Satish Mohan Your Data, Your Search Tuesday, 12 March 13
May 10, 2015
Satish Mohan
Your Data,Your Search
Tuesday, 12 March 13
Enterprises today are collecting and have access to more data points in their ecosystem then ever.
Tuesday, 12 March 13
File Store Example
File / Folder Navigation
Integration - Mount Points
Limited Metadata
Hierarchical Structure
Regular File Store
Tuesday, 12 March 13
• Find a document from December 2011 about transfer containing proposal and David
• Find the document received from John containing David and transfer
• Find the revisions of transfer document
File Store Example
File / Folder Navigation
Integration - Mount Points
Limited Metadata
Hierarchical Structure
Tuesday, 12 March 13
• Find a document from December 2011 about transfer containing proposal and David
• Find the document received from John containing David and transfer
• Find the revisions of transfer document
File Store Example
File / Folder Navigation
Integration - Mount Points
Limited Metadata
Hierarchical Structure Collections / Documents
Local / Distributed Integrations
Semantic Metadata
Declarative Queries
Automatic Indexing
Provenance
Automatic Organization
Virtual Collections
Regular File Store
Intelligent File Store
Tuesday, 12 March 13
ElasticSearch is an open source, scalable, distributed, cloud-ready, highly-available full-text search engine and database with powerful aggregation features, communicating by JSON over RESTful HTTP, based on Apache Lucene.
Tuesday, 12 March 13
Structured Data
Unstructured Data Data Refinery
Message Queues
Inverted index
Transaction Log Versioning
Source Document
Capture & Curate
Index
Streams
Analyse
Search
MemoryShared FS FS + MemoryLocal FS
Document Store
Playing with ElasticSearch
Tuesday, 12 March 13
Playing with ElasticSearchRivers
• Data flows from sources using Rivers
• Continues to add data as it flows
• Can be added, removed, configured dynamically
Tuesday, 12 March 13
Playing with ElasticSearchRivers
• Data flows from sources using Rivers
• Continues to add data as it flows
• Can be added, removed, configured dynamically
ES NodeData Source
Data Source
Data Source
River
River
River
ES Index
Tuesday, 12 March 13
Playing with ElasticSearchRivers
• Data flows from sources using Rivers
• Continues to add data as it flows
• Can be added, removed, configured dynamically
ES NodeData Source
Data Source
Data Source
River
River
River
ES Index
Tuesday, 12 March 13
Playing with ElasticSearchRiver Modules
• CouchDB • JDBC
• MongoDB • Solr
• Wikipedia • Jira
• Twitter • CSV
• ActiveMQ • FileSystem
• RabbitMQ • SysInfo
• NSQ • Logs
• RSS • LDAP
Tuesday, 12 March 13
Playing with ElasticSearchIndex
• Describes document structure to the search engine
• Automatically created with sensible defaults
• Explicit mapping can be provided (generally, a good idea)
• Simple:
• string, integer/long, float/double, boolean, and null
• Complex:
• array, object
Tuesday, 12 March 13
Structured Data
Unstructured Data Data Refinery
Message Queues
Inverted index
Transaction Log Versioning
Source Document Shards
Replication Load Balancing Nodes
Distributed
Capture & Curate
Index
Streams
Analyse
Search
MemoryShared FS FS + MemoryLocal FS
Document Store
Playing with ElasticSearch
Tuesday, 12 March 13
Playing with ElasticSearchDistributed Model
• Number of shards is the scaling unit [ #shards > #nodes ]
• each one is a separate Lucene index thus, many per-index settings are available
• Moving shards around is faster than splitting them (no reindex)
• Replicas also serves reads, allowing to scale search
• # of replicas can be updated dynamically after index creation
Node 1
user (0)
user (1)
Node 2
user1 (0)
user (1)
Node 3
user (0)
user2 (0)
Automatic Discovery Protocol
Replica
Shard
Tuesday, 12 March 13
Playing with ElasticSearchIndex Aliases
curl -X POST 'http://localhost:9200/_aliases' -d '{
"actions" : [{
"add" : { “index” : “users”,
“alias” : “user_1”,
“filter” : { “term” : { “user” : “1” } }, “routing” : “1”
}} ]
}'
Indexing and search happens on the alias, with automatic use of routing and filtering
Tuesday, 12 March 13
Playing with ElasticSearchIndex Aliases
curl -X POST 'http://localhost:9200/_aliases' -d
' {"actions" : [ { "add" : {
"index" : "user_1", "alias" : "users"
}
},{ "add" : {
"index" : "user_2", "alias" : "users"
}
} ]}'
users
user_1
user_2
curl -X GET "http://localhost:9200/users/_search?q=..."
Tuesday, 12 March 13
Structured Data
Unstructured Data Data Refinery
Message Queues
Inverted index
Transaction Log Versioning
Source Document Shards
Replication Load Balancing Nodes
Distributed
Capture & Curate
Index
Streams
Analyse
Search
Transport
HTTP WebSockets
Thrift
ZeroMQ
memcached
TCP
MemoryShared FS FS + MemoryLocal FS
Document Store
Playing with ElasticSearch
Tuesday, 12 March 13
Structured Data
Unstructured Data Data Refinery
Message Queues
Inverted index
Transaction Log Versioning
Source Document
Data Sources
Tokenisers
Retrieval Models
Structured Results
Language Bindings Transport
Shards
Replication Load Balancing Nodes
Distributed
Capture & Curate
Index
Streams
Analyse
Search
Transport
HTTP WebSockets
Thrift
ZeroMQ
memcached
TCP
Modules
Extend
MemoryShared FS FS + MemoryLocal FS
Document Store
Playing with ElasticSearch
Tuesday, 12 March 13
Structured Data
Unstructured Data Data Refinery
Message Queues
Inverted index
Transaction Log Versioning
Source Document
Data Sources
Tokenisers
Retrieval Models
Structured Results
Language Bindings Transport
Shards
Replication Load Balancing Nodes
Distributed
Zen EC2
Capture & Curate
Index
Streams
Analyse
Search
Transport
HTTP WebSockets
Thrift
ZeroMQ
memcached
TCP
Modules
Extend
Discovery
MemoryShared FS FS + MemoryLocal FS
Document Store
Playing with ElasticSearch
Tuesday, 12 March 13
Structured Data
Unstructured Data Data Refinery
Message Queues
Inverted index
Transaction Log Versioning
Source Document
Data Sources
Tokenisers
Retrieval Models
Structured Results
Language Bindings Transport
Shards
Replication Load Balancing Nodes
Distributed
Zen EC2
mvel Python
Groovy
Javascript
Capture & Curate
Index
Streams
Analyse
Search
Transport
HTTP WebSockets
Thrift
ZeroMQ
memcached
TCP
Modules
Extend
Script
Discovery
MemoryShared FS FS + MemoryLocal FS
Document Store
Playing with ElasticSearch
Tuesday, 12 March 13
Structured Data
Unstructured Data Data Refinery
Message Queues
Inverted index
Transaction Log Versioning
Source Document
Data Sources
Tokenisers
Retrieval Models
Structured Results
Language Bindings Transport
Shards
Replication Load Balancing Nodes
Distributed
Zen EC2
mvel Python
Groovy
Javascript
Capture & Curate
Index
Streams
Analyse
Search
Transport
HTTP WebSockets
Thrift
ZeroMQ
memcached
TCP
Modules
Extend
Script
Monitor
Discovery
MemoryShared FS FS + MemoryLocal FS
Document Store
Playing with ElasticSearch
Tuesday, 12 March 13
Structured Data
Unstructured Data Data Refinery
Message Queues
Inverted index
Transaction Log Versioning
Source Document
Data Sources
Tokenisers
Retrieval Models
Structured Results
Language Bindings Transport
Shards
Replication Load Balancing Nodes
Distributed
Zen EC2
mvel Python
Groovy
Javascript
Capture & Curate
Index
Streams
Analyse
Search
Transport
HTTP WebSockets
Thrift
ZeroMQ
memcached
TCP
Modules
Extend
Script
Monitor
Discovery
RESTful
MemoryShared FS FS + MemoryLocal FS
Document Store
Playing with ElasticSearch
Tuesday, 12 March 13
REST API : http://host:port/[ index]/[type]/[_action/id]HTTP Methods : GET, POST, PUT, DELETE
Playing with ElasticSearch
Tuesday, 12 March 13
REST API : http://host:port/[ index]/[type]/[_action/id]HTTP Methods : GET, POST, PUT, DELETE
Playing with ElasticSearch
Some Definitions.....
• index -> Like a database
• type -> Like a table
• id -> Like a row in a table
Tuesday, 12 March 13
Playing with ElasticSearchREST API : http://host:port/[index]/[type]/_action/id]HTTP Methods : GET, POST, PUT, DELETE
curl -X POST "http:// localhost:9200/art ic les/art ic le/1 " -d '{
" t i t le" : "E last icSearch Understands JSON!" , "body" : "E last icSearch not only “works” with JSON, i t understands i t ! Let ’s f i rst . . . " , "publ ished_on" : "2013/02/06 10:00:00" , " tags" : [ "search" , " json"] , "author" : {
" f i rst_name" : "Bruce" , " last_name" : "Croft" , "emai l" : "bruce@croft .org"
}} '
requ
est
Tuesday, 12 March 13
Playing with ElasticSearchREST API : http://host:port/[index]/[type]/_action/id]HTTP Methods : GET, POST, PUT, DELETE
curl -X POST "http:// localhost:9200/art ic les/art ic le/1 " -d '{
" t i t le" : "E last icSearch Understands JSON!" , "body" : "E last icSearch not only “works” with JSON, i t understands i t ! Let ’s f i rst . . . " , "publ ished_on" : "2013/02/06 10:00:00" , " tags" : [ "search" , " json"] , "author" : {
" f i rst_name" : "Bruce" , " last_name" : "Croft" , "emai l" : "bruce@croft .org"
}} '
{"ok" : true,"_ index":"art ic les" ,"_type":"art ic le" ,"_ id" :"1" ,"_version":1
}
requ
est
resp
onse
Tuesday, 12 March 13
Playing with ElasticSearchREST API : http://host:port/[index]/[type]/_action/id]HTTP Methods : GET, POST, PUT, DELETE
requ
est
curl -X GET "http://localhost:9200/articles/_search?q=author.first_name:BRUCE"
Tuesday, 12 March 13
Playing with ElasticSearchREST API : http://host:port/[index]/[type]/_action/id]HTTP Methods : GET, POST, PUT, DELETE
{"took":1,"t imed_out" : fa lse,"_shards" :{"total" :5 ,"successful" :5 ," fa i led" :0} ,"hits" : {
" total" :1 , "max_score":0.30685282, "hits" : [{ "_ index":"art ic les" , "_type":"art ic le" , "_ id" :"1" , "_score":0.30685282, "_source" :
{ " t i t le" : "E last icSearch Understands JSON!" , "body" : "E last icSearch not only “works” with JSON, i t understands i t ! Let ’s f i rst . . . " , "publ ished_on" : "2013/02/06 10:00:00" , " tags" : [ "search" , " json"] , "author" : {
" f i rst_name" : "Bruce" , " last_name" : "Croft" , "emai l" : "bruce@croft .org"
}} } ] } }
requ
est
curl -X GET "http://localhost:9200/articles/_search?q=author.first_name:BRUCE"
resp
onse
Tuesday, 12 March 13
Playing with ElasticSearchREST API : http://host:port/[index]/[type]/_action/id]HTTP Methods : GET, POST, PUT, DELETE
{"took":1,"t imed_out" : fa lse,"_shards" :{"total" :5 ,"successful" :5 ," fa i led" :0} ,"hits" : {
"total" :1 , "max_score":0.30685282, "hits" : [{ "_index":"art ic les", "_type":"art ic le", "_ id":"1", "_score":0.30685282, "_source" :
{ "t i t le" : "Elast icSearch Understands JSON!", "body" : "Elast icSearch not only “works” with JSON, i t understands it ! Let ’s f irst . . ." , "published_on" : "2013/02/06 10:00:00", "tags" : ["search", " json"] , "author" : {
"f irst_name" : "Bruce", " last_name" : "Croft" , "email" : "[email protected]"
}}
} ] } }
requ
est
curl -X GET "http://localhost:9200/articles/_search?q=author.first_name:BRUCE"
resp
onse
Location & ID
Document Source
Total number of documents
Tuesday, 12 March 13
Structured Data
Unstructured Data Data Refinery
Message Queues
Inverted index
Transaction Log Versioning
Source Document
Data Sources
Tokenisers
Retrieval Models
Structured Results
Language Bindings Transport
Shards
Replication Load Balancing Nodes
Distributed
Zen EC2
mvel Python
Groovy
Javascript
Capture & Curate
Index
Streams
Analyse
Search
Transport
HTTP WebSockets
Thrift
ZeroMQ
memcached
TCP
Modules
Extend
Script
Monitor
Discovery
RESTful Micro Apps
MemoryShared FS FS + MemoryLocal FS
Document Store
Playing with ElasticSearch
Tuesday, 12 March 13
Structured Data
Unstructured Data Data Refinery
Message Queues
Inverted index
Transaction Log Versioning
Source Document
Data Sources
Tokenisers
Retrieval Models
Structured Results
Language Bindings Transport
Shards
Replication Load Balancing Nodes
Distributed
Zen EC2
mvel Python
Groovy
Javascript
HTML5/CSS3 Javascript
Capture & Curate
Index
Streams
Analyse
Search
Transport
HTTP WebSockets
Thrift
ZeroMQ
memcached
TCP
Modules
Extend
Script
Monitor
Discovery
RESTful Micro Apps
MemoryShared FS FS + MemoryLocal FS
Document Store
Playing with ElasticSearch
Tuesday, 12 March 13
Micro Applications
Rich, interactive single-page web applications powered by JavaScript, HTML and CSS.
Tuesday, 12 March 13
Micro Applications
Tuesday, 12 March 13
Micro Applications
Tuesday, 12 March 13
Micro Applications
Tuesday, 12 March 13
Micro Applications
Tuesday, 12 March 13
Micro ApplicationsRich, interactive single-page web applications powered by JavaScript, HTML and CSS.
• A self-described framework for ambitious applications
• Rails-inspired “convention over configuration” approach
• High level abstractions, two-way binding and auto-updating templates
Data Model
ControllerRouter
View
Model Model
Controller
View View
View
Tuesday, 12 March 13
Micro ApplicationsRich, interactive single-page web applications powered by JavaScript, HTML and CSS.
• A self-described framework for ambitious applications
• Rails-inspired “convention over configuration” approach
• High level abstractions, two-way binding and auto-updating templates
• Ember Data
• Client side storage adapter
• Provides a common interface to persist application data
• RESTful HTTP service - primary endpoint
• Browser’s localStorage
• Emerging web databases such as IndexedDB
Data Model
ControllerRouter
View
Model Model
Controller
View View
View
Tuesday, 12 March 13
Playing with ElasticSearchMore Features.....
• document oriented • load balancing
• versioning • plugins
• parent/child docs • more_like_this
• scripting • multi_field mapping
• dynamic mapping templates • percolation
• bulk indexing • facets
• geo location • index aliases
• auto-complete • ngrams & edge-ngrams
• histograms • rivers
Tuesday, 12 March 13
Structured Data
Unstructured Data Data Refinery
Message Queues
Inverted index
Transaction Log Versioning
Source Document
Data Sources
Tokenisers
Retrieval Models
Structured Results
Language Bindings Transport
Shards
Replication Load Balancing Nodes
Distributed
Zen EC2
mvel Python
Groovy
Javascript
HTML5/CSS3 Javascript
Capture & Curate
Index
Streams
Analyse
Search
Transport
HTTP WebSockets
Thrift
ZeroMQ
memcached
TCP
Modules
Extend
Script
Monitor
Discovery
RESTful Micro Apps
MemoryShared FS FS + MemoryLocal FS
Document Store
An alternative that would allow scientists or even casual users to perform analysis of distributed data regardless of where the data resides.
Tuesday, 12 March 13
Search is the primary interface for getting information today. Let’s build on it.
Search
DiscoverAnalyse
Tuesday, 12 March 13
Tuesday, 12 March 13
Data Management Tools - Challenges
• Interactive queries, data exploration or iterative query refinement poses significant challenges for current methods
• Building and running jobs and queries requires deep understanding of cluster size and structure, job performance, etc.
• Time-consuming to set up, deploy and use
Tuesday, 12 March 13