Three Problems, Three Myths, Three Realities

Three Problems, Three Myths, Three Realities, Tom Fisher, EVP, Chief Technology Officer, MapR

Transcript

1. Tom Fisher EVP, Chief Technology Officer Three Problems, Three Myths, Three Realities
2. 2 © 2018 MapR Technologies, Inc. // MapR Confidential Forrester Research predicts that by 2020, businesses adopting Machine Learning, AI, and Deep Learning, the Internet of Things (IoT), and Big Data will take away more than $1.2 trillion from their less-informed peers. We Are All Here Because This Has Big Potential Impact
3. 3 © 2018 MapR Technologies, Inc. // MapR Confidential MapR Technologies COMPANY SUCCESSFUL CUSTOMERS ACCLAIMED INNOVATOR Silicon valley software company Proven Preferred choice for production AI & analytics deployments Better value, lower TCO AI, Machine Learning & Analytics at scale IoT & Edge Analytics Stateful container apps Hybrid and multi cloud Reliable distributed file & object store
4. 4 © 2018 MapR Technologies, Inc. // MapR Confidential ON-PREMISES, MULTI-CLOUD, IoT EDGE COMMODITY SERVER VIRTUAL MACHINE IoT & Edge APIs: NFS, POSIX, REST, S3, HDFS, HBASE, JSON, KAFKA Powerful Founding Vision, Now the Leading Platform for AI & Analytics Distributed High-scale, High-resiliency Core Distributed File and Object Store NoSQL Database Event Store For Apache Kafka
5. Myths Realities 3
6. 6 © 2018 MapR Technologies, Inc. // MapR Confidential AI is all about the right algorithms. AI is different from “big data” Myth #1:
7. 7 © 2018 MapR Technologies, Inc. // MapR Confidential It Is A Journey Along An Evolving Continuum Data Warehouses Data Lakes Machine Learning Analysts Data Scientists Centralized Cloud Multi-cloud WORKLOADS DEPLOYMENT MODELS AI Targeted Offers Fraud Detection Smart Cars Containerized Targeted Offers Security Predictive Maintenance Security
8. 8 © 2018 MapR Technologies, Inc. // MapR Confidential 9.9 M Weekly Transactions DATA PLATFORM ETL PROCESSING Spark, Pig, Hive Machine Learning Scala, Python,H20.ai, Zeppelin, XGBoost OFFLOAD / RELOAD Customer Datasets Data Warehouses Customer Loyalty data from Retailer 14 M Weekly Transactions Property Ownership Data 275M Property Facts Facebook Data Social Feeds Media Consumption Data Remote Clicks from 110k households Vehicle/Property Insurance Data Banking Transactions 5M Insurance Members Data Data Science Platform Data Ingestion Services Data Transformation Services Execution Training Models Decision Making Cloud based AI Platform for segment marketing Customer Analytics & Insights Pricing and Credit Risk Decisions Market Insights One Customer’s Architecture: Multiple Tools & Engines Needed
9. 9 © 2018 MapR Technologies, Inc. // MapR Confidential Data Scientists Want Access To Fast Moving Innovation & Agility /f1
10. 10 © 2018 MapR Technologies, Inc. // MapR Confidential • Algorithms are important, but model management is essential for any reasonable production scale • 90+% of Machine Learning Success is Data Logistics • Data scientists are DataOps people, not interested in “IT” • ”IT” people see Data scientists as demanding & difficult to please From DevOps To DataOps Ted Dunning & Ellen Friedman Model Management in the Real World Machine Learning Logistics https://mapr.com/ebook/machine-learning-logistics
11. 11 © 2018 MapR Technologies, Inc. // MapR Confidential It is a continuum from batch, to microbatch to ML, DL, AI It cannot be a separate silo and doesn’t have to be Enabling a DataOps lifestyle is critical AI is about algorithms AI is separate tech Myth Reality AI & ML:Myth & Reality
12. 12 © 2018 MapR Technologies, Inc. // MapR Confidential Containers are only good for simple stateless apps. Doesn’t help complex stateful apps. Myth #2:
13. 13 © 2018 MapR Technologies, Inc. // MapR Confidential Container Challenges with Microservices Microservices Require Data Systems that Scale to Millions of Tables, Streams, Billions of Files Stream Microservice A Microservice B Microservice C Stream Stateless containerized application • Applications deployed as single monolith • Single instance of storage, database • Inter-process communication internal to server • Applications are transaction-based • Applications deployed as many microservices • Per-microservice storage, database • Inter-process communication via Event Streams, REST APIs • Applications are event-based Old New
14. 14 © 2018 MapR Technologies, Inc. // MapR Confidential Containerized Stateful Apps On A Data Platform MAPR DATA PLATFORM Data Science Refinery Container TENANT 1 TENANT N Application Application ….. Stateful app container MAPR KUBERNETES VOLUME DRIVER CUSTOMER APPLICATION S MAPR POSIX CLIENT FOR CONTAINERS MAPR CLIENT FOR CONTAINER S
15. 15 © 2018 MapR Technologies, Inc. // MapR Confidential Stateful apps in containers with persisted data is a radical enhancement for ML/AI and microservices Containers are only for stateless apps Myth Reality Containers: Myth & Reality
16. 16 © 2018 MapR Technologies, Inc. // MapR Confidential Going all-cloud with the cloud’s native services is the simplest path to the future Myth #3
17. 17 © 2018 MapR Technologies, Inc. // MapR Confidential • Cloud providers offer platform services leave customers siloed. • APIs are specific to one type of cloud but are not compatible with other clouds • Costs in cloud add up over time, often with opaque billings • Cloud is easy to get in, but hard to get out • No SLAs on performance on Public Clouds and minimal penalties if missed • The moment you move to the cloud, you also hand over the control over your data. • No solution for multi-cloud strategies spanning different clouds or hybrid strategies in combination with on-prem When my cloud instance was down, I felt helpless. The other cloud is cheaper now. I will not give away control over my sensitive data Marketing started to use yet another cloud I’ve made infrastructure investments Auditing is a huge topic for us Data is hard to move It feels like managing contracts not services No consistent performance. The Cloud Reality
18. 18 © 2018 MapR Technologies, Inc. // MapR Confidential Edge Private Cloud On Premise Public Cloud Public Cloud Public Cloud API Application API Connector API API API API ü Cloud Alone: Difficult to establish a unified data access or security concept
19. 19 © 2018 MapR Technologies, Inc. // MapR Confidential Edge Private Cloud On Premise Public Cloud Public Cloud Public Cloud Open APIs Application • Unified Security Model • Data access decoupled from physical storage location. Globally. • Data made portable • No lock-in to proprietary APIs • Full openness API Connector GLOBAL DATA MANAGEMENT ü Cloud With A Data Platform: Portable, Unified Access & Security Silo Problem solved!
20. 20 © 2018 MapR Technologies, Inc. // MapR Confidential A Data Platform Inside The Cloud Pays Off • Lower total annual cloud billings by 16% to 41% • Lower new use case development costs by 33% • Lower cumulative 3 year cloud costs for an average sized analytics use case by about 24% or an estimated $584K • Lower the cumulative 3 years cloud costs for an average sized converged or complex use case by about 21% or an estimated $2.6M * Value propositions based on 2 AWS DC locations, 10% hot MapR data tier, 10% annual growth in billing metrics (TB volume), and other standard MapR metrics estimated by defined use case type T o p ic T o p ic T o p ic On-Premises S 3 MAP R EDGE MAPR DATA PLATFORM MAPR DATA PLATFORM MAPR DATA PLATFORM MAPR DATA PLATFORM
21. 21 © 2018 MapR Technologies, Inc. // MapR Confidential A data platform inside a cloud makes it easy to move data and applications between clouds or on-prem Can lower cost, speed development time, avoid lock-in The cloud provider’s features and capabilities are the best to use Myth Reality Cloud: Myth & Reality
22. 22 © 2018 MapR Technologies, Inc. // MapR Confidential Three Myths Three Realities AI is about the right algorithms & needs it’s own silo. Containers are only good for stateless apps. Going all-cloud with one provider is best. A data platform approach simplifies the journey to AI, and speeds time to value A data platform is a radical enabler for stateful microservices A data platform inside cloud makes data & app movement easy, saves money, avoids lock in
23. 23 © 2018 MapR Technologies, Inc. // MapR Confidential Thank You! Busting Big Data Myths Articles and Videos
24. © 2017 MapR Technologies MapR Confidential 24 Q & A