1、Versant数据库与非结构化海量数据存储及应用案例,Versant Database in Non-Structural, SQL/NO-SQL, Big Data Applications,Robert Greene Vice President Technology Versant Co. Ltd.,Agenda,Who is Versant What is the Technology Why it is Chosen Customer Examples Technology,Hamburg,Shanghai,Redwood City, CA,Bangalore,Innovator
2、of high performance databases since 1988 Stock Exchange (NASDAQ:VSNT) since 1996 Global Presence 4 major locations Redwood City, California Hamburg, Germany Shanghai, China Bangalore, IndiaWorldwide installations including: Telecommunications, Defense, Financial Services, Transportation, Manufacturi
3、ng, other.,Who is Versant,Versant Big Data and the Cloud,Data Volume Volume Complexity Speed Changing analysis model,Agenda,Who is Versant What is the Technology Why it is Chosen Customer Examples Technology,What is the Technology,A complete object oriented approach to scalable, extreme performance
4、Cloud Database data management,Distributed, Clustered, Parallel, Fault Tolerant,Compute Node 2,Compute Node 3,Compute Node 4,Compute Node 1,Compute Node ,Parallel Thread Execution On Compute Nodes,Versant,Versant,Transparent Data Distribution / Data Balancing,The Better NoSQL,. find by key( ) + a lo
5、t of custom application code,.find by key( ) or select from where operators + JPA,VSNT =,partition key:graph,connected,disconnected,Object Database Industry Leadership,Versant Performance vs. RDBMS,performance improvement,50x30x10x5x3x,Complexity level of Data Model for transactions,Simple Queries C
6、omplex transactions,Simple,Complex,ING Barings Risk Management,CIBC Global Markets Front Office Trading,IBM Benchmark,Medium,Ericsson Network Management,Verizon Fraud Management,British Airways Yield Management,52x faster,32x faster,35x faster,85x faster,10-50x faster,15x faster,Agenda,Who is Versan
7、t What is the Technology Why it is Chosen Customer Examples Technology,Why is it chosen,Handles model complexity of modern business Benchmarked significantly Faster than competition Designed to Scale with a modern architecture Provides high QoS up time ( 99.9999 Availability ) Provides extreme opera
8、tional Efficiency,Versant Customers,Agenda,Why it is Chosen Example Customer,The Challenge at ESA: Planck,Complex Big Data Captured Analyzed Shared,X=4329832809458930458099853,Y=348328409328,ANALYSIS SEQUENCE /,Technology Selection Complexity,Key Versant Benefits Persistence of compute schema as sto
9、rage schema (no mapping) Supporting data connectivity across physical database nodes,Object Oriented Analysis & Design Simplifies the computation space Modeling of Groups No need for complex, computational expensive queries Framework for dealing with scale and distribution,Technology Selection Datab
10、ase Size,Computational Schema Size Big Data as previously discussed ( 200M+ per Map ) Simulated Data Bolometer Experiments,Key Versant Benefits 75% less disk space required,Technology Selection Analysis / Performance,Versant 10 times faster on analysis,Technology Selection Creation,ESA High Availabi
11、lity - Online Administration,On-line distributed backup Full and incremental Roll forward archiving High Availability Backup On-line data volume addition On-line Index creation/deletion On-line schema evolution GUI Admin tools and browsers,How,Why it is Chosen Complex information management,Same Log
12、ical and Physical Modeling,How does it handle complexity,Business Developer DBA speak the same language,Cut Development cost - Code Less - Conceive more - Run Faster,Analyse Design,C, C+, Java classes,Application test,Versant CDFS II (Cloud Depiction and Forecasting),Precise Worldwide Cloud Forecast
13、s near real-time global cloud analysis and forecasting. Operated by the Air Force Weather Agency (AFWA) 6 times higher frequency 4 times higher accuracy,Boeing: Multi-Data Source Integration in Anti-Missile System,Fast Multi Intelligent Data Source Processing Large Volume Data Processing.,China Reso
14、urce: Address Database,Tree Organized Address Database; Distributed Address Database, Ready to Get Deployed on Cloud Computing Environment Multiple Layers/Tiers, Multiple Data Connections, Multiple Calculation Models; Ready to Get Deployed with GIS.,How,Why it is Chosen Extreme performance,Model dri
15、ven, not data driven NoSQL based design Data relationships established mostly once Relationships semi-static, changing infrequently Data values are what change often Eliminated need for slow JOIN operations RDB JOIN is runtime relationship calculation Terrible for richly linked, unstructured and gro
16、wing data Versant stores relationship natively, so no JOIN Query on relations, but no runtime calculation of relations,How does it achieve better performance,SabreSonic Inventory,Inventory Manager 90K in production complex read transactions per second 500 write transactions per second Claims - highe
17、st business transaction processing system in world Chose Versant for: Huge performance advantage Availability processor caches entire database in server farms, transacts to disk Uses CORBA messaging for refresh of changed objects Cloud testing at SUN labs achieved 150K txn/sec on 14 node cluster Rel
18、iability 5 nines reliability using our Fault Tolerant Server Online backup and no need for offline rebuilds,Quantel Pablo Platform,How,Why it is Chosen Big Data Scalability and Cloud Distribution,How does it handle cloud scalability -virtual data store-,Objects may be partitioned into any database i
19、n network and can reference any other object in any other database and you can move them freely from database to database Allows for efficient partitioning and parallelization of access Data store is now scalable and it is easy to minimize network traffic Allows for “locality“ performance gains,Veri
20、zon - Sheriff Architecture,file:/E:/p0007.gif,Verizon - System Highlights,Transactions in a 12 hour period 1.8B events inserted into the database size of transaction 550 bytes 5 million cases built size of transaction 3K bytes Hardware configuration Dell x86 Linux Servers EMC Disk Array Veritas file
21、 system with stripped raw devices,How,Why it is Chosen High Reliability, Always On System,How does it achieve its availability,Synchronous fault tolerant replication Asynchronous master-slave, peer-peer replication Hardware integrations for active-passive availability Visual and API onlineadministra
22、tion,Real-time GIS Support in Battle Field Management,Thales - Field Battle Management System; Fast & Applied to Multiple Simulation Models.,How,Why it is Chosen Operational Efficiency lower cost of cloud,Extreme Operational Efficiencies -storage-,Save greater than 50% in disk storage No duplicate d
23、ata for JOINs No excessive indexes to rebuild relations No duplicate data for performance optimization (de-normalization ),European Space Agency Planck Study,Extreme Operational Efficiencies -CPU-,Save up to 50% in server capacity No CPU intensive mapping processing in mid tier No CPU intensive JOIN
24、 operations in database server,* enJin is an application server branded packaging of Versant,IBM study,Versant ExxonMobil EMPower,ExxonMobil uses the Versant database for its reservoir simulation system. This system is used to evaluate geological information and answer key questions such as: Where t
25、o place wells When to bring on line When & where to inject fluids The performance and object orientation of the Versant database is a perfect fit for real time simulation and modeling systems.,Game Logic Server,Chat,Player Actions,A.I.,Action & Response Multiplexor,Client-specific game “world views”
26、,Game World,Client,Client,Client,Client,UniversalC - Gundam Online,Publisher: Bandai, Developed by DIMPS 1 billion objects in the game world at the start Integration of multiple game shards on one single cluster provides capacity for more than 100,000 concurrent users in a single game space (Supra-M
27、ultiple-Player-Connection possible),Game Logic Server,Gundam Implementation,Production event numbers 35,000 concurrent users 28 CPUs, 40G RAM,Agenda,Who is Versant What is the Technology Why it is Chosen Where it Matters,Where it matters for Cloud,It is crystal clear traditional relational databases
28、 are failing under the modern demands of complexity, concurrency and large data volumeIt is equally clear modern information model complexity needs more than a key:value store,ISV & othersMicrofocus/Borland Echelon Eidos Media IBM Invensys Oracle Quantel Siemens PLM OpenText Versata,TelcoEquipment A
29、lcatel-Lucent Ericsson NEC Nokia Siemens Nortel Operators AT&T France Telecom Samsung Verizon,TransportationAmadeus Air France BNSF British Airways Frontier Airlines GE Transportation Indra Norfolk Southern Qantas Sabre Airline Solutions,DefenseBAE Systems Boeing EADS ESA Lockheed Martin Northrup Gr
30、umman Qinetiq Raytheon Thales US Air Force,Versant Market Leadership,Life ScienceBeckman Coulter Becton Dickinson Biomerieux Clarity Cornell Weill Med MDS Sciex Nat. Lib. Of Medicine Siemens Medical,Robert Greene : MP: +1 9255774634 E-mail: Tiger Lau: CTO / China MP: 13818062567 MSN: longguo_ E-mail: Tiger.L,Versant Contacts,
