SpringOne2GX-2014-Presentation-Notes.rtf

{\rtf1\ansi\ansicpg1252\cocoartf1138\cocoasubrtf510
{\fonttbl\f0\fswiss\fcharset0 ArialMT;\f1\fswiss\fcharset0 Helvetica;}
{\colortbl;\red255\green255\blue255;\red38\green38\blue38;}
\margl1440\margr1440\vieww20160\viewh16000\viewkind0
\pard\tx720\tx1440\tx2160\tx2880\tx3600\tx4320\tx5040\tx5760\tx6480\tx7200\tx7920\tx8640\pardirnatural

\f0\b\fs48 \cf0 SpringOne 2GX-2014, Dallas, TX - September 8th - September 12th, 2014
\b0\fs24 \
\

\i\fs40 Effective Application Development with GemFire using Spring Data GemFire
\i0\fs24 \
by John Blum (Luke Shannon)\
\
\

\b\fs32 GemFire Overview
\b0\fs24 \
\
1. 
\b What is GemFire?
\b0 \
\
* What IMDG means to GemFire...\
* Persistence, Local/Global Transactions, WAN Replication\
\
2. 
\b In-Memory
\b0 \
\
* Stop-the-world Garbage Collection, GC pauses\
* JVM Heap/GC tuning depends entirely on the application (use cases) of GemFire and access data patterns, along with the type of JVM (Sun, IBM)\
* Serialized objects graphs (byte arrays, especially with PDX) use less memory\
* IMDG with Off-Heap Support: GridGain\
\
\
3. 
\b In-Memory Continued\'85
\b0 \
\
* 
\i critical-heap-percentage
\i0  - percentage of heap at which the cache is considered in danger of becoming inoperable due to garbage collection pauses or OutOfMemoryErrors.\
* 
\i eviction-heap-percentage
\i0  - percentage of heap at which eviction will begin on Regions configured for HeapLRU eviction\
\
\
4. 
\b Data Grid
\b0 \
\
* 
\i Gfsh
\i0  is similar in function to 
\i mysql
\i0  if using MySQL and 
\i sqlplus
\i0  if using Oracle\
\
\
5. 
\b Shared Nothing Architecture
\b0 \
\
* \cf2 Failures in 
\i cache
\i0 /disks in one member do not affect the ability of another cache instance to operate safely on its disk files\
* Continuous Availability\cf0 \
\
\
6.
\b  High Scalability
\b0 \
\
* GemFire is elastic\'85 Scale-up/Scale-down as load patterns, demand changes\
* 
\b Collocate
\b0  application business logic (Function(s)) with the data; data is not moved around over the network -> 
\i locality of scale
\i0 \
* 
\b Data-aware Function routing
\b0  is more appropriate for applications requiring iteration over the data set (e.g. aggregation, querying)\
* 
\b Write-Behind
\b0  data changes are (redundantly) queued in-order to avoid bottlenecks.\
\
* Scalability is also improved when using asynchronous, \'93write-behind\'94 of data changes to external data stores (e.g. RDBMS)\
\
7. 
\b High-Throughput / Low-Latency
\b0 \
\
* How does GemFire achieve high-throughput, low-latency\'85 concurrent, in-memory data structures (java.util.concurrent)\
\
\

\b\fs32 GemFire Data Storage Model
\b0\fs24 \
\
* So how does GemFire store data...\
* Data Policy and Region Shortcuts\
\
\
8. 
\b REPLICATE vs. PARTITION
\b0 \
\
\pard\tx720\tx1440\tx2160\tx2880\tx3600\tx4320\tx5040\tx5760\tx6480\tx7200\tx7920\tx8640\pardirnatural

\f1 \cf0 * Oracle Coherence just refers to 
\i Regions
\i0  as 
\i caches
\f0\i0 \
* A Region implement java.util.Map\
* Keys are typically scalar values (Integer/Long, String), but can be objects
\f1\fs36 \
\pard\tx720\tx1440\tx2160\tx2880\tx3600\tx4320\tx5040\tx5760\tx6480\tx7200\tx7920\tx8640\pardirnatural

\f0\fs24 \cf0 \
\
9. 
\b Region Configuration
\b0 \
\
* 
\i Time-to-Live (TTL)
\i0  is baed on last time object was either created or updated (put)\
* 
\i Idle Timeout (TTI)
\i0  is based last time object was accessed (get); resets TTL\
* 
\i Conflation
\i0  only \'93latest\'94 queued value is sent for a particular key.\
\
\
X. 
\b Concurrent Updates
\b0 \
\
* 
\i concurrent updates
\i0  is when 2 (or more) members hosting the same Replicate Region update the same key at the same time without a lock\'85\
* due to network latency an 
\i out-of-order update
\i0  occurs\'85\
* where an update from 1 member is distributed at a later time, where those members already applied a more recent update\
\

\b\fs32 GemFire Topology
\b0\fs24 \
\
X. 
\b Client/Server\
\

\b0 * GemFire partitions 
\b Subscription Management
\b0  (
\i Interest Registration
\i0  and 
\i Continuous Queries
\i0 ) across server data stores resulting in improvements in CPU use and bandwidth utilization thereby improving throughput and reduced latency for clients.\
\
* 
\b Single-hop
\b0  clients store metadata about bucket locations for 
\i Partition Region
\i0  data.\
\
\

\b\fs32 Spring Data GemFire Overview
\b0\fs24 \
\
1. Advantages of Spring Data GemFire\
	- uses Spring's powerful, expressive programming model\
	Integrates with:\
	- Spring Framework's Caching Abstraction (Customer Case Study)\
	- Spring XD\
\
2. Baseline\
	- Spring Framework 4.0.7\
	- Spring Data Commons 1.9\
	- GemFire 7.0.2\
	- Built/Runs on JDK 7 & 8\
	- And when using Spring Framework 4.1, GemFire can be used as a JSR-107 caching provider\
\
4. Behavioral Feature Changes\
	- Manual Start of GatewayReceivers\
	- Region Lookup/Create (--ignore-if-exists)\
\
3. Convenience Features\
	- Spring Boot Starter\
	- GemfireTemplate \
	- Exception Translation\
	- Disk Store Disk Directory Location automatic creation\
\
4. Auto Region Lookups\
\
5. Region Templates\
\
6. Use Cases\
	- DEMO configuration (integrate with 
\i Gfsh
\i0 ).\
\
7. Spring Data Commons 
\i @Repository
\i0  Abstraction\
	- Querying (No support for Pagination/Slices although Sorting is supported, No Top/First, No Geo-spatial support)\
	- Persisting to multiple Regions.\
	- DEMO Cache/Global Transactions\
\
8. GemFire Functions\
\
9. Customer Case Study - CIBC\
\
\

\b\fs32 Tips-
\b0\fs24 \
\

\i Region.get(key)
\i0  returns direct reference to the object\'85\
\
1. 
\b Do not directly modify cached values.
\b0  Modifying a value in place bypasses the GemFire distribution framework, including cache listeners and writers, expiration activities, and transaction management, and can produce undesired results.\
\
	-- Instead, make a copy of the data using 
\i --copy-on-read
\i0  or 
\i CopyHelper
\i0 .\
\
2. Make sure Classes used as Keys implement Object methods: 
\i equals(:Object)
\i0  and 
\i hashCode()
\i0 ;\
\
3. GemFire serializes data entry key and values during distribution, transfer between client/server, therefore keys/values must be serializable.\
\
4. Application classes must be each peer member's CLASSPATH using the data in non-serialized form.\
\
5. Close the Region to release the resources held\
\
6. Distributed System and Region configuration setting must match to be compatible systems\
\
7. \
\
\

\b\fs32 General- ?
\b0\fs24 \
\
* GemFire keeps data in the form given (if serialized, then serialized).\
\
* Can set the JVMs -XX:MaxDirectMemorySize, which is used by Socket buffers, byte buffers, etc\
}