100% CPU usage! Industry View of Oracle Architecture “Oracle has never really provided the most elegant database technology …” Hambrecht & Quist 2/97 “Oracle will need to perform a significant redesign of its DBMS engine...” Gartner Group, 3/96 • Oracle’s database architecture was designed and written in the early 1980s • It is widely believed that Oracle needs to rewrite its DBMS engine Oracle’s Architectural Approach Oracle Paralle Server Oracle Personal Lite Video Server Network Computing Architectur e Oracle 8 Oracle 6 Rdb IRI Oracle 7 Replication Engine • Informix has one architecture: DSA • In constrast, Oracle has: – Multiple incompatible servers – User written integration code – Incomplete parallel processing – High OS overhead – No API for Database Exensibility – Unclear object strategy and distributed computing vision Architectural Comparison: Performance/Scalability • Informix – Consistent high performance across all platforms and applications – Best return on hardware and application investment • Oracle – Performance dependent on servers and configuration – ROI dependent on OS, platform and application Architectural Comparison: Extensibility • Informix – Adaptable API for business specific data and processing – Multiple integrated datatypes from leading vendors – Easily adjusts to new business requirements • Oracle – No user-defined datatypes or functions – Only useful for Numeric, Alpha, Date Order – New business requirements may require extensive coding or rewrite Architectural Comparison: Cost of Ownership • Informix – Single, integrated server from desktop to data warehouse – Lower cost of maintenance and training – Architectural simplicity means faster deployment; better ROI • Oracle – Multiple non-integrated servers • Personal Oracle, Oracle 7.x/8.x, Oracle Express, etc... – Requires more maintenance; higher level of training – Complexity slows deployment, lessens ROI Customer Testimonials: Cost of Ownership • Wells Fargo converted from Oracle databases to Informix. – Working with both vendors showed “…it takes about 60% longer to deploy a project on Oracle than on Informix.” • Total Cost of Ownership wins over Oracle – Bank of America – GTE Topics Architecture – The Dynamic Scalable Architecture provides better performance and lower cost of ownership than Oracle’s many faceted solution. • OLTP • Data Warehousing • Complex Data / Content Management OLTP Success Factors • Performance – Transaction rates, scalability, multi-user access to data • Administration and Maintenance – Low labor costs, compatibility with enterprise • Replication – Availability of data, ease of recovery • Extensibility – Planning for the future of OLTP • Cost of Ownership – How all factors impact the bottom line OLTP Application Comparison Informix is significantly faster than Oracle in realworld OLTP applications. • BAAN Triton 3.1 Client/Server Benchmark – Informix 1,510 users: ~2X better than Oracle on identical hardware – Informix developers and support engineers on-site at BAAN • SAP R/3 Benchmark – New world record on Windows NT: 1,116 users – Only Informix has joint development lab at SAP headquarters in Germany. Industry View of Oracle OLTP Support “[Oracle] consumes more OS resources, has a relatively weak query optimizer and was later to the game than Informix in parallelization.” Hambrecht & Quist, 2/97 • Oracle has ‘bolted on’ features rather than rewriting and integrating • Many of Oracle’s ‘benchmark special’ OLTP features are largely unusable Components of OLTP Performance • Cost-based optimizer – efficient queries, best utilization of hardware • Parallel Insert, Update, Delete (DML) – high transaction rates, good multi-user access • Integrated log-based replication – reliability, recovery, ease of administration Informix vs. Oracle: Optimizer • • Informix – Cost based, automatically chooses optimum query strategy Oracle inferior in these areas: – Limited support for parallel processing – Incompatible with standard SQL • Requires user to define non-standard ‘hints’ based on table characteristics – Inefficient • Optimizer can not be used by standard tools and applications due to need for ‘hints’ • Most users turn it off Informix vs. Oracle: Parallel DML • • Informix -- best in industry – Optimal use of hardware for scalability and performance Oracle inferior in these areas: – Data Integrity • No serializability => no recoverability • No multi-statement transactions • No referential integrity – Advanced Database functionality • No support for triggers, replication, variable arrays, nested tables, large objects – Multi-user concurrency • Row level locking must be turned off • Parallel insert locks entire table Informix vs. Oracle: Replication • Informix – Log based. Fast and reliable – Open. Replicates to other vendors’ databases – Simple to implement and maintain • Oracle – Trigger based. Management nightmare. – Closed. • Can only replicate Oracle data. – High Maintenance overhead. • requires a separate server • incompatible with objects, partitions, parallel DML • triggers Replication in Practice Five steps involved in Informix log-based replication: Step 1: Transaction updates tables Step 2: DSA will log these updates using regular (parallel) logging mechanism Step 3: Replication threads read from in-memory log buffers Step 4: Apply threads will fan-out the messages to remote servers in parallel using optimized protocol Step 5: Remote servers (targets) will apply transactions in parallel Informix Advantages: 3x the I/O, half the # of steps, 5-10x faster! This means: Higher Availability of Data Less chance for data loss Easier recovery Ten steps involved in Oracle trigger-based replication: Step 1: Transaction updates tables Step 2: Oracle will log these updates using regular logging mechanism Step 3: Replication triggers are fired (one per table involved) = additional CPU processing Step 4: Replication triggers call procedures to update change queue (additional inserts occur to change queue tables) = additional I/O processing Step 5: Oracle will log these inserts using regular logging mechanism = additional I/O processing Step 6: Changes get sent via SQL*Net protocol to target DBMS = significant overhead Step 7: Target DBMS updates tables Step 8: Oracle will log these changes using regular logging mechanism = additional I/O processing Step 9: Deletes from change queue tables additional I/I Step 10: Oracle will log these deletes using regular logging mechanism = additional I/O processing Informix vs. Oracle: Administration • • Informix -- complete and flexible – Informix Enterprise Command Center • Automated, Scalable, Open (Java, IIOP, SNMP) – Enterprise Framework Integration • HP OpenView, CA UniCenter, DEC PolyCenter, IBM NetView/6000, Sun Solstice – Integration with Tools Partners • BMC, Platinum, Tivoli, Compuware Oracle -- provides an “Oracle only” solution – Enterpise Manager • Closed, Proprietary Architecture The Next OLTP Advantage: Extensibility • • Infrastructure Integration – Move complex business from applications to central control of the database server – Add management of today’s “unmanaged data” • documents, e-mail, images Meet highly specialized needs – user defined functions on server – pre-witten data types (DataBlades) on server • time series, text management, spatial queries – high performance indexes for complex relationships, hiearchies, business rules Informix vs. Oracle: Extensibility • Informix – A single, Universal Server -- available today. – Full database support for user defined datatypes, abstract datatypes, user defined functions – Faster response to changing business requirements; reusable code; flexible systems • Oracle – Not currently available in server – Conflicting extensibility messages – Promised in Oracle8 since 1993 – Now promised in Oracle middleware (NCA) in 1998 Cost of Ownership: Informix Advantages Informix Result Better performance & scalability Lower hardware costs Fewer upgrades Fully implemented features Less user code Faster deployment Easier maintenance Extensibility Keep pace with changing requirements Extends lifespan of applications Topics Architecture OLTP – Informix has decisive advantages for OLTP in the critical areas of Performance, Administration, Replication, and Extensibility. • Data Warehousing • Complex Data / Content Management Data Warehousing Success Factors • • • Hardware Investment – Hardware cost for 500 GB Data Warehouse: • $3.5 million on SMP systems • $7 million on MPP systems Performance / Processing Time – Judged in hours and days, not seconds Ease of Use and Total Cost of Ownership – Obtain the right answer, easily customize – Avoid specialized training and tasks – Minimize hardware investment Industry View of Oracle DW Performance “[Oracle] is notoriously poor at scaling above 2 to 4 nodes.....” Meta Group • Oracle parallel processing was introduced as a late response to Informix • Its capabilities are limited and incompatible with other performance oriented features Result: Informix Wins in Head-toHead Contests “Informix has beaten both Oracle and IBM for some very high visibility terabyte accounts - namely MCI, Visa International, Sears, and Fidelity Investments” Patricia Seybold Group 2/97 MCI Sears HCIA First Union Fleet Bank Fidelity Investment Telecom Retail Healthcare Financial Financial Financial 6 TB 5 TB 3 TB 2-3 TB 1-2 TB 800 GB TPC-D Performance (Scale Factor 300 GB) 4000 TPC-D Power (QppD) 3.416 TPC-D Throughput (QthD) 3000 2.009 2000 1.673 1.241 916 1000 0 • • • • 1.360 Informix (HP) Oracle (Sun) Oracle (IBM) 1.501 1.028 Teradata (NCR) Informix vs. Oracle TPC-D: 70% faster on 25% less hardware – Informix: 48 CPU’s, Oracle: 64 CPU’s Informix means lowest overall hardware investment Best price / performance Less HW = lower admin overhead , lower labor costs TPC-D Total Running Time (Scale Factor 300 GB) 6:00 hh:mm 0:00 Load Time 18:0 12:0 6:00 0:00 • Query Time 3:07 14:53 4:07 5:34 15:09 13:39 14:36 Oracle (Sun) Oracle (IBM) Teradata (NCR) 7:47 Informix (HP) Data Warehousing encompasses all functions, not just queries • Informix completed entire benchmark before Oracle completed loading • Informix has complete parallel architecture: load, index, query, backup/restore, management, etc. Performance: Informix Extended Parallel Server (XPS) Node 1 Node 2 Node N High Speed Interconnect 1 Informix most efficient! Update requires : 1 message; no I/O • • • Highly efficient shared nothing architecture – Each node controls its own log and data Flexible data partitioning Function shipping eliminates need for distributed lock manager Performance: Oracle Parallel Server (OPS) Node 1 Node 2 1 7 same task, Oracle requires: 12 Messages 2 I/Os 8 11 DLM Node N 2 6 3 12 5 4 Oracle Virtual Shared Disk • Oracle simulates shared disk – inflexible partitioning scheme • Requires huge overhead: distributed lock manager, many extra messages and I/Os 10 9 Performance: Review of Key Informix Features • Shared nothing architecture – Easily locates all data; avoids locking conflicts • Data partitioning – Spreads data across multiple disks and nodes for highest performance – Six different table/index partitioning options • Advanced parallel processing – Best use of hardware Application transparent parallel everything: query, sort, scan, join, aggregates, update, delete, insert, ... – Pipelined parallelism; superior hash-based join Ease of use: Informix Metacube vs. Oracle Express • • Informix MetaCube: ROLAP Decision Support – Integrated tool for browsing and query of the database – Analyzes standard Informix RDBMS data • no separate server to administer • no translation code, no extra overhead Oracle’s DW Solution requires Oracle Express – Oracle Express is a separate database server • does not use RDMBS, but a multi-dimensional DB – Must load data into Express from main Oracle DB • user must write “pass through” code • updating of data requires a re-load Performance: Oracle Data Partitioning • Incomplete – Supports only simple range partitioning – No round-robin, hashing, random, direct, expression, schema , ... – No migration tools from Oracle 7.3 • Incompatible – Partitioned tables may not contain objects, triggers, constraints, data clustering, index clustering • Inefficient – Range partitioning data skew causes poor performance – Performance degraded due to inability to use existing features Informix Customer Story: MCI MCI Data Warehouse facts at a glance: • 6 TB of Data • Friends and Family lead generation program • 3500 data points on each of 190 million people • Oracle beaten in head to head competition – Oracle could not successfully complete the benchmark • Informix demonstrated DW expertise – Awarded 1996 Best Practices award by the Data Warehousing Institute. – Project established industry’s TPC-D benchmark. MCI Data Warehouse Benefits • Informix DW enables better use of essential business data. – Long Distance Providers face – 50% annual customer turnover – Better analysis of leads is a strategic advantage – Reduction in lead generation cost: • 1994: $0.90 / lead • 1996: $0.07 / lead $/Lead $0,70 $0,60 $0,50 $0,40 $0,30 $0,20 $0,10 $0,00 1994 Informix outperformed next competitor by 2X – initial hardware savings of $2-3 million – more savings over time: HW will = 4X software cost 1996 The Next Data Warehousing Advantage: Extensibility • New types of queries – add spatial analysis – add unique business rules, unique data types • Better information – Compute the lowest fare route between New York and Baton Rouge and plot alternative route on a map • Faster response for complex queries – Find all high tech stocks that have a risk/reward ratio lower than current portfolio Informix vs. Oracle: DW Extensibility • Informix – Numerous complex and new media datatypes – Database integration => better performance, easier administration and lower maintenance – Data Warehouse becomes central information repository • Oracle – Partial database integration of text, image and spatial data – Others promised as separate servers or middleware in Oracle 8 – Much higher overhead, harder to manage Data Warehouse Cost of Ownership: Informix Advantage Better performance & scalability Dramatically faster load times E xtensibility built into server Result F aster answers Less hardware required E nables applications not possible with O racle M ore frequent updates M ore reliable results F aster RO I on W arehouse investment F aster deployment E asier maintenance E xtends lifespan of applications Topics Architecture OLTP Data Warehouse – Informix has consistently demonstrated superiority to Oracle in Data Warehousing by delivering on customers’ systems and drastically cutting their costs. • Complex Data / Content Management Content Management Critical Success Factors • • • Integration – Connectivity to: web servers, browsers, java applets, legacy data Management, Performance, Cost of Ownership – single server for all data – unified and industry endorsed architecture – scalable architecture Advanced data types – audio, video, text, spatial, java, VRML, etc… – user written and pre-defined (DataBlades) Informix: Integrated, Simple, Complete. Simple interface means fast development Optimized access means fast applications Single data source means easy administration No need for file, scripts, or programs outside the DB. Single Interface Optimized Access Integrated Data Types (DataBlades): Documents HTML Images Video Time Series Corporate Data ... Single Universal Server Informix Connects the Entire Enterprise to the Web Java/IIOP Browser Web Server ActiveX INFORMIXUniversal Web Connect™ • Java • C/C++ •New Era 4GL •IIOP/CORBA •DCOM •Publish and •Subscribe Development Tools Jworks, Web DataBlade, NewEra, Data Director, Partners... INFORMIX-Universal Server • Web Site Content Legacy Data • Application Logic • Java Gateways Management Tools Informix Enterprise Command Center, Tools Partners... Industry View of Informix Content Management “For serious new media projects, Informix offers the most content-inclusive strategy.” Forrester Research, 9/97 • Analysts agree that Informix’s server integration is the most efficient architecture • Informix has numerous off-the-shelf DataBlade modules currently available Oracle’s NCA -- Not a Unified Architecture 3rdThird partyParty ORBORB • Non-integrated servers • No extensibility • No new indexing • No cartridges CORBA Objects Communications Overhead Massive Data Movement Video Server OLAP Engine Client Application Server • CORBA Client • No standard DML • Not parallel User code • No optimizer • No data integration Standard Relational Performance Problems of Oracle’s NCA Call Overhead Local Lightweight Thread Informix Oracle CORBA Object 1x 10x 100x 1,000x 10,000x relative time Oracle’s NCA is 1000X slower than Informix’s Universal Server in performing a function on an advanced data type. Performance Problems of Oracle’s NCA Data Movement Overhead Local Lightweight Thread CORBA Object Informix Oracle local object Oracle remote object 1x 10x 100x 1,000x 10,000x relative time Oracle’s NCA is 100 to1000X slower than Informix’s Universal Server at transferring complex data types (objects) from the server. Cost of Ownership: Oracle ‘Do-itYourself’ Approach ‘Do-it-Yourself’ Custom Code Oracle Code Relational Data BLOBs Data Integrator Optimizer Object Simulator Complex Geometry Engine Text Spatial Points Cost of Ownership: Oracle NCA Integrity Issues Transaction Boundary Data Catridges Separate Oracle Database Servers Oracle Application Server • • • • Cartridges outside transaction boundary No integrity constraints No standard data manipulation language No certification program Oracle’s Competing Architectures • • • • Oracle 7.3 with hard-wired support for 4 data types – spatial, time series, text media – separate servers employed for each Oracle 8.0 – mainly provides patches to scalability and performance issues – extremely limited object support Network Computing Architecture (NCA) – “data catridges” as middle-ware, CORBA interface – advanced data outside the control and optimization of server Data Cartridges as part of the server – perhaps in Oracle 8.1 or 8.2 Performance Appraisal: Industry View of Oracle’s NCA “Oracle faces a significant challenge in achieving adequate performance when interfacing Data Cartridge functions with core database services.” Patricia Seybold Group, 2/97 “Oracle 8 will (at least initially) suffer performance shortcomings compared with the IUS single database engine..” Meta Group, 2/97 Informix vs. Oracle: Predefined Data Types • Oracle: Few if any Data Catridges are shipping. • Informix: A sample of available Data Blades... • • Data Warehousing/Financial • – DataCleanser DataBlade (EDD) – OptiLink DataBlade (CPT) – Fuzzy Search DataBlade (Expector) – TimeSeries DataBlade (Fame) – Reporting DataBlade (SLP-Statistiques) – Decision Series DataBlade (NeoVista) Text – Text DataBlade (Verity) – Text Search/Management DataBlade (Excalibur) – NameTag DataBlade (IsoQuest) – PLS Text DataBlade (PLS) – Document Objects DataBlade (ArborText) • Web/Electronic Commerce Digital Media – Video DataBlade (VXtreme) – LiveLink DataBlade (Open Text) – PixFactory DataBlade (Kodak) – DesCrypt DataBlade (Prime Factors) – Internet Commerce DataBlade (Open Market) – Image Search DataBlade (Excalibur) – Face Recognition DataBlade (Excalibur) – Real-time Profiling DataBlade (Excalibur) – Scene Detection DataBlade (Excalibur) – Content Management DataBlade (MKS) Geospatial – Visual Info Retrieval DataBlade (Virage) – GeoCoding DataBlade (MapInfo) – Audio Information Retrieval DataBlade – Global/Interval DataBlade (Telcontar) (Muscle Fish) – Visualization DataBlade (ECOlogic) Oracle’s Negative Campaigning Against Informix “In one of the most vicious smear campaigns in recent software history, Oracle has been attacking the integrity of [Informix’s] database” PC Week 2/17/97 “We believe the actual exposures with Informix’s DataBlades have been miniscule.” Gartner Group, 12/96 Web / Conent Wins Over Oracle • • • • SouthWestern Bell Nike Fidelity IVillage Topics Architecture OLTP Data Warehouse Content Management – Informix’s Universal Server was designed from the start for advanced content management. Oracle 8 and the NCA are cobbled together components which will not fare well in real world applications. • Summary Summary: Informix • Informix – – – – Mission: Deliver absolute best database technology Single and extensible server arhitecture Tight integration with best-of-breed application partners Results for the Customer: • lower cost-to-serve • flexibility and simplicity • extended lifespan of applications • clear support for the future enterprise with Universal Server Summary: Oracle • Oracle – Mission: Cover as many bases as possible in the market with combinations of existing products. – Lock customers into Oracle products; partner only when necessary – Results for the Customer • higher overall costs due to non-integration • performance limitations • expensive services • questionable path to future with Oracle 8 and NCA