intel ai devcon intro to gigaspaces - bigdl technical...for ultra low-latency, high-throughput...
TRANSCRIPT
-
OperationalizingArtificialIntelligenceInstantInsightstoActions
2018
RajivShahDirectorofSolutionArchitectandProfessionalServices
-
GigaSpaces provides the leading in-memory computing platform for fast data analytics and extreme transaction processing. With GigaSpaces InsightEdge Platform, and its core engine XAP, organizations can overcome the complex obstacles of big data and advanced analytics to deliver real-time business impact. GigaSpaces offices are locatedin the US, Europe and Asia.
About GigaSpacesDirect customers300+
Fortune / Organizations50+ / 500+
Large installations in production (OEM)5,000+
ISVs25+
-
In-memory Insight Platform which unifies fast-data analytics, AI and transactional processing
Leverages XAP as its core technology
In-Memory Computing Platform for ultra low-latency, high-
throughput transaction processing
Leveraging the strength of combining innovation & mature technology
-
InsightEdge Platform is an open-sourcein-memory insight platform unifying fast-data analytics, artificial intelligence and transactional processing, for instant business insights and actions
EXTREMEPERFORMANCE
INSTANTINSIGHTS
TCOREDUCTION
-
TRANSACTIONAL/ANALYTICALPROCESSINGTRADITIONAL
TRANSACTIONAL PROCESSING
ANALYTICS
DATA REPLICATIONSLOWFEEDBACKLOOP
FASTFEEDBACKLOOP
TRANSACTIONAL PROCESSING
ANALYTICS
IN-MEMORYDATA GRID
TRANSACTIONAL/ANALYTICALPROCESSINGUNIFIED
IMPACTSReal-time analyticsGreater situationawarenessSimplified architecture
Traditional vs. Unified “Translytical” Processing
-
ACTIVE APP CENTER
CLOUD / CONTAINERS / ON-PREMISES
The Digital Core: Unified Fast Data & Microservices
HIGHAVAILABILITY
TIEREDSTORAGE
SECURITY &MULTI-TENANCY
MONITORING &OPERATIONS
INFRASTRUCTURE ORCHESTRATION AND SCHEDULING
TRANSACTIONAL APPLICATIONS
/
ANALYTICS REAL-TIME DATA STORE
SQL / NOSQL
EVENT PROCESSING
SEARCH
STREAMING
BATCH
MACHINE LEARNING
MICROSERVICES + CONTAINERSTRANSACTIONAL ANALYTICS FAST DATA + ANALYTICS
-
ANALYTICS & BIG DATA APPS & MICROSERVICES
MICROSERVICES(REST)
EVENTPROCESSING
RPC & MAP/REDUCE
.NET JAVA MICROSERVICES(REST)
EVENTPROCESSING
WEB CONTAINERS RPC & MAP/REDUCEDATA MODELS(SPATIAL, POJO, JSON)
EVENTPROCESSING
STREAMING
IN-MEMORYDATA GRID RAM SSD
STORAGESTORAGE-CLASSMEMORY
DATA REPLICATION& PERSISTENCE
CLUSTER MANAGEMENT & SERVICE DISCOVERY
SEARCH, BI & QUERY
SECU
RITY
AND
AUD
ITIN
G
MANA
GEME
NT A
ND M
ONIT
ORIN
G
REST
ORC
HEST
RATI
ON SPARK SQLMACHINELEARNINGSQL/JDBC SEARCH
MOBILE WEB IOT
ON-PREMISE CLOUD HYBRID
InsightEdge Architecture Overview
-
ANALYTICS & BIG DATA APPS & MICROSERVICES
MICROSERVICES(REST)
EVENTPROCESSING
RPC & MAP/REDUCE
.NET JAVA MICROSERVICES(REST)
EVENTPROCESSING
WEB CONTAINERS RPC & MAP/REDUCEDATA MODELS(SPATIAL, POJO, JSON)
EVENTPROCESSING
STREAMING
IN-MEMORYDATA GRID RAM SSD
STORAGESTORAGE-CLASSMEMORY
DATA REPLICATION& PERSISTENCE
CLUSTER MANAGEMENT & SERVICE DISCOVERY
SEARCH, BI & QUERY
SECU
RITY
AND
AUD
ITIN
G
MANA
GEME
NT A
ND M
ONIT
ORIN
G
REST
ORC
HEST
RATI
ON SPARK SQLMACHINELEARNINGSQL/JDBC SEARCH
MOBILE WEB IOT
ON-PREMISE CLOUD HYBRID
InsightEdgeUnifying Fast Data Analytics, AI and Transactional Processing
Clo
ud N
ativ
e M
anag
emen
t, O
rche
stra
tion,
and
M
onito
ring
Analytics and AISQL and BI Real-time Microservices
In-Memory Data Grid
Multi-Tiered Data torage and Replication
High Availability and Clustering
-
ANALYTICS & BIG DATA
STREAMING
CLUSTER MANAGEMENT & SERVICE DISCOVERY
SEARCH, BI & QUERY
SECU
RITY
AND
AUD
ITIN
G
MANA
GENE
NT A
ND M
ONIT
ORIN
G
MANA
GENE
NT A
ND M
ONIT
ORIN
G SPARKL SQLMACHINELEARNING
SQL/JDBC SEARCH
MOBILE WEB IOT
ON-PREMISIS CLOUD HYBRID
Ultra-low latency and high throughput transactional processing IMDG
RPC & MAP/REDUCE
WEB CONTAINERS RPC & MAP/REDUCEDATA MODELS(SPATIAL, POJO, JSON)
EVENTPROCESSING
IN-MEMORYDATA GRID RAM SSD
STORAGESTORAGE-CLASSMEMORY
DATA REPLICATION& PERSISTENCE
APPS & MICROSERVICES
MICROSERVICES(REST)
EVENTPROCESSING
.NET JAVA MICROSERVICES(REST)
EVENTPROCESSING
Partitioned In-Memory GridShared-nothing, linear scalability, elastic capacity
Co-Location of Data and Business LogicCo-located ops, event-driven, fast indexing
Event-Driven Processing and Map/Reduce
No DowntimeAuto-healing, multi-data center replication, fault tolerance
Fast Indexing Multi-Data ModelPOJO, .NET, Document/JSON, GeoSpatial
Seamless Integration wihJava/Scala ecosystem
Cloud Native
-
CLUSTER MANAGEMENT & SERVICE DISCOVERY
SEARCH, BI & QUERY
SECU
RITY
AND
AUD
ITIN
G
MANA
GENE
NT A
ND M
ONIT
ORIN
G
MANA
GENE
NT A
ND M
ONIT
ORIN
G SQL/JDBC SEARCH
MOBILE WEB IOT
ON-PREMISE CLOUD HYBRID
Co-located Analytics and AI with Transactional Processing
RPC & MAP/REDUCE
WEB CONTAINERS RPC & MAP/REDUCEDATA MODELS(SPATIAL, POJO, JSON)
EVENTPROCESSING
IN-MEMORYDATA GRID RAM SSD
STORAGESTORAGE-CLASSMEMORY
DATA REPLICATION& PERSISTENCE
APPS & MICROSERVICES
MICROSERVICES(REST)
EVENTPROCESSING
.NET JAVA MICROSERVICES(REST)
EVENTPROCESSING
ANALYTICS & BIG DATA
STREAMING SPARK SQLMACHINELEARNING
Full Spark Distribution
Push-down predicate for ultra-low latency filter (30x faster)
Shared RDDs/DataFrames
Streaming with 99.999% availability
Deep Learning with Intel BigDL
Graph processing, text mining, geospatial
SEARCH, BI & QUERY
SQL/JDBC SEARCH
Distributed SQL-99
Real-time integration with Tableau and Business Intelligence tools
JDBC driver
-
ANALYTICS & BIG DATA
STREAMING SPARKL SQLMACHINELEARNING
CLUSTER MANAGEMENT & SERVICE DISCOVERY
SECU
RITY
AND
AUD
ITIN
G
MANA
GENE
NT A
ND M
ONIT
ORIN
GMOBILE WEB IOT
ON-PREMISE CLOUD HYBRID
Multi-Tiered Data Storage and Replication
RPC & MAP/REDUCE
WEB CONTAINERS RPC & MAP/REDUCEDATA MODELS(SPATIAL, POJO, JSON)
EVENTPROCESSING
IN-MEMORYDATA GRID
APPS & MICROSERVICES
MICROSERVICES(REST)
EVENTPROCESSING
.NET JAVA MICROSERVICES(REST)
EVENTPROCESSING
SEARCH, BI & QUERY
SQL/JDBC SEARCH
RAM SSDSTORAGE
STORAGE-CLASSMEMORY
DATA REPLICATION& PERSISTENCE
REST
ORC
HEST
RATI
ONIn-Memory Data Processing (RAM)
Intelligent Data Tiering between RAM, SSD and Storage-Class Memory such as Intel 3DXPoint -Optane SSD/NVMe and Apache Pass *Leverages RocksDB
Multi-Data Center Replication
Asynchronous Persistence to SQL/NoSQL
* Apache Pass support in Q4 2018
-
MemoryXtend Module for Big Data – Optimized TCO
-
Cost Delay MORELESSHIGHERLOWER
The Right Information in the Right Layer at the Right TimeHigher Performance – Optimized TCO
Define which data resides on which layer per class and per field
-
ANALYTICS & BIG DATA
STREAMING SPARKL SQLMACHINELEARNING
SECU
RITY
AND
AUD
ITIN
G
MANA
GENE
NT A
ND M
ONIT
ORIN
GMOBILE WEB IOT
ON-PREMISE CLOUD HYBRID
High Availability & Clustering
RPC & MAP/REDUCE
WEB CONTAINERS RPC & MAP/REDUCEDATA MODELS(SPATIAL, POJO, JSON)
EVENTPROCESSING
IN-MEMORYDATA GRID
APPS & MICROSERVICES
MICROSERVICES(REST)
EVENTPROCESSING
.NET JAVA MICROSERVICES(REST)
EVENTPROCESSING
SEARCH, BI & QUERY
SQL/JDBC SEARCH
RAM SSDSTORAGE
STORAGE-CLASSMEMORY
DATA REPLICATION& PERSISTENCE
CLUSTER MANAGEMENT & SERVICE DISCOVERY
REST
ORC
HEST
RATI
ONZooKeeper-based clustering for 1000s of nodes
Back-up and auto-healing for each grid container
N + 1 redundancy
Unicast or Multicast discovery
-
ANALYTICS & BIG DATA APPS & MICROSERVICES
MICROSERVICES(REST)
EVENTPROCESSING
RPC & MAP/REDUCE
.NET JAVA MICROSERVICES(REST)
EVENTPROCESSING
WEB CONTAINERS RPC & MAP/REDUCEDATA MODELS(SPATIAL, POJO, JSON)
EVENTPROCESSING
STREAMING
IN-MEMORYDATA GRID RAM SSD
STORAGESTORAGE-CLASSMEMORY
DATA REPLICATION& PERSISTENCE
CLUSTER MANAGEMENT & SERVICE DISCOVERY
SEARCH, BI & QUERY
SECU
RITY
AND
AUD
ITIN
G
MANA
GEME
NT A
ND M
ONIT
ORIN
G
REST
ORC
HEST
RATI
ON SPARK SQLMACHINELEARNINGSQL/JDBC SEARCH
MOBILE WEB IOT
ON-PREMISE CLOUD HYBRID
InsightEdge Architecture Overview
-
ANALYTICS & BIG DATA APPS & MICROSERVICES
MICROSERVICES(REST)
EVENTPROCESSING
RPC & MAP/REDUCE
.NET JAVA MICROSERVICES(REST)
EVENTPROCESSING
WEB CONTAINERS RPC & MAP/REDUCEDATA MODELS(SPATIAL, POJO, JSON)
EVENTPROCESSING
STREAMING
IN-MEMORYDATA GRID RAM SSD
STORAGESTORAGE-CLASSMEMORY
DATA REPLICATION& PERSISTENCE
CLUSTER MANAGEMENT & SERVICE DISCOVERY
SEARCH, BI & QUERY
SECU
RITY
AND
AUD
ITIN
G
MANA
GEME
NT A
ND M
ONIT
ORIN
G
REST
ORC
HEST
RATI
ON SPARK SQLBigDLSQL/JDBC SEARCH
MOBILE WEB IOT
ON-PREMISE CLOUD HYBRID
InsightEdge Architecture Overview
-
GigaSpacesInsightEdge Platform
In-memory insight platform unifying fast-data analytics,
artificial intelligence and real-time applications
Apache Spark
Open source big data analytics framework built around speed, ease of use, and sophisticated
analytics
BigDL
Distributed deep learning framework designed for
Apache Spark and Intel Xeon
Intel® Xeon® Scalable ProcessorEfficiency, Hybrid Cloud, Analytics, Security
-
Fast Data Analytics Requires Real-Time Data Processing
• Unifies analytics, AI and real-time applications • Reduces cluster and component sprawl• Triggers transactional workflows based on prediction criteria and scoring
• Efficient scale-out computing• Distributed model training
• Lowers TCO/Decreases Deployment Costs – train and run large-scale deep learning workloads on Intel Xeon Scalable processors without relying on GPUs
• High performance – optimized with Intel® Math Kernel Library (Intel® MKL)
-
Demo Architecture
-
Streaming, Machine Learning and AISolution Architecture
-
Automated stock trading using InsightEdge and Machine learning
-
Stop Pressing 0 Or *
Automatic routing to the right agent for the perfect personalized experience
I have a windows MAC
problem
training, prediction, and tuning
Route to the MAC expertNLP Processing
User speaks using web interface
Browser converts speech to text and sends to controller
Spark job listens on Kafka topic and
using BigDL model, creates prediction
Controller writes data to
InsightEdge and to Kafka topic
BiGDL writes Prediction to
InsightEdge data grid
InsightEdge event processor listens for Prediction data and routes call session
-
Innovate with Confidence
INSTANT INSIGHTS
EXTREME PERFORMANCE
TCO REDUCTION
TOTAL CONFIDENCE
-
Try it
http://insightedge.io
Learn it
http://software.intel.com/bigdl
Build it
http://github.com/InsightEdge
-
THANK YOU