Coherence

nlighten
Oracle Coherence cache

for a high performance
eCommerce Application
WHITE PAPER
J u n e 2 0 1 0
Oracle Coherence cache for a high performance eCommerce Application | June 2010
TABLE OF CONTENTS
Abstract 3
Abbreviations 3
Challenges in Ecommerce Application 4
Caching solution – which one to use? 5
Coherence 6
Cache Topologies 7
Coherence Cache Settings 7
Coherence Extend 11
Coherence Configuration 14
Best Practices 18
Common Issues 18
Conclusion 20
Reference 20
© 2010, HCL Technologies. Reproduction Prohibited. This document is protected under Copyright by the Author, all rights reserved.

Abstract
This document discusses how the in-memory Data-Grid based on
Oracle’s Coherence product can be leveraged to implement a very high
performing and highly-scalable e-commerce application where it provides
the basic knowledge for understanding, planning and setting-up coherence
Data-Grid(s) and then also touches-upon a few critical tasks involved in the
implementation such as setting the appropriate cache size and configuring the
relevant cache parameters. Finally, the document covers typical issues which
can be encountered during implementation and the suggested resolutions
for them.
Abbreviations
This section lists abbreviations required to accurately interpret the details in
this document.
• ORM – Object Relational Mapping
• JDBC – Java Database Connectivity
• JVM – Java Virtual Machine
• GC – Garbage Collection
• JMX – Java Management Extentions

Challenges in E-commerce Application

A typical e-commerce application will have a very high performance &
scalability requirement, such as page response time of < 3 seconds, 120K
customer sessions per hour and 100K Orders per day.
Basically, ORM technologies will be used in the e-commerce application.

Result of the ORM technologies are multiple select queries used compared
to pure JDBC, high concurrent load, frequent database access for mostly read
only data, high session memory resulting in garbage collection frequency
and non-linear growth of mostly read only data over period of time. These
all are the factors will lead to performance impact in the application. This is
the major problem domain for the e-commerce application.
Mostly, read only data plays a major role in the e-commerce application
performance. Number of JVM per process and Caching read only data will
reduce the database access bottlenecks and performance impact. By default
ORM supports orthogonal caching. Each caching technique has pros and
cons. Example of OpenJPA, where extremely high frequency of data reads
causes the thread contention and also a different object instance even for the
same entity. Application caching is explicitly having logic in the application
layer for retrieving or populating cache. Content cache is the example for
application caching.
Caching more data in the application memory will impact more GC and
will directly lead to performance degradation. Distributed caching is the best
option for a large amount of cached data. Avoiding unnecessary caching,
optimizing the heap memory size and finding suitable garbage collector will
reduce the frequent GC.
Size of read only data will differ based on the number of countries,
language, vendor and sites used in the application. The example below
explains the predicted growth of the product content data in the ecommerce
application.
Site, Country, Lang, Vendor 3744
Site, Country & Lang 2496
Site & Lang 322

y = 28.179e0.9907x
Site Level (One Lang) 161
R² = 0.9319
Lang Level 104
Example of predicted growth of mostly read only data of e-commerce application

Caching solution – which one to use?

Lots of caching solutions are available for java web application. Based on the
performance & scalability requirement, suitable caching solution should be
implemented in the web application. Different caching solution as follows:
a) Java Caching System

JCS is an open source distributed caching system written in java for web
applications. It is designed to speed up dynamic web applications by providing
a means to manage cached data of various dynamic natures. Like any caching
system, the JCS is most useful for high read, low put applications.
b) WhirlyCache
Whirlycache is a fast, configurable in-memory object cache for Java. Whirly
cache is open source software. It can be used, to increase the performance of
a website or an application by caching objects.
c) Jofti
Jofti is a simple to use, faster object indexing and searching solution for
Objects in a Caching layer or storage structure that supports the Map
interface.
d) EHCache
A replacement for JCS, EHCache started out as some patches for JCS to
correct threading and memory leak problems. EHCache is faster than JCS
and acts as a pluggable cache for Hibernate 2.1.
e) JBossCache
JBoss Cache is a product designed to cache frequently accessed Java objects
in order to dramatically improve the performance of e-business applications.
By avoiding unnecessary database access, JBoss Cache decreases network
traffic and increases the scalability of applications.
f) OSCache
OSCache is a widely used, high performance J2EE caching solution. Its
features include fast in-memory caching, persistent caching, clustering
support, flexible caching System, comprehensive API, exception handling,
cache flushing, portable caching and i18n aware.
g) OpenJPA
ORM products also have in-built caching solutions. OpenJPA and hibernates
are the best example of this. Apache OpenJPA is a Java persistence framework
that can be used as a stand-alone POJO persistence layer or integrated into
any Java EE compliant container and many other lightweight frameworks.

h) Hibernate
Hibernate is a powerful, ultra-high performance object/relational persistence

and query service for Java. Hibernate lets the developer develop persistent
objects following common Java idiom - including association, inheritance,
polymorphism, composition and the Java collections framework. Hibernate
is now the most popular ORM solution for Java.
Coherence
What is Coherence?
Coherence provides data performance, reliability and scalability in

distributed or replicated manner for caching services. Coherence is unique
in utilizing a caching grid that is highly available and extensible, providing
the ability to linearly add capacity as demand increases. It also runs outside
of any applications in its own JVM which negates the need to tie it to any
particular application.
Coherence supports unicast and multicast UDP clustering protocols. The

cluster can add dynamically to the already existing coherence grid servers.
If any server is disconnected from the network, it will automatically
redistribute the data from other clustered nodes. Coherence provides
continuous availability for application data and processing and provides
standard JMX APIs.
Why Coherence?
For a typical e-commerce application with high performance & high

scalability requirements, the following are the features which should be
considered in a caching solution:
• A true In-Memory Distributed Data-Grid (not just the notion of
distributed cache which is either based on replication or invalidation in
distributed/clustered environments)
• Provides Number of Topologies (Distributed, Replicated, Distributed-
Near & Local)
• Write-Behind Caches (this is feature will be used for Synchronizing caches
b/w Datacenters)
• Dynamic Discovery and Automatic-Failover of Clustered Node in the
Data-Grid
• JTA Transactional support where the Cache acts as Resource
• Reliable Peer-to-peer communication using a proprietary protocol over
UDP and proprietary Object serialization, making the communication
extremely faster

• Comprehensive Monitoring & Management

• Cache partitioning
Although nowadays most of the open-source and commercial caching

solutions provide the above set of features which makes it really difficult to
choose one over the other, it seems Oracle Coherence is the most matured
and reliable and has been considered as the caching/data-grid solution in rest
of the document.
Cache Topologies
Coherence supports three types of caches. These are Distributed, Near and
Replicated cache. Distributed cache is each node in the server containing a
unique set of application data in the cache. To scale the capacity of cache,
increase the nodes in the cluster. Any type of cache will involve serialization
/de-serialization and network transfers for application data read and write
access in the cache. Distributed cache is best when the application requires
heavy volume of read and write application data.
Near cache is each client node containing small amount of data in the local
cache and larger amount of data in the distributed cache and these caches
are synchronized with each other. There is some overhead involved with
synchronizing the caches.
Replicated cache is each node in the cluster will contains all the application
data in the cache. Replicated cache is best when application requires less
application data and highly read access from cache.
Coherence Cache Settings

Deciding the number of nodes, number of JVMs per node, size of the heap
in JVM, type of serialization, sizing the cache, cache invalidation strategy
and cache eviction policy.
Sizing
When trying to determine how many systems and how many JVMs per
system are required, the coherence cache is a good starting point to determine
how much data need to cache in the application. Once this is calculated all
other factors can be decided later.
Initially find out the cached object size by perform the below stapes:
• Enable the unit calculator to BINARY in the local scheme cache

configuration file
– <unit-calculator>BINARY</unit-calculator>

• Add JMX monitoring settings to the coherence grid server JVMs

– Dcom.sun.management.jmxremote
– Dtangosol.coherence.management.remote=true
– Dtangosol.coherence.management=all
– Dcom.sun.management.jmxremote.port=<JMX port>
– Dcom.sun.management.jmxremote.authenticate=false
– Dcom.sun.management.jmxremote.ssl=false
• Once the application is started, connect the coherence grid server JVM
using JConsole and the mbean tab find out grid server cached objects
size and count. Single cached object size can be calculated using this
information.
Based on the application requirement, calculate the number of objects to

be stored in the coherence grid server cache. Number of objects and size of
the single cached object decide the heap size of JVM and no of grid server
nodes.
To maintain a highly available cluster, sufficient cache servers should be

deployed in the clusters. Generally provide one additional grid server, so
that the application works well in grid server failure scenario also.
Serialization
Network is the important factor in the coherence, since over the wire
lots of data will be transferred and all the clustered instances will be
intercommunicated. Serializing the application objects in the cache will
give better performance. POF serializer is the best one, as it consumes less
serialization /deserialization time. Application objects should implement the
PortableObject interface and the PortableObjectSerializer implementation
is used as the PofSerializer.
Cache Invalidation Strategy
Invalidating objects in the cache will sync the local cache with grid server
cache. Coherence has pre-defined events for the invalidation strategy. Based
on the invalidation strategy local cache will be updated from back cache.
Invalidation strategies are Listen None, Listen Present, Listen All and
Listen Auto.
• Listen none strategy is not to listen for invalidation objects. Based on the
eviction policy data freshness will be maintained in the cache.
• Listen present strategy is local cache will listen to the back cache only
objects in the local cache.
• Listen all strategy will listen to the back cache for the all the objects.
• Listen auto strategy is local cache automatically use listen present and listen
all based on the cache statistics. This is the default invalidation strategy.

Cache Eviction Policy
Mainly, eviction is used to free the memory in the cache and remove the unused
objects in the cache. Eviction policies are LRU, LFU and HYBRID.
• LRU is where cache will keep the objects those are all least recently
used only.
• LFU is the least frequently used eviction policy that chooses which entries
to evict based on how often they are being accessed, evicting those that are
accessed least first.
• HYBRID is a cache that will keep the objects those are all how often
and recently accessed, evicting those that are accessed least frequently
and were not accessed for the longest period first. This is the default
eviction policy.
TCMP Protocol
Coherence uses TCMP, a clustered IP-based protocol, for server discovery,

cluster management, service provisioning and data transmission. To ensure
true scalability, the TCMP protocol is completely asynchronous, meaning
that communication is never blocked, even when many threads on a server
are communicating at the same time. Further, the asynchronous nature
also means that the latency of the network does not affect the cluster
throughput, although it will affect the speed of certain operations. TCMP
uses a combination of UDP/IP multicast, UDP/IP unicast and TCP/IP
as follows:
a) Multicast
• Cluster discovery: Is there a cluster already running that a new member

can join?
• Cluster heartbeat: The most senior member in the cluster issues a periodic
heartbeat via multi-cast; the rate is configurable and defaults to once per
second.
• Message delivery: Messages that need to be delivered to multiple cluster

members will often be sent via multicast, instead of unicasting the message
one time to each member.
b) Unicast
• Direct member-to-member (“point-to-point”) communication, including

messages, asynchronous acknowledgments (ACKs), asynchronous
negative acknowledgments (NACKs) and peer-to-peer heartbeats.
• Under some circumstances, a message may be sent via unicast even if the
message is directed to multiple members. This is done to shape traffic flow
and to reduce CPU load in very large clusters.

c) TCP
• An optional TCP/IP ring is used as an additional “death detection”

mechanism, to differentiate between actual node failure and an
unresponsive node, such as when a JVM conducts a full GC.
• TCP/IP is not used as a data transfer mechanism due to the intrinsic

overhead of the protocol and its synchronous nature.
Multicast Vs WKA
Multicast is always best compared to WKA (Well Known Address) settings.

But network infrastructure should also support Multicast, otherwise use
the WKA. This configuration can be defined in the tangosol configuration
xml file.
Node discovery can be done using multicast or well known address. The
choice is made based on the network topology. If coherence clients and
servers are in the same subnet then multicast is the best, since it will use less
configuration. If clients and servers are located in the different subnet i.e.
datacenters or network didn’t support multicast packets then well known
address is used with predefined configuration. Disadvantage of WKA is to
update the configurations and restart all the servers and clients for adding
new coherence server to the cluster.
The element multicast listener is used to define the address and port that
cluster will use for cluster wide and point to multipoint communications. All
the clients and servers should use the same address and port. This multicast
IP address and port will be used for socket listening or publishing. Sample
configuration file for multicast is as follows:
<multicast-listener>
<address system-property=”tangosol.coherence.clusteraddress”>xx.xx.xx.xx</address>
<port system-property=”tangosol.coherence.clusterport”>9000</port>
<time-to-live system-property=”tangosol.coherence.ttl”>1</time-to-live>
<packet-buffer>
<maximum-packets>64</maximum-packets>
</packet-buffer>
<priority>8</priority>
<join-timeout-milliseconds>30000</join-timeout-milliseconds>
<multicast-threshold-percent>25</multicast-threshold-percent>
</multicast-listener>
10
The element socket address is used to define the WKA list IP addresses and
ports. If one or more WKA is specified, for a member to join the cluster it will
either have to be WKA or there will have to be at least one WKA member
running. If this is empty or unspecified then multicast communication will
be used. Sample configuration file for WKA is as follows:
<unicast-listener>
<well-known-addresses>
<socket-address id=”1”>
<address>xx.xx.xx.xx</address>
<port>8088</port>
</socket-address>
</well-known-addresses>
</unicast-listener>
Coherence Extend
Coherence Extend extends the reach of the Coherence TCMP cluster to a
wider ranch of consumers, including desktops, remote servers and machines
located across WAN connections. Typical uses of coherence extend include
providing desktop applications with access to coherence caches and coherence
cluster “bridges” that link together multiple coherence clusters connected
via a high-latency, unreliable WAN.
Coherence extend consists of two basic components. A client and a Coherence

extend clustered service hosted by one or more DefaultCacheServer processes.
The adapter library includes implementations of both the CacheService
and InvocationService interfaces that route all requests to a Coherence
extend clustered service instance running within the Coherence cluster.
The Coherence extend clustered service in turn responds to client requests
by delegating to an actual Coherence clustered service. The client adapter
library and Coherence extend clustered service use a low-level messaging
protocol to communicate with each other.
Transport bindings for these protocols are extend JMS and extend TCP.
Extend JMS use existing JMS infrastructure to connect to the server. Extend
TCP provides high performance and scalable TCP/IP based communication
layer to connect to the server.
11
Example
Two data centers have the same application and data. One data center data
is updated through application then other data center data is replicated by
streaming in the backend database. Same functionality may be required in
the application cache level also. This is achieved by coherence extend feature.
Define one data center is a master server and other data center as a replica
server. Here is the example of <remote-cache-scheme> element for master
server configuration file as follows:
<remote-cache-scheme>
<scheme-name>example-remote</scheme-name>
<service-name>RemoteCache</service-name>
<initiator-config>
<tcp-initiator>
<remote-addresses>
<socket-address>
<address>hostname </address>
<port>9099</port>
</socket-address>
</remote-addresses>
</tcp-initiator>
<outgoing-message-handler>
<request-timeout>10s</request-timeout>
</outgoing-message-handler>
</initiator-config>
</remote-cache-scheme>
12
Example of <remote-cache-scheme> element for replica server configuration

file as follows:
<proxy-scheme>
<service-name>Proxy</service-name>
<thread-count>4</thread-count>
<acceptor-config>
<tcp-acceptor>
<local-address>
<address>hostname</address>
<port>9099</port>
<reusable>true</reusable>
</local-address>
</tcp-acceptor>
</acceptor-config>
<autostart>true</autostart>
</proxy-scheme>
13
Coherence Configuration
Three configuration files are used in the coherence. Files are tangosol
configuration xml, cache configuration xml and pof configuration xml.
Tangosol Config XML
Tangosol configuration xml file will specify the operational and runtime
elements that control clustering, communication and data management
services and override the tangosol configuration xml file in the java argument
by using -Dtangosol.coherence.override. Example xml file as follows:
<?xml version=’1.0’?>
<coherence>
<cluster-config>
<member-identity>
<cluster-name system-property=”tangosol.coherence.cluster”>example-node1</cluster-name>
</member-identity>
<unicast-listener>
<well-known-addresses>
<socket-address id=”1”>
<address>xx.xx.xx.xx</address>
<port>8088</port>
</socket-address>
</well-known-addresses>
</unicast-listener>
<packet-publisher>
<packet-delivery>
<timeout-milliseconds>60000</timeout-milliseconds>
</packet-delivery>
</packet-publisher>
<service-guardian>
<timeout-milliseconds system-property=”tangosol.coherence.guard.timeout”>65000</timeout-milliseconds>
</service-guardian>
</cluster-config>
<logging-config>
<severity-level system-property=”tangosol.coherence.log.level”>5</severity-level>
<character-limit system-property=”tangosol.coherence.log.limit”>1048576</character-limit>
</logging-config>
</coherence>
14
Cache Config XML
Coherence cache attributes and settings are defined in the cache configuration
xml file. Cache attributes are cache type, cache invalidation strategy, cache
policy, type of serialization and objects count of the cache. Specify the cache
configuration xml file in the java argument by using -Dtangosol.coherence.
cacheconfig. Example xml file as follows:
<?xml version=”1.0”?>
<cache-config>
<caching-scheme-mapping>
<cache-mapping>
<cache-name>demoCache</cache-name>
<scheme-name>demo-near-content</scheme-name>
</cache-mapping>
</caching-scheme-mapping>
<caching-schemes>
<distributed-scheme>
<scheme-name>distributed-content</scheme-name>
<service-name>DistributedCache-pof</service-name>
<serializer>
<class-name>com.tangosol.io.pof.ConfigurablePofContext
</class-name>
<init-params>
<init-param>
<param-value>C:/coherence/pof-config.xml</param-value>
<param-type>String</param-type>
</init-param>
</init-params>
</serializer>
<backup-count>1</backup-count>
<thread-count>8</thread-count>
<backing-map-scheme>
<local-scheme>
<scheme-ref>dist-binary-backing-map</scheme-ref>
</local-scheme>
</backing-map-scheme>
</distributed-scheme>
15
<local-scheme>
<scheme-name>dist-binary-front-map</scheme-name>
<eviction-policy>HYBRID</eviction-policy>
<high-units>50000</high-units>
<expiry-delay>1h</expiry-delay>
</local-scheme>
<local-scheme>
<scheme-name>dist-binary-backing-map</scheme-name>
<eviction-policy>HYBRID</eviction-policy>
<high-units>1000000</high-units>
<expiry-delay>1h</expiry-delay>
</local-scheme>
<near-scheme>
<scheme-name>demo-near-content</scheme-name>
<front-scheme>
<local-scheme>
<scheme-ref>dist-binary-front-map</scheme-ref>
</local-scheme>
</front-scheme>
<back-scheme>
<distributed-scheme>
<scheme-ref>distributed-content</scheme-ref>
</distributed-scheme>
</back-scheme>
<invalidation-strategy>all</invalidation-strategy>
</near-scheme>
</caching-schemes>
</cache-config>
16
POF Config XML
This xml file is used to define the objects to serialize and deserialized by
using pof serializer. The name and location of this file can be configured
in the coherence cache configuration xml file or by the system property
tangosol.pof.config. It is recommended that all the nodes within a cluster use
identical POF user type descriptors. Example file as follows:
<?xml version=”1.0”?>
<!DOCTYPE pof-config SYSTEM “pof-config.dtd”>
<pof-config>
<user-type-list>
<include>coherence-pof-config.xml</include>
<user-type>
<type-id>1004</type-id>
<class-name>com.example.test.SampleOneDataImpl</class-name>
</user-type>
<user-type>
<class-name> com.example.test.SampleTwoDataImpl</class-name>
</user-type>
<user-type>
<class-name> com.example.test.SampleThreeDataImpl </class-name>
</user-type>
</user-type-list>
</pof-config>
The application cached objects should implement the com.tangosol.io.pof.

PortableObject interface to achieve POF serialization. The portable object
interface is a simple interface and has two methods. The two methods are
readExternal() and writeExternal(). The below example shows the sample
code for PortableObject interface implementation:
public void readExternal(final PofReader reader) throws IOException {
uidPk = reader.readLong(0);
value = reader.readString(1);
contentType = (SampleContentType) reader.readObject(2);
contentData = (Set<SampleContentData>) reader.readCollection(3, contentData); }
public void writeExternal(final PofWriter writer) throws IOException {
writer.writeLong(0, uidPk);
writer.writeString(1, value);
writer.writeObject(2, contentType);
writer.writeCollection(3, contentData, SampleOneDataImpl.class); }
17
Best Practices
This section describes the best practices to follow while implementing the
coherence cache in the application.
Cache Warm Up
Preloading all the objects in the coherence cache will give better performance
in the application start up using cache warm up utility. Write a small program
that warms application cache server by causing it to fetch the data directly
from the source which in most cases is the database. This warm up program
should be run in the application start up process.
Minimal Data Cache Size in Application Heap

Some applications contain large amount of application data that will cause
performance and memory insufficiency problems. Coherence distributed
cache provides the capability to have small amount of data in the application
heap and remaining data in the coherence grid server cache. Adding more
nodes in the cluster will increase the size of the cache. Usually small amount
of the application data being used in the coherence local cache depends upon
the application requirement.
Number of JVMs per System

Depending on the application data size in the cache and number of processors
per node, the number of JVMs per node is decided. More JVMs will cause
context switching and contention on processors. Less number of JVMs will
lead to the memory insufficiency problem. More than one JVMs per server
will impact the network resources efficiency.
Common Issues
This section describes the common issues encountered when implementing
coherence. This section guides the reader how to address these issues and
solve them.
Unique Cluster Name

All the coherence client and grid server instances should have a unique cluster
name. Based on this name all the client’s applications will be connected to
the coherence grid servers. This name is defined in the tangosol configuration
xml file as follows:
<cluster-name system-property=”tangosol.coherence.cluster”>demo-sample1</cluster-name>
If this cluster name is not properly defined, then some other application
which is also using coherence within the network will be participated in the
coherence cluster. This will lead to weird behavior in the application like
optimistic lock issue from the DAO layer or dead lock threads.
18
Communication Delay Warning
If the application experience frequent multi-second communication pauses

across multiple cluster, nodes may be required to increase switch’s buffer
space. These communication pauses can be identified by a series of Coherence
log messages identifying communication delays with multiple nodes which
are not attributable to local or remote GCs.
Experienced a 8275 ms communication delay (probable remote GC) with

Member(Id=7,Timestamp=2010-04-0312:15:47.511,Address=255.255.255.255:8088,
MachineId=13838); 320 packets rescheduled, PauseRate=0.31, Threshold=512
Some switches such as the Cisco 6500 series support configuration the
amount of buffer space available to each ethernet port or ASIC. In high
load applications it may be necessary to increase the default buffer space. On
Cisco this can be accomplished by executing:
fabric buffer-reserve high
Some other switches contact with oracle coherence.
UnicastUdpSocket Failed Warning
Experiencing packet loss warning message in the coherence may be need to

increase the maximum socket buffer size to get good performance. Warning
message as follows:
2010-01-14 22:35:30.606/2.051 Oracle Coherence GE 3.5/459 <Warning>

(thread=Main Thread, member=n/a): UnicastUdpSocket failed to set receive
buffer size to 1428 packets (2096304 bytes); actual size is 92 packets (135168 bytes).
Consult your OS documentation regarding increasing the maximum socket buffer
size. Proceeding with the actual value may cause sub-optimal performance.
Though it is safe to operate with the smaller buffers it is recommended that

configure OS to allow for larger buffers.
On Linux execute (as root):

sysctl -w net.core.rmem_max=2096304
sysctl -w net.core.wmem_max=2096304
On Solaris execute (as root):

ndd -set /dev/udp udp_max_buf 2096304
On AIX execute (as root):

no -o rfc1323=1
no -o sb_max=4194304
19
12. Conclusion
Coherence has extensive features like dynamic high availability, automatic
load balancing, transactional consistency, capacity on demand, grid wide
JMX for management and monitoring and true linear scalability. It is
important to have a matured and comprehensive caching solution to meet
the non-functional requirements (performance & scalability) of a typical e-
commerce application, and if the budget can afford, Coherence is definitely
a reliable proposition to meet the non-functional objectives.
13. Reference
Website:
http://www.oracle.com/technology/products/coherence/index.html
http://www.manageability.org/blog/stuff/distributed-cache-java
http://java-source.net/open-source/cache-solutions
About the Author

Rajendran Sundaram is a sun certificated 1.4 (SCJP) professional.
He has 5 years and 8 months of experience in J2EE application development,
performance analysis, tuning and fixing complex production issues, and
is working as a performance analyst with HCL Technologies Limited
Chennai.
20
ENGINEERING AND Hello, I’m from HCL’s Engineering and R&D Services. We enable
R&D SERVICES technology led organizations to go to market with innovative products &
solutions. We partner with our customers in building world class products
& creating the associated solution delivery ecosystem to help build market
CUSTOM APPLICATION leadership. Right now, 13000+ of us are developing engineering products,
SERVICES solutions and platforms across Aerospace and Defense, Automotive,
Consumer Electronics, Industrial Manufacturing, Medical Devices,
ENTERPRISE APPLICATION Networking & Telecom, Office Automation, Semiconductor, Servers &
SERVICES Storage for our customers.
For more details contact eootb@hcl.in
ENTERPRISE Follow us on twitter http://twitter.com/hclers and our blog http://ers.
TRANSFORMATION SERVICES hclblogs.com/
Visit our website http://www.hcltech.com/engineering-services/
IT INFRASTRUCTURE
MANAGEMENT
BUSINESS PROCESS
OUTSOURCING
About HCL
About HCL Technologies
HCL Technologies is a leading global IT services company, working with

clients in the areas that impact and redefine the core of their businesses. Since
its inception into the global landscape after its IPO in 1999, HCL focuses
on ‘transformational outsourcing’, underlined by innovation and value
creation, and offers integrated portfolio of services including software-led IT
solutions, remote infrastructure management, engineering and R&D services
and BPO. HCL leverages its extensive global offshore infrastructure and
network of offices in 26 countries to provide holistic, multi-service delivery
in key industry verticals including Financial Services, Manufacturing,
Consumer Services, Public Services and Healthcare. HCL takes pride in its
philosophy of ‘Employee First’ which empowers our 58,129 transformers
to create a real value for the customers. HCL Technologies, along with its
subsidiaries, had consolidated revenues of US$ 2.6 billion (Rs. 12,048 crores),
as on 31st March 2010 (on LTM basis). For more information, please visit
www.hcltech.com
About HCL Enterprise
HCL is a $5 billion leading global Technology and IT Enterprise that comprises

two companies listed in India - HCL Technologies & HCL Infosystems.
Founded in 1976, HCL is one of India’s original IT garage start-ups, a pioneer of
modern computing, and a global transformational enterprise today. Its range
of offerings spans Product Engineering, Custom & Package Applications,
BPO, IT Infrastructure Services, IT Hardware, Systems Integration, and
distribution of ICT products across a wide range of focused industry verticals.
The HCL team comprises over 62,000 professionals of diverse nationalities,
who operate from 26 countries including over 500 points of presence in
India. HCL has global partnerships with several leading Fortune 1000 firms,
including leading IT and Technology firms. For more information,
please visit www.hcl.in

Coherence

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Coherence

Uploaded by

Copyright:

Available Formats

nlighten

Oracle Coherence cache

• ORM – Object Relational Mapping

• JDBC – Java Database Connectivity

• JVM – Java Virtual Machine

• JMX – Java Management Extentions

Challenges in E-commerce Application

Basically, ORM technologies will be used in the e-commerce application.

Site, Country, Lang, Vendor 3744

Site, Country & Lang 2496

Site & Lang 322

Lang Level 104

Example of predicted growth of mostly read only data of e-commerce application

Caching solution – which one to use?

a) Java Caching System

Hibernate is a powerful, ultra-high performance object/relational persistence

Coherence provides data performance, reliability and scalability in

Coherence supports unicast and multicast UDP clustering protocols. The

For a typical e-commerce application with high performance & high

• Comprehensive Monitoring & Management

Although nowadays most of the open-source and commercial caching

Coherence Cache Settings

• Enable the unit calculator to BINARY in the local scheme cache

• Add JMX monitoring settings to the coherence grid server JVMs

Based on the application requirement, calculate the number of objects to

To maintain a highly available cluster, sufficient cache servers should be

Cache Invalidation Strategy

Cache Eviction Policy

Coherence uses TCMP, a clustered IP-based protocol, for server discovery,

• Cluster discovery: Is there a cluster already running that a new member

• Message delivery: Messages that need to be delivered to multiple cluster

• Direct member-to-member (“point-to-point”) communication, including

• An optional TCP/IP ring is used as an additional “death detection”

• TCP/IP is not used as a data transfer mechanism due to the intrinsic

Multicast is always best compared to WKA (Well Known Address) settings.

Coherence extend consists of two basic components. A client and a Coherence

Example of <remote-cache-scheme> element for replica server configuration

Tangosol Config XML

Cache Config XML

POF Config XML

The application cached objects should implement the com.tangosol.io.pof.

contentType = (SampleContentType) reader.readObject(2);

contentData = (Set<SampleContentData>) reader.readCollection(3, contentData); }

public void writeExternal(final PofWriter writer) throws IOException {

writer.writeCollection(3, contentData, SampleOneDataImpl.class); }

Minimal Data Cache Size in Application Heap

Number of JVMs per System

Unique Cluster Name

Communication Delay Warning

If the application experience frequent multi-second communication pauses

Experienced a 8275 ms communication delay (probable remote GC) with

UnicastUdpSocket Failed Warning

Experiencing packet loss warning message in the coherence may be need to

2010-01-14 22:35:30.606/2.051 Oracle Coherence GE 3.5/459 <Warning>

Though it is safe to operate with the smaller buffers it is recommended that

On Linux execute (as root):

On Solaris execute (as root):

On AIX execute (as root):

About the Author

HCL Technologies is a leading global IT services company, working with

About HCL Enterprise

HCL is a $5 billion leading global Technology and IT Enterprise that comprises

You might also like