IoT and Big Data - Iot Asia 2014

Presented at IoT Asia 2014 Workshop

Published in: Data & Analytics, Technology

IoT and Big Data - Iot Asia 2014 from John Berns

Transcript

2. © 2014 MapR Technologies 2 What This Is; What This Is Not • It’s not specific to IoT – It’s not about any specific type of data or protocol – It’s not specific to any particular industry • It’s about processing big data – IoT data can be big data – IoT might be the biggest data of the coming decade – But it’s just big data – Same strategies & technologies apply
5. © 2014 MapR Technologies 5 When Does Data Become ―Big?‖ • When the size of the data, itself, becomes a problem • When the ―old way‖ of processing data just doesn’t work effectively • It’s ―big‖ when we have to rethink: – How we store that much data – How we move that much data – How we extract, load & transform that much data – How we explore and analyze that much data – How we process and get meaningful insights from that much data
6. © 2014 MapR Technologies 6 C’mon! What does that mean in size? • Not gigabytes • Most likely not a few terabytes • Possibly not 10’s of terabytes • Probably 100’s of terabytes • Definitely petabytes
9. © 2014 MapR Technologies 9 Big Data Analytics Bridging classic & big data worlds “Capture only what’s needed” SQL performance and structure Hadoop scale and flexibility IT delivers a platform for storing, refining, and analyzing all data sources Business explores data for questions worth answering Big Data Method Multi-structured & iterative analysis IT structures the data to answer those questions Business determines what questions to ask Classic Method Structured & Repeatable Analysis “Capture in case it’s needed”
10. © 2014 MapR Technologies 10 Philosophical Differences Traditional Methods • More power • Summarize data • Transform and store • Pre-defined schema • Move data -> compute • Less data / more complex algorithms Big Data • More machines • Keep all data • Transform on demand • Flexible / no schema • Move compute -> data • Mode data / simple algorithms
11. © 2014 MapR Technologies 11 answer = f(all data) • Save all raw data • Data immutability • Transform as needed • Result is based on the raw data
15. © 2014 MapR Technologies 15 Forces of Adoption Hadoop TAM comes from disrupting enterprise data warehouse and storage spending Data IT Budgets • Gartner, "Forecast Analysis: Enterprise IT Spending by Vertical Industry Market, Worldwide, 2010-2016, 3Q12 Update.― • Wall Street Journal, ―Financial Services Companies Firms See Results from Big Data Push‖, Jan. 27, 2014 $9,000 $40,000 <$1,000 2013 ENTERPRISE STORAGE IT BUDGETS GROWING AT 2.5% 2014 2015 2016 2017 DATABASE WAREHOUSE DATA GROWING AT 40% $ PER TERABYTE HADOOP
18. © 2014 MapR Technologies 18 Typical Compute Node • Two CPUs, each with 4-8 cores per CPU • 32-128 GB Memory • 6-24 hard disks • 2-4 10GB Network cards
22. © 2014 MapR Technologies 22 SQL on Hadoop • Generally data has no inherent ―schema‖ • Schema is defined by user / interpreted from structure • Schema is applied during processing • One file can have many schemas applied • Works for many kinds of data—but not all – Temperature sensor data? Sure – Video feeds? Not really
23. © 2014 MapR Technologies 23 Key Use Cases • Exploratory analysis on large scale raw data • Unknown value • No defined schema • Variety of data types • Large-scale SQL queries on long history • Well defined schema • Known value, but high cost in existing systems 2 Big Data Analysis Big Data Exploration
24. © 2014 MapR Technologies 24 What is Driving the Need for SQL-on-Hadoop? Organizations are looking for • Reuse existing tools and skills to unlock Hadoop data to broader audience • Analysis on new types of data • More complete data analysis • More up-to-date and real-time data analysis (not just ―after the fact‖)
25. © 2014 MapR Technologies 25 Drill 1.0 Hive 0.13 with Tez Impala 1.x Presto 0.56 Shark 0.8 Vertica Latency Low Medium Low Low Medium Low Files Yes (all Hive file formats) Yes (all Hive file formats) Yes (Parquet, Sequence, …) Yes (RC, Sequence, Text) Yes (all Hive file formats) Yes (all Hive file formats) HBase/M7 Yes Yes Various issues No Yes No Schema Hive or schema- less Hive Hive Hive Hive Proprietary or Hive SQL support ANSI SQL HiveQL HiveQL (subset) ANSI SQL HiveQL ANSI SQL + advanced analytics Client support ODBC/JDBC ODBC/JDBC ODBC/JDBC ODBC/JDBC ODBC/JDBC ODBC/JDBC, ADO.NET, … Large joins Yes Yes No No No Yes Nested data Yes Limited No Limited Limited Limited Hive UDFs Yes Yes Limited No Yes No Transactions No No No No No Yes Optimizer Limited Limited Limited Limited Limited Yes Concurrency Limited Limited Limited Limited Limited Yes SQL on Hadoop: Many Options Flexibility to choose when to use which based on use case
26. © 2014 MapR Technologies 26 ENTERPRISE DATA HUB MARKETING ANALYTICS RISK ANALYTICS OPERATIONS INTELLIGENCE • Multi-structured data staging & archive • ETL / DW optimization • Mainframe optimization • Data exploration • Recommendation engines & targeting • Ad optimization • Pricing analysis • Lead scoring • Network security monitoring • Security information & event management • Fraudulent behavioral analysis • Supply chain & logistics • System log analysis • Manufacturing quality assurance • Preventative maintenance • Sensor analysis Proven Hadoop Production Success
29. © 2014 MapR Technologies 29 • Map Reduce Framwork • Lingual: SQL-like operations • Pattern: Machine Learning Applications • Scalding: Cascading for Scala • Cascalog: Cascading for Clojure
30. © 2014 MapR Technologies 30 • Python, Scala and Java • Spark powers a stack of high-level tools including – Shark for SQL, – MLlib for machine learning, – GraphX, and – Spark Streaming. • You can combine these frameworks seamlessly in the same application.
31. © 2014 MapR Technologies 31 • Machine Learning / Predictive Analytics – Collaborative Filtering – Linear / Logistic Regression – Naïve Bayes – Random Forests – K-Mean Clustering – Canopy Clustering – Principal Component Analysis
32. © 2014 MapR Technologies 32 • Database on Hadoop • Highly scalable • Columnar – Flexible schema • Data source for Map Reduce and Spark jobs
36. © 2014 MapR Technologies 36 NoSQL Databases • No-SQL or ―Not only‖ SQL • Give up some of the functionality of traditional relational databases for speed and scalability • Types – Key-Value – Columnar – Document – Graph • NoSQL databases favor flexible schemas
39. © 2014 MapR Technologies 39 Queues • Just like a queue at an amusement park • First-in-first out • Queues messages or events
42. © 2014 MapR Technologies 42 Stream Processing • Handles data at high velocity • If Hadoop is the ocean, streams are the firehose • Processing in near real-time

Top of the world

Monday, April 20, 2015

IoT and Big Data - Iot Asia 2014

IoT and Big Data - Iot Asia 2014

Transcript