SlideShare a Scribd company logo
1 of 22
Download to read offline
GC Tuning
Call for Papers open through May 8
lucenerevolution.org
October 13-16 Ÿ Austin, TX
•  Time-based partitioning scheme
•  Transient collections for high-volume indexing
•  Schema design considerations
•  Q&A
Webinar: Fast Log Indexing with Fusion / Solr
Large-scale log analytics
•  Index billions of log events per day, near real-time
•  Analyze logs over time: today, recent, past week,
past 30 days, …
•  Easy to use dashboards to visualize common
questions and allow for ad hoc analysis
•  Ability to scale linearly as business grows …
with sub-linear growth in costs!
•  Easy to setup, easy to manage, easy to use
•  Combines the power of open
source Solr, Logstash, Kibana, and
more to deliver a feature rich UI for
search and data visualization.
•  Gives user the power to perform ad
hoc search and analysis on massive
amounts of multi-structured and
time series data stored in Solr.
•  Users can quickly transform results
into sharable visualizations and
dashboards.
Lucidworks Fusion Dashboard
Why Fusion / Solr for Log Analytics?
•  Speed
•  Proven scalability
•  Community
•  Cost effective
Fusion Collection
A collection is a distributed index defined by:
•  configuration stored in ZooKeeper (schema.xml, solrconfig.xml, …)
•  one or more shards: documents are distributed across N partitions of the index
•  document routing strategy: how documents get assigned to shards
•  replication factor: how many copies of each document in the collection
•  replica placement strategy: rack awareness, etc (see SOLR-6620)
Sharding increases parallelism during indexing and query execution
Replication enables load-balancing and fault-tolerance
https://twitter.com/bretthoerner/status/476830302430437376	
  
Fusion
Log Analytics
Dashboard
logs_feb26
(daily collection)
logs_feb25
(daily collection)
logs_feb01
(daily collection)
h00
(shard)
h22
(shard)
h23
(shard)
h00
(shard)
h22
(shard)
h23
(shard)
Add replicas
to support higher
query volume &
fault-tolerance
recent_logs
(colllection alias)
Use a collection
alias to make multiple
collections look like a
single collection; minimize
exposure to partitioning
strategy in client layer
Every daily collection has 24 shards (h00-h23), each covering 1-hour blocks of log messages
Time-based Data Partitioning Scheme
Time-based Data Partitioning Benefits
•  Optimizing read performance for write-once, read-many-times
type data
•  Recent data stays “hot” in Solr memory-based caches and OS
cache
•  Queries can be directed to specific partitions based on time
range filters
Show me errors occurring in my database layer in the past 10 minutes
Visualize activity for the past 2 hours …
•  Easy to drop or move older data to more cost-effective hardware
•  Easier to scale-out by adding more nodes, no shard-splitting,
rebalancing, or reindexing
But …
Time-based Data Partitioning Limitations
•  Indexing into a single hourly shard cannot keep up with write-volume
•  15-20K docs/sec per shard is normal
•  Not unusual to need an avg. of 50K docs/sec with ability to scale to 120K
•  Shard being written to would be a hotspot that would hurt query performance
•  Partitioning details exposed to indexing applications
logs_feb26
(daily collection)
logs_feb25
(daily collection)
logs_feb01
(daily collection)
h00
(shard)
h22
(shard)
h23
(shard)
recent_logs
(colllection alias)
Time-based Data Partitioning with Transient Collection
shard4
shard3
shard2
shard1
Latest 1-hour shard
gets built from
merging shards at
time bucket boundary
logs_feb26_h23
(transient collection)
current_log_writer
(colllection alias)
log indexing app
(logstash, flume)
current_log_writer collection alias
rolls over to a new transient collection
every hour; the shards in the transient
collection are merged into the 1-hour
shard and added to the daily collection
Transient Collection Highlights
•  Hash-based document routing across all shards
•  Scale up the number of shards to meet peak write-volume reqts.
•  Indexing applications write to an alias, e.g. “current_log_writer”
•  Rollover to a new transient collection every hour
•  Use Solr merge index support to create a single, immutable hourly shard
•  Read alias includes current transient collection (to query very recent data)
•  Achieved >120K docs per second with replication in a cluster with 20 servers
Direct Updates to Leaders
server-sideclient-side
logs_feb26
(daily collection)
logs_feb25
(daily collection)
h00
(shard)
h22
(shard)
h23
(shard)
recent_logs
(colllection alias)
Merge Transient Collection into Daily Collection
current_log_writer
(colllection alias)
log indexing app
(logstash, flume)
shard4
shard3
shard2
shard1
logs_feb26_h23
(transient collection)
logs_feb27_h00
(transient collection)
h23
(shard)
2015-02-27 00:00:00
Solr merge
Nuts and Bolts
•  Most features are provided out-of-the-box with Fusion
- Connectors
- Banana Dashboard
- SolrCloud & Solr Scale Toolkit
- Collection Management
- Monitoring, security, etc …
•  Transient collection rollover
- Custom Solr UpdateRequestProcessor
- Creates new transient collection when needed
- Updates collection aliases (reads and writes)
- Tricky in a distributed cluster, uses ZooKeeper to coordinate
•  Merge transient shards into single hourly shard
- Custom Python script
- Pulls shards from transient collection to a single host (throttled)
- Runs Solr merge to create 1 index from many shards
- Activates new hourly shard & removes old transient
- Adds replicas for new hourly shard
Moving older data to less expensive storage
•  SSD preferred for recent data, older data can be on “spinning rust”
•  Move older data off SSD to slower disks
- ADDREPLICA API to add new replica on another host
- Solr 5+ supports throttling replication transfer speed
- or, simply scp –l 500 … (-l for limit copy at 0.5 Mbit/s)
Schema Design
•  At large scale, disk (SSD) and memory are still expensive!
•  Balance flexibility with performance and efficiency (disk usage)
•  Fast unique IDs
Use UUID v1 if possible: http://johannburkard.de/software/uuid/
http://blog.mikemccandless.com/2014/05/choosing-fast-unique-identifier-uuid.html
•  DocValues for fields that you facet and sort on frequently
•  Trie-based date for range queries
•  Catch all text field for improved recall
DocValues
•  Faster sorting and faceting, both of which are used extensively in log analytics
•  Column-oriented view of a field in a Solr index (uninverted)
•  Disk & OS cache vs. Java Heap
“… we tuned everything about the default codec to keep all the data structures on disk but still be fast.”
Robert Muir, Lucene/Solr Committer, LUCENE-5761
•  Helps reduce pressure on Java GC
GC Tuning
•  Dashboards put a lot of pressure on the Java GC
•  Stop-the-world GC pauses can lead to ZooKeeper session expiration (which is always bad)
•  Good success with 32gb max heap (-Xmx32g) but have to tune it …
•  MMapDirectory relies on sufficient memory available to the OS cache (off-heap)
•  Enable verbose GC logging (even in prod) so you can troubleshoot issues:
-­‐verbose:gc	
  –Xloggc:gc.log	
  -­‐XX:+PrintHeapAtGC	
  -­‐XX:+PrintGCDetails	
  	
  
-­‐XX:+PrintGCDateStamps	
  -­‐XX:+PrintGCCause	
  	
  
-­‐XX:+PrintTenuringDistribution	
  -­‐XX:+PrintGCApplicationStoppedTime	
  
GC Tuning
-­‐XX:CMSTriggerPermRatio=80	
  
-­‐XX:CMSFullGCsBeforeCompaction=1	
  
-­‐XX:+ParallelRefProcEnabled	
  
-­‐XX:+CMSParallelRemarkEnabled	
  
-­‐XX:CMSMaxAbortablePrecleanTime=6000	
  
-­‐XX:CMSInitiatingOccupancyFraction=50	
  
-­‐XX:+UseCMSInitiatingOccupancyOnly	
  
-­‐XX:PretenureSizeThreshold=128m	
  
-­‐XX:+CMSScavengeBeforeRemark	
  
-­‐XX:ParallelGCThreads=9	
  
-­‐XX:ConcGCThreads=9	
  
-­‐XX:+UseParNewGC	
  
-­‐XX:+UseConcMarkSweepGC	
  
-­‐XX:MaxTenuringThreshold=12	
  
-­‐XX:TargetSurvivorRatio=90	
  
-­‐XX:SurvivorRatio=6	
  
-­‐XX:NewRatio=5	
  
-­‐Xmx32g	
  
-­‐Xms16g	
  
-­‐Xss256k	
  
Thank you.
Download Fusion: http://lucidworks.com/fusion/
Webinar recording will be available soon.
Questions?
Webinar: Faster Log Indexing with Fusion

More Related Content

What's hot

ScyllaDB: What could you do with Cassandra compatibility at 1.8 million reque...
ScyllaDB: What could you do with Cassandra compatibility at 1.8 million reque...ScyllaDB: What could you do with Cassandra compatibility at 1.8 million reque...
ScyllaDB: What could you do with Cassandra compatibility at 1.8 million reque...
Data Con LA
 

What's hot (20)

What database
What databaseWhat database
What database
 
Scylla Summit 2016: Scylla at Samsung SDS
Scylla Summit 2016: Scylla at Samsung SDSScylla Summit 2016: Scylla at Samsung SDS
Scylla Summit 2016: Scylla at Samsung SDS
 
Scylla Summit 2016: Why Kenshoo is about to displace Cassandra with Scylla
Scylla Summit 2016: Why Kenshoo is about to displace Cassandra with ScyllaScylla Summit 2016: Why Kenshoo is about to displace Cassandra with Scylla
Scylla Summit 2016: Why Kenshoo is about to displace Cassandra with Scylla
 
Cassandra Summit 2014: Apache Cassandra Best Practices at Ebay
Cassandra Summit 2014: Apache Cassandra Best Practices at EbayCassandra Summit 2014: Apache Cassandra Best Practices at Ebay
Cassandra Summit 2014: Apache Cassandra Best Practices at Ebay
 
Scylla Summit 2016: Outbrain Case Study - Lowering Latency While Doing 20X IO...
Scylla Summit 2016: Outbrain Case Study - Lowering Latency While Doing 20X IO...Scylla Summit 2016: Outbrain Case Study - Lowering Latency While Doing 20X IO...
Scylla Summit 2016: Outbrain Case Study - Lowering Latency While Doing 20X IO...
 
Introducing Cloudian HyperStore 6.0
Introducing Cloudian HyperStore 6.0Introducing Cloudian HyperStore 6.0
Introducing Cloudian HyperStore 6.0
 
Disney+ Hotstar: Scaling NoSQL for Millions of Video On-Demand Users
Disney+ Hotstar: Scaling NoSQL for Millions of Video On-Demand UsersDisney+ Hotstar: Scaling NoSQL for Millions of Video On-Demand Users
Disney+ Hotstar: Scaling NoSQL for Millions of Video On-Demand Users
 
HBaseCon 2015 General Session: Zen - A Graph Data Model on HBase
HBaseCon 2015 General Session: Zen - A Graph Data Model on HBaseHBaseCon 2015 General Session: Zen - A Graph Data Model on HBase
HBaseCon 2015 General Session: Zen - A Graph Data Model on HBase
 
Run Cloud Native MySQL NDB Cluster in Kubernetes
Run Cloud Native MySQL NDB Cluster in KubernetesRun Cloud Native MySQL NDB Cluster in Kubernetes
Run Cloud Native MySQL NDB Cluster in Kubernetes
 
How to Monitor and Size Workloads on AWS i3 instances
How to Monitor and Size Workloads on AWS i3 instancesHow to Monitor and Size Workloads on AWS i3 instances
How to Monitor and Size Workloads on AWS i3 instances
 
Paul Dix [InfluxData] | InfluxDays Opening Keynote | InfluxDays Virtual Exper...
Paul Dix [InfluxData] | InfluxDays Opening Keynote | InfluxDays Virtual Exper...Paul Dix [InfluxData] | InfluxDays Opening Keynote | InfluxDays Virtual Exper...
Paul Dix [InfluxData] | InfluxDays Opening Keynote | InfluxDays Virtual Exper...
 
August 2013 HUG: Removing the NameNode's memory limitation
August 2013 HUG: Removing the NameNode's memory limitation August 2013 HUG: Removing the NameNode's memory limitation
August 2013 HUG: Removing the NameNode's memory limitation
 
Seastar Summit 2019 Keynote
Seastar Summit 2019 KeynoteSeastar Summit 2019 Keynote
Seastar Summit 2019 Keynote
 
Cassandra Community Webinar: Apache Spark Analytics at The Weather Channel - ...
Cassandra Community Webinar: Apache Spark Analytics at The Weather Channel - ...Cassandra Community Webinar: Apache Spark Analytics at The Weather Channel - ...
Cassandra Community Webinar: Apache Spark Analytics at The Weather Channel - ...
 
ScyllaDB: What could you do with Cassandra compatibility at 1.8 million reque...
ScyllaDB: What could you do with Cassandra compatibility at 1.8 million reque...ScyllaDB: What could you do with Cassandra compatibility at 1.8 million reque...
ScyllaDB: What could you do with Cassandra compatibility at 1.8 million reque...
 
C* Summit 2013: Cassandra at eBay Scale by Feng Qu and Anurag Jambhekar
C* Summit 2013: Cassandra at eBay Scale by Feng Qu and Anurag JambhekarC* Summit 2013: Cassandra at eBay Scale by Feng Qu and Anurag Jambhekar
C* Summit 2013: Cassandra at eBay Scale by Feng Qu and Anurag Jambhekar
 
NewSQL overview, Feb 2015
NewSQL overview, Feb 2015NewSQL overview, Feb 2015
NewSQL overview, Feb 2015
 
DatEngConf SF16 - Apache Kudu: Fast Analytics on Fast Data
DatEngConf SF16 - Apache Kudu: Fast Analytics on Fast DataDatEngConf SF16 - Apache Kudu: Fast Analytics on Fast Data
DatEngConf SF16 - Apache Kudu: Fast Analytics on Fast Data
 
HBaseConAsia2018 Keynote1: Apache HBase Project Status
HBaseConAsia2018 Keynote1: Apache HBase Project StatusHBaseConAsia2018 Keynote1: Apache HBase Project Status
HBaseConAsia2018 Keynote1: Apache HBase Project Status
 
MySQL NDB Cluster 8.0 SQL faster than NoSQL
MySQL NDB Cluster 8.0 SQL faster than NoSQL MySQL NDB Cluster 8.0 SQL faster than NoSQL
MySQL NDB Cluster 8.0 SQL faster than NoSQL
 

Similar to Webinar: Faster Log Indexing with Fusion

Rackspace: Email's Solution for Indexing 50K Documents per Second: Presented ...
Rackspace: Email's Solution for Indexing 50K Documents per Second: Presented ...Rackspace: Email's Solution for Indexing 50K Documents per Second: Presented ...
Rackspace: Email's Solution for Indexing 50K Documents per Second: Presented ...
Lucidworks
 

Similar to Webinar: Faster Log Indexing with Fusion (20)

Alluxio 2.0 & Near Real-time Big Data Platform w/ Spark & Alluxio
Alluxio 2.0 & Near Real-time Big Data Platform w/ Spark & AlluxioAlluxio 2.0 & Near Real-time Big Data Platform w/ Spark & Alluxio
Alluxio 2.0 & Near Real-time Big Data Platform w/ Spark & Alluxio
 
Mail Search As A Sercive: Presented by Rishi Easwaran, Aol
Mail Search As A Sercive: Presented by Rishi Easwaran, AolMail Search As A Sercive: Presented by Rishi Easwaran, Aol
Mail Search As A Sercive: Presented by Rishi Easwaran, Aol
 
Taking Splunk to the Next Level - Architecture Breakout Session
Taking Splunk to the Next Level - Architecture Breakout SessionTaking Splunk to the Next Level - Architecture Breakout Session
Taking Splunk to the Next Level - Architecture Breakout Session
 
Highlights of AWS ReInvent 2023 (Announcements and Best Practices)
Highlights of AWS ReInvent 2023 (Announcements and Best Practices)Highlights of AWS ReInvent 2023 (Announcements and Best Practices)
Highlights of AWS ReInvent 2023 (Announcements and Best Practices)
 
Solving Office 365 Big Challenges using Cassandra + Spark
Solving Office 365 Big Challenges using Cassandra + Spark Solving Office 365 Big Challenges using Cassandra + Spark
Solving Office 365 Big Challenges using Cassandra + Spark
 
Rackspace: Email's Solution for Indexing 50K Documents per Second: Presented ...
Rackspace: Email's Solution for Indexing 50K Documents per Second: Presented ...Rackspace: Email's Solution for Indexing 50K Documents per Second: Presented ...
Rackspace: Email's Solution for Indexing 50K Documents per Second: Presented ...
 
Building a near real time search engine & analytics for logs using solr
Building a near real time search engine & analytics for logs using solrBuilding a near real time search engine & analytics for logs using solr
Building a near real time search engine & analytics for logs using solr
 
Managing Security At 1M Events a Second using Elasticsearch
Managing Security At 1M Events a Second using ElasticsearchManaging Security At 1M Events a Second using Elasticsearch
Managing Security At 1M Events a Second using Elasticsearch
 
Solr 4
Solr 4Solr 4
Solr 4
 
A Closer Look at Apache Kudu
A Closer Look at Apache KuduA Closer Look at Apache Kudu
A Closer Look at Apache Kudu
 
Benchmarking Solr Performance at Scale
Benchmarking Solr Performance at ScaleBenchmarking Solr Performance at Scale
Benchmarking Solr Performance at Scale
 
CosmosDB for DBAs & Developers
CosmosDB for DBAs & DevelopersCosmosDB for DBAs & Developers
CosmosDB for DBAs & Developers
 
Pulsar - flexible pub-sub for internet scale
Pulsar - flexible pub-sub for internet scalePulsar - flexible pub-sub for internet scale
Pulsar - flexible pub-sub for internet scale
 
Dissecting Scalable Database Architectures
Dissecting Scalable Database ArchitecturesDissecting Scalable Database Architectures
Dissecting Scalable Database Architectures
 
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage CostLeverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
 
Introduction to SolrCloud
Introduction to SolrCloudIntroduction to SolrCloud
Introduction to SolrCloud
 
Cosmos db
Cosmos dbCosmos db
Cosmos db
 
Scalable relational database with SQL Azure
Scalable relational database with SQL AzureScalable relational database with SQL Azure
Scalable relational database with SQL Azure
 
Cosmos DB at VLDB 2019
Cosmos DB at VLDB 2019Cosmos DB at VLDB 2019
Cosmos DB at VLDB 2019
 
MyHeritage backend group - build to scale
MyHeritage backend group - build to scaleMyHeritage backend group - build to scale
MyHeritage backend group - build to scale
 

More from Lucidworks

Intelligent Insight Driven Policing with MC+A, Toronto Police Service and Luc...
Intelligent Insight Driven Policing with MC+A, Toronto Police Service and Luc...Intelligent Insight Driven Policing with MC+A, Toronto Police Service and Luc...
Intelligent Insight Driven Policing with MC+A, Toronto Police Service and Luc...
Lucidworks
 

More from Lucidworks (20)

Search is the Tip of the Spear for Your B2B eCommerce Strategy
Search is the Tip of the Spear for Your B2B eCommerce StrategySearch is the Tip of the Spear for Your B2B eCommerce Strategy
Search is the Tip of the Spear for Your B2B eCommerce Strategy
 
Drive Agent Effectiveness in Salesforce
Drive Agent Effectiveness in SalesforceDrive Agent Effectiveness in Salesforce
Drive Agent Effectiveness in Salesforce
 
How Crate & Barrel Connects Shoppers with Relevant Products
How Crate & Barrel Connects Shoppers with Relevant ProductsHow Crate & Barrel Connects Shoppers with Relevant Products
How Crate & Barrel Connects Shoppers with Relevant Products
 
Lucidworks & IMRG Webinar – Best-In-Class Retail Product Discovery
Lucidworks & IMRG Webinar – Best-In-Class Retail Product DiscoveryLucidworks & IMRG Webinar – Best-In-Class Retail Product Discovery
Lucidworks & IMRG Webinar – Best-In-Class Retail Product Discovery
 
Connected Experiences Are Personalized Experiences
Connected Experiences Are Personalized ExperiencesConnected Experiences Are Personalized Experiences
Connected Experiences Are Personalized Experiences
 
Intelligent Insight Driven Policing with MC+A, Toronto Police Service and Luc...
Intelligent Insight Driven Policing with MC+A, Toronto Police Service and Luc...Intelligent Insight Driven Policing with MC+A, Toronto Police Service and Luc...
Intelligent Insight Driven Policing with MC+A, Toronto Police Service and Luc...
 
[Webinar] Intelligent Policing. Leveraging Data to more effectively Serve Com...
[Webinar] Intelligent Policing. Leveraging Data to more effectively Serve Com...[Webinar] Intelligent Policing. Leveraging Data to more effectively Serve Com...
[Webinar] Intelligent Policing. Leveraging Data to more effectively Serve Com...
 
Preparing for Peak in Ecommerce | eTail Asia 2020
Preparing for Peak in Ecommerce | eTail Asia 2020Preparing for Peak in Ecommerce | eTail Asia 2020
Preparing for Peak in Ecommerce | eTail Asia 2020
 
Accelerate The Path To Purchase With Product Discovery at Retail Innovation C...
Accelerate The Path To Purchase With Product Discovery at Retail Innovation C...Accelerate The Path To Purchase With Product Discovery at Retail Innovation C...
Accelerate The Path To Purchase With Product Discovery at Retail Innovation C...
 
AI-Powered Linguistics and Search with Fusion and Rosette
AI-Powered Linguistics and Search with Fusion and RosetteAI-Powered Linguistics and Search with Fusion and Rosette
AI-Powered Linguistics and Search with Fusion and Rosette
 
The Service Industry After COVID-19: The Soul of Service in a Virtual Moment
The Service Industry After COVID-19: The Soul of Service in a Virtual MomentThe Service Industry After COVID-19: The Soul of Service in a Virtual Moment
The Service Industry After COVID-19: The Soul of Service in a Virtual Moment
 
Webinar: Smart answers for employee and customer support after covid 19 - Europe
Webinar: Smart answers for employee and customer support after covid 19 - EuropeWebinar: Smart answers for employee and customer support after covid 19 - Europe
Webinar: Smart answers for employee and customer support after covid 19 - Europe
 
Smart Answers for Employee and Customer Support After COVID-19
Smart Answers for Employee and Customer Support After COVID-19Smart Answers for Employee and Customer Support After COVID-19
Smart Answers for Employee and Customer Support After COVID-19
 
Applying AI & Search in Europe - featuring 451 Research
Applying AI & Search in Europe - featuring 451 ResearchApplying AI & Search in Europe - featuring 451 Research
Applying AI & Search in Europe - featuring 451 Research
 
Webinar: Accelerate Data Science with Fusion 5.1
Webinar: Accelerate Data Science with Fusion 5.1Webinar: Accelerate Data Science with Fusion 5.1
Webinar: Accelerate Data Science with Fusion 5.1
 
Webinar: 5 Must-Have Items You Need for Your 2020 Ecommerce Strategy
Webinar: 5 Must-Have Items You Need for Your 2020 Ecommerce StrategyWebinar: 5 Must-Have Items You Need for Your 2020 Ecommerce Strategy
Webinar: 5 Must-Have Items You Need for Your 2020 Ecommerce Strategy
 
Where Search Meets Science and Style Meets Savings: Nordstrom Rack's Journey ...
Where Search Meets Science and Style Meets Savings: Nordstrom Rack's Journey ...Where Search Meets Science and Style Meets Savings: Nordstrom Rack's Journey ...
Where Search Meets Science and Style Meets Savings: Nordstrom Rack's Journey ...
 
Apply Knowledge Graphs and Search for Real-World Decision Intelligence
Apply Knowledge Graphs and Search for Real-World Decision IntelligenceApply Knowledge Graphs and Search for Real-World Decision Intelligence
Apply Knowledge Graphs and Search for Real-World Decision Intelligence
 
Webinar: Building a Business Case for Enterprise Search
Webinar: Building a Business Case for Enterprise SearchWebinar: Building a Business Case for Enterprise Search
Webinar: Building a Business Case for Enterprise Search
 
Why Insight Engines Matter in 2020 and Beyond
Why Insight Engines Matter in 2020 and BeyondWhy Insight Engines Matter in 2020 and Beyond
Why Insight Engines Matter in 2020 and Beyond
 

Recently uploaded

AI Mastery 201: Elevating Your Workflow with Advanced LLM Techniques
AI Mastery 201: Elevating Your Workflow with Advanced LLM TechniquesAI Mastery 201: Elevating Your Workflow with Advanced LLM Techniques
AI Mastery 201: Elevating Your Workflow with Advanced LLM Techniques
VictorSzoltysek
 
TECUNIQUE: Success Stories: IT Service provider
TECUNIQUE: Success Stories: IT Service providerTECUNIQUE: Success Stories: IT Service provider
TECUNIQUE: Success Stories: IT Service provider
mohitmore19
 
+971565801893>>SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHAB...
+971565801893>>SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHAB...+971565801893>>SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHAB...
+971565801893>>SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHAB...
Health
 

Recently uploaded (20)

Learn the Fundamentals of XCUITest Framework_ A Beginner's Guide.pdf
Learn the Fundamentals of XCUITest Framework_ A Beginner's Guide.pdfLearn the Fundamentals of XCUITest Framework_ A Beginner's Guide.pdf
Learn the Fundamentals of XCUITest Framework_ A Beginner's Guide.pdf
 
Right Money Management App For Your Financial Goals
Right Money Management App For Your Financial GoalsRight Money Management App For Your Financial Goals
Right Money Management App For Your Financial Goals
 
Azure_Native_Qumulo_High_Performance_Compute_Benchmarks.pdf
Azure_Native_Qumulo_High_Performance_Compute_Benchmarks.pdfAzure_Native_Qumulo_High_Performance_Compute_Benchmarks.pdf
Azure_Native_Qumulo_High_Performance_Compute_Benchmarks.pdf
 
A Secure and Reliable Document Management System is Essential.docx
A Secure and Reliable Document Management System is Essential.docxA Secure and Reliable Document Management System is Essential.docx
A Secure and Reliable Document Management System is Essential.docx
 
HR Software Buyers Guide in 2024 - HRSoftware.com
HR Software Buyers Guide in 2024 - HRSoftware.comHR Software Buyers Guide in 2024 - HRSoftware.com
HR Software Buyers Guide in 2024 - HRSoftware.com
 
%in Bahrain+277-882-255-28 abortion pills for sale in Bahrain
%in Bahrain+277-882-255-28 abortion pills for sale in Bahrain%in Bahrain+277-882-255-28 abortion pills for sale in Bahrain
%in Bahrain+277-882-255-28 abortion pills for sale in Bahrain
 
The Guide to Integrating Generative AI into Unified Continuous Testing Platfo...
The Guide to Integrating Generative AI into Unified Continuous Testing Platfo...The Guide to Integrating Generative AI into Unified Continuous Testing Platfo...
The Guide to Integrating Generative AI into Unified Continuous Testing Platfo...
 
Define the academic and professional writing..pdf
Define the academic and professional writing..pdfDefine the academic and professional writing..pdf
Define the academic and professional writing..pdf
 
%in Midrand+277-882-255-28 abortion pills for sale in midrand
%in Midrand+277-882-255-28 abortion pills for sale in midrand%in Midrand+277-882-255-28 abortion pills for sale in midrand
%in Midrand+277-882-255-28 abortion pills for sale in midrand
 
10 Trends Likely to Shape Enterprise Technology in 2024
10 Trends Likely to Shape Enterprise Technology in 202410 Trends Likely to Shape Enterprise Technology in 2024
10 Trends Likely to Shape Enterprise Technology in 2024
 
AI Mastery 201: Elevating Your Workflow with Advanced LLM Techniques
AI Mastery 201: Elevating Your Workflow with Advanced LLM TechniquesAI Mastery 201: Elevating Your Workflow with Advanced LLM Techniques
AI Mastery 201: Elevating Your Workflow with Advanced LLM Techniques
 
W01_panagenda_Navigating-the-Future-with-The-Hitchhikers-Guide-to-Notes-and-D...
W01_panagenda_Navigating-the-Future-with-The-Hitchhikers-Guide-to-Notes-and-D...W01_panagenda_Navigating-the-Future-with-The-Hitchhikers-Guide-to-Notes-and-D...
W01_panagenda_Navigating-the-Future-with-The-Hitchhikers-Guide-to-Notes-and-D...
 
Shapes for Sharing between Graph Data Spaces - and Epistemic Querying of RDF-...
Shapes for Sharing between Graph Data Spaces - and Epistemic Querying of RDF-...Shapes for Sharing between Graph Data Spaces - and Epistemic Querying of RDF-...
Shapes for Sharing between Graph Data Spaces - and Epistemic Querying of RDF-...
 
TECUNIQUE: Success Stories: IT Service provider
TECUNIQUE: Success Stories: IT Service providerTECUNIQUE: Success Stories: IT Service provider
TECUNIQUE: Success Stories: IT Service provider
 
+971565801893>>SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHAB...
+971565801893>>SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHAB...+971565801893>>SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHAB...
+971565801893>>SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHAB...
 
Unveiling the Tech Salsa of LAMs with Janus in Real-Time Applications
Unveiling the Tech Salsa of LAMs with Janus in Real-Time ApplicationsUnveiling the Tech Salsa of LAMs with Janus in Real-Time Applications
Unveiling the Tech Salsa of LAMs with Janus in Real-Time Applications
 
Unlocking the Future of AI Agents with Large Language Models
Unlocking the Future of AI Agents with Large Language ModelsUnlocking the Future of AI Agents with Large Language Models
Unlocking the Future of AI Agents with Large Language Models
 
Direct Style Effect Systems - The Print[A] Example - A Comprehension Aid
Direct Style Effect Systems -The Print[A] Example- A Comprehension AidDirect Style Effect Systems -The Print[A] Example- A Comprehension Aid
Direct Style Effect Systems - The Print[A] Example - A Comprehension Aid
 
%in tembisa+277-882-255-28 abortion pills for sale in tembisa
%in tembisa+277-882-255-28 abortion pills for sale in tembisa%in tembisa+277-882-255-28 abortion pills for sale in tembisa
%in tembisa+277-882-255-28 abortion pills for sale in tembisa
 
Crypto Cloud Review - How To Earn Up To $500 Per DAY Of Bitcoin 100% On AutoP...
Crypto Cloud Review - How To Earn Up To $500 Per DAY Of Bitcoin 100% On AutoP...Crypto Cloud Review - How To Earn Up To $500 Per DAY Of Bitcoin 100% On AutoP...
Crypto Cloud Review - How To Earn Up To $500 Per DAY Of Bitcoin 100% On AutoP...
 

Webinar: Faster Log Indexing with Fusion

  • 1.
  • 2. GC Tuning Call for Papers open through May 8 lucenerevolution.org October 13-16 Ÿ Austin, TX
  • 3. •  Time-based partitioning scheme •  Transient collections for high-volume indexing •  Schema design considerations •  Q&A Webinar: Fast Log Indexing with Fusion / Solr
  • 4. Large-scale log analytics •  Index billions of log events per day, near real-time •  Analyze logs over time: today, recent, past week, past 30 days, … •  Easy to use dashboards to visualize common questions and allow for ad hoc analysis •  Ability to scale linearly as business grows … with sub-linear growth in costs! •  Easy to setup, easy to manage, easy to use
  • 5. •  Combines the power of open source Solr, Logstash, Kibana, and more to deliver a feature rich UI for search and data visualization. •  Gives user the power to perform ad hoc search and analysis on massive amounts of multi-structured and time series data stored in Solr. •  Users can quickly transform results into sharable visualizations and dashboards. Lucidworks Fusion Dashboard
  • 6. Why Fusion / Solr for Log Analytics? •  Speed •  Proven scalability •  Community •  Cost effective
  • 7. Fusion Collection A collection is a distributed index defined by: •  configuration stored in ZooKeeper (schema.xml, solrconfig.xml, …) •  one or more shards: documents are distributed across N partitions of the index •  document routing strategy: how documents get assigned to shards •  replication factor: how many copies of each document in the collection •  replica placement strategy: rack awareness, etc (see SOLR-6620) Sharding increases parallelism during indexing and query execution Replication enables load-balancing and fault-tolerance https://twitter.com/bretthoerner/status/476830302430437376  
  • 8. Fusion Log Analytics Dashboard logs_feb26 (daily collection) logs_feb25 (daily collection) logs_feb01 (daily collection) h00 (shard) h22 (shard) h23 (shard) h00 (shard) h22 (shard) h23 (shard) Add replicas to support higher query volume & fault-tolerance recent_logs (colllection alias) Use a collection alias to make multiple collections look like a single collection; minimize exposure to partitioning strategy in client layer Every daily collection has 24 shards (h00-h23), each covering 1-hour blocks of log messages Time-based Data Partitioning Scheme
  • 9. Time-based Data Partitioning Benefits •  Optimizing read performance for write-once, read-many-times type data •  Recent data stays “hot” in Solr memory-based caches and OS cache •  Queries can be directed to specific partitions based on time range filters Show me errors occurring in my database layer in the past 10 minutes Visualize activity for the past 2 hours … •  Easy to drop or move older data to more cost-effective hardware •  Easier to scale-out by adding more nodes, no shard-splitting, rebalancing, or reindexing But …
  • 10. Time-based Data Partitioning Limitations •  Indexing into a single hourly shard cannot keep up with write-volume •  15-20K docs/sec per shard is normal •  Not unusual to need an avg. of 50K docs/sec with ability to scale to 120K •  Shard being written to would be a hotspot that would hurt query performance •  Partitioning details exposed to indexing applications
  • 11. logs_feb26 (daily collection) logs_feb25 (daily collection) logs_feb01 (daily collection) h00 (shard) h22 (shard) h23 (shard) recent_logs (colllection alias) Time-based Data Partitioning with Transient Collection shard4 shard3 shard2 shard1 Latest 1-hour shard gets built from merging shards at time bucket boundary logs_feb26_h23 (transient collection) current_log_writer (colllection alias) log indexing app (logstash, flume) current_log_writer collection alias rolls over to a new transient collection every hour; the shards in the transient collection are merged into the 1-hour shard and added to the daily collection
  • 12. Transient Collection Highlights •  Hash-based document routing across all shards •  Scale up the number of shards to meet peak write-volume reqts. •  Indexing applications write to an alias, e.g. “current_log_writer” •  Rollover to a new transient collection every hour •  Use Solr merge index support to create a single, immutable hourly shard •  Read alias includes current transient collection (to query very recent data) •  Achieved >120K docs per second with replication in a cluster with 20 servers
  • 13. Direct Updates to Leaders server-sideclient-side
  • 14. logs_feb26 (daily collection) logs_feb25 (daily collection) h00 (shard) h22 (shard) h23 (shard) recent_logs (colllection alias) Merge Transient Collection into Daily Collection current_log_writer (colllection alias) log indexing app (logstash, flume) shard4 shard3 shard2 shard1 logs_feb26_h23 (transient collection) logs_feb27_h00 (transient collection) h23 (shard) 2015-02-27 00:00:00 Solr merge
  • 15. Nuts and Bolts •  Most features are provided out-of-the-box with Fusion - Connectors - Banana Dashboard - SolrCloud & Solr Scale Toolkit - Collection Management - Monitoring, security, etc … •  Transient collection rollover - Custom Solr UpdateRequestProcessor - Creates new transient collection when needed - Updates collection aliases (reads and writes) - Tricky in a distributed cluster, uses ZooKeeper to coordinate •  Merge transient shards into single hourly shard - Custom Python script - Pulls shards from transient collection to a single host (throttled) - Runs Solr merge to create 1 index from many shards - Activates new hourly shard & removes old transient - Adds replicas for new hourly shard
  • 16. Moving older data to less expensive storage •  SSD preferred for recent data, older data can be on “spinning rust” •  Move older data off SSD to slower disks - ADDREPLICA API to add new replica on another host - Solr 5+ supports throttling replication transfer speed - or, simply scp –l 500 … (-l for limit copy at 0.5 Mbit/s)
  • 17. Schema Design •  At large scale, disk (SSD) and memory are still expensive! •  Balance flexibility with performance and efficiency (disk usage) •  Fast unique IDs Use UUID v1 if possible: http://johannburkard.de/software/uuid/ http://blog.mikemccandless.com/2014/05/choosing-fast-unique-identifier-uuid.html •  DocValues for fields that you facet and sort on frequently •  Trie-based date for range queries •  Catch all text field for improved recall
  • 18. DocValues •  Faster sorting and faceting, both of which are used extensively in log analytics •  Column-oriented view of a field in a Solr index (uninverted) •  Disk & OS cache vs. Java Heap “… we tuned everything about the default codec to keep all the data structures on disk but still be fast.” Robert Muir, Lucene/Solr Committer, LUCENE-5761 •  Helps reduce pressure on Java GC
  • 19. GC Tuning •  Dashboards put a lot of pressure on the Java GC •  Stop-the-world GC pauses can lead to ZooKeeper session expiration (which is always bad) •  Good success with 32gb max heap (-Xmx32g) but have to tune it … •  MMapDirectory relies on sufficient memory available to the OS cache (off-heap) •  Enable verbose GC logging (even in prod) so you can troubleshoot issues: -­‐verbose:gc  –Xloggc:gc.log  -­‐XX:+PrintHeapAtGC  -­‐XX:+PrintGCDetails     -­‐XX:+PrintGCDateStamps  -­‐XX:+PrintGCCause     -­‐XX:+PrintTenuringDistribution  -­‐XX:+PrintGCApplicationStoppedTime  
  • 20. GC Tuning -­‐XX:CMSTriggerPermRatio=80   -­‐XX:CMSFullGCsBeforeCompaction=1   -­‐XX:+ParallelRefProcEnabled   -­‐XX:+CMSParallelRemarkEnabled   -­‐XX:CMSMaxAbortablePrecleanTime=6000   -­‐XX:CMSInitiatingOccupancyFraction=50   -­‐XX:+UseCMSInitiatingOccupancyOnly   -­‐XX:PretenureSizeThreshold=128m   -­‐XX:+CMSScavengeBeforeRemark   -­‐XX:ParallelGCThreads=9   -­‐XX:ConcGCThreads=9   -­‐XX:+UseParNewGC   -­‐XX:+UseConcMarkSweepGC   -­‐XX:MaxTenuringThreshold=12   -­‐XX:TargetSurvivorRatio=90   -­‐XX:SurvivorRatio=6   -­‐XX:NewRatio=5   -­‐Xmx32g   -­‐Xms16g   -­‐Xss256k  
  • 21. Thank you. Download Fusion: http://lucidworks.com/fusion/ Webinar recording will be available soon. Questions?