community experience distilled

P U B L I S H I N G

Getting Started with Hazelcast

Mat Johns

Chapter No. 7 "Typical Deployments"

In this package, you will find:

A Biography of the author of the book A preview chapter from the book, Chapter NO.7 "Typical Deployments" A synopsis of the books content Information on where to buy this book

About the Author

Mat Johns is an agile software engineer, hands-on architect, and a general technologist based in London. His experience with the Web reaches all the way back to his misspent youth and some rather hacktastic code, but eventually he grew up to graduate from the University of Southampton with a Masters in Computer Science with Distributed Systems and Networks. He has worked for a number of startups on various web projects and systems since then and nowadays he specializes in designing and creating high performance and scalable web services, currently in the Internet TV world. Away from technology, he is an avid explorer and endeavors to seek out new destinations and adventures as much as possible. He is also a qualified yacht skipper and regularly races in, around, and beyond the Solent. You can follow him on Twitter at @matjohns. I would like to dedicate this book to my incredible parents and grandparents (Trevor, Alison, David, and Sheila) as without your encouragement and support over so many years, this book would not have been possible.

For More Information: www.packtpub.com/getting-started-with-hazelcast/book

Getting Started with Hazelcast

Hazelcast is an innovative new approach to data, in terms of storage, processing, and distribution; it provides an accessible solution to the age-old problem of application and data scalability. Getting Started with Hazelcast introduces this great open source technology in a step-by-step, easy-to-follow manner, from the why to the how to wow!

What This Book Covers

Chapter 1, What is Hazelcast?, helps us to get introduced with the technology, its place in an application's stack, and how it has evolved from traditional approaches to data. Chapter 2, Getting Off the Ground, helps us start coding and get acquainted with the standard distributed data store collections on offer. Chapter 3, Going Concurrent, helps us expand to look at more distributed and concurrent capabilities we can bring into our applications. Chapter 4, Divide and Conquer, helps us look at how data is split up and split across many nodes to provide both performance and resilience. Chapter 5, Listen Out, helps us discover that we can register to receive notifications from the cluster to enable our application to be aware of the goings on. Chapter 6, Spreading the Load, helps us move beyond the data storage, and we investigate the distributed execution service and how Hazelcast is more than just a database. Chapter 7, Typical Deployments, helps us explore the various ways we can use or install Hazelcast into our application or infrastructure, looking at the architectural decisions, reasons, and trade-offs behind each one. Chapter 8, From the Outside Looking In, helps us look at popular alternative access we have to our data rather than using the provided drivers for integrating with a Hazelcast cluster. Chapter 9, Going Global, helps us explode onto the world stage using the public cloud infrastructure and asynchronous remote replication to take our data all around the globe. Chapter 10, Playing Well with Others, helps us bring the technology together with popular companion frameworks to see how we might start to bring the technology to work with legacy applications. Appendix, Configuration Summary, helps us overview of the configurations we have used throughout the book.

For More Information: www.packtpub.com/getting-started-with-hazelcast/book

Typical Deployments
So far we have been looking at Hazelcast in one particular type of deployment, however, there are a number of congurations we could use depending on our particular architecture and application needs. Each deployment strategy tends to be best suited to certain types of conguration or application deployment; so in this chapter we will look at: The issues of co-locating data too close to the application Thin client connectivity, where it's best used and the issues that come with it Lite member node (nee super client) as a middle ground option Overview of the architectural choices

All heap and nowhere to go

One thing we may have noticed with all the examples we have been working on so far is that as we are running Hazelcast in an embedded mode, each of the JVM instances will provide both the application's functionality and also house the data storage. Hence the persisted cluster data is held within the heap of the various nodes, but this does mean that we will need to control the provisioned heap sizes more accurately as it is now more than just a non-functional advantage to have more; size matters. However, depending on the type of application we are developing, it may not be convenient or suitable to directly use the application's heap on the running instance for storing in the data. A pertinent example of this situation would be a web application, especially one that runs in a potentially shared web application container (for example, Apache Tomcat). This would be rather unsuitable for storing extensive amounts of data within the heap as our application's storage requirements drastically increase, we will either need to provision more web application containers, or potentially put ours and other applications running with that container cluster at risk from excessive garbage collection, or worse still running out of heap altogether.

For More Information: www.packtpub.com/getting-started-with-hazelcast/book

Typical Deployments

Stepping back from the cluster

To avoid this situation we can separate our application away from the data cluster through the use of a thin client driver that looks and appears very similar to a direct Hazelcast instance; however, in this case, the operations performed are delegated out to a wider cluster of real instances. This has the benet of separating our application away from the scaling of the Hazelcast cluster, allowing us to scale up our own application without having to scale everything together, maximizing the utilization efciency of the resources we are running on. However, we can still scale up our data cluster by adding more nodes which will lead to a bottleneck, either for memory storage requirements or performance and compute necessities. If we create a "server side" vanilla instance to provide us with a cluster of nodes we can connect out to from a client.
public class VanillaInstanceExample { public static void main(String[] args) { Config conf = new Config(); HazelcastInstance hz = Hazelcast.newHazelcastInstance(conf); } }

If we run this a few times to establish a cluster of a number of instances:

Members [3] { Member [127.0.0.1]:5701 Member [127.0.0.1]:5702 Member [127.0.0.1]:5703 this }

Now we need to bring in a new dependency, with our original downloaded archive is the hazelcast-client-2.6.jar, and we can use this to create ClientExample to connect to the cluster and perform operations against the data that is held there. As the client is delegating the operations out to the wider cluster, the data persisted will out-live the client.
public class ClientExample { public static void main(String[] args) { ClientConfig conf = new ClientConfig(); conf.addAddress("127.0.0.1:5701"); HazelcastClient hzc = HazelcastClient.newHazelcastClient(conf); IMap<String, String> capitals = hzc.getMap("capitals");

[ 78 ]

For More Information: www.packtpub.com/getting-started-with-hazelcast/book

Chapter 7 if (capitals.isEmpty()) { System.err.println("Empty capitals map, adding entries"); capitals.put("GB", capitals.put("FR", capitals.put("US", capitals.put("AU", } System.err.println( "Known capital cities: " + capitals.size()); System.err.println( "Capital city of GB: " + capitals.get("GB")); hzc.shutdown(); } } "London"); "Paris"); "Washington DC"); "Canberra");

In running our client multiple times we can see that the rst run will initialize the capitals map with our starting set of data, before shutting down the client which will allow the JVM instance to complete and exit cleanly. However, when we run the client again, the data has been successfully persisted by the still running cluster so that we won't repopulate it a second time. Our client is currently connecting through to one of the clients specically, however it learns about the existence of the other nodes once it is running. So should our supporting member node die, the client will simply connect over to another one of the other nodes and continue on as normal. The only critical phase is the initial connection, that unlike the member nodes we don't have an auto-discovery mechanism in place; so that needs to be congured explicitly. If the node we have listed is down at the time of our client starting up, we will fail to connect to the cluster irrespective of the state of other nodes or the cluster as a whole. We can address this by listing a number of seed nodes within our client's conguration, as long as one of these nodes is available we can connect to the cluster and go from there.
ClientConfig conf = new ClientConfig(); conf.addAddress("127.0.0.1:5701"); conf.addAddress("127.0.0.1:5702", "127.0.0.1:5703");

[ 79 ]

For More Information: www.packtpub.com/getting-started-with-hazelcast/book

Typical Deployments

By default, the ordering of the nodes we attempt to connect to is consistent depending on the conguration, should the rst nodes we try to connect to be down or having networking issues, we might have to wait until the congured connection time-out to be reached before moving on to the next to try block. To prevent a consistent issue proving to be an ongoing issue for clients starting up, we can set the client to randomly order the target nodes list from its conguration. In this way we would get a faster connection time, at least for a proportion of the time, this may be preferable to a possible consistent issue.
conf.setShuffle(true);

Serialization and classes

One issue we do introduce when using the thin client driver( )is that while our cluster can hold, persist, and serve classes it doesn't have to and might not actually hold the POJO class itself; rather a serialization of the object. This means that as long as each of our clients holds the appropriate class in its classpath we can successfully serialize (for persistence) and de-serialize (for retrieval), but our cluster nodes can't. You can most notably see this if we try to retrieve entries via the TestApp console for custom objects, this will produce ClassNotFoundException. The process used to serialize objects to the cluster starts by checking whether the object is a well-known primitive-like class (String, Long, Integer, byte[], ByteBuffer, Date); if so, these are serialized directly. If not, Hazelcast next checks to see if the object implements com.hazelcast.nio.DataSerializable and if so uses the appropriate methods provided to marshal the object. Otherwise it falls back to standard Java serialization. However in the case of using the distributed executor, as the execution will actually be performed on the cluster nodes themselves, those classes must be present on the classpath of each cluster node.

Lite cluster members

One issue with the client method of connecting to the cluster is that most operations will require multiple hops in order to perform an action. This is as we only maintain a connection to a single node of the cluster and run all our operations through it. With the exception of operations performed on partitions owned by that node, all other activities must be handed off to the node responsible out in the wider cluster, with the single node acting as a proxy for the client.

[ 80 ]

For More Information: www.packtpub.com/getting-started-with-hazelcast/book

Chapter 7

This will add latency to the requests made to the cluster. Should that latency be too high, there is an alternative method of connecting to the cluster known as a lite member (originally known as a super client). This is effectively a non-participant member of the cluster, in that it maintains connections to all the other nodes in the cluster and will directly talk to partition owners, but does not provide any storage or computation to the cluster. This avoids the double hop required by the standard client, but adds the additional complexity and overhead of participating in the cluster. For most uses cases using the standard client is preferable as it is much simpler to congure and use, and can work over higher latency connections; however, should you need higher levels of performance and throughput, you could consider using a lite member. Lite members are set up as you would set up a standard node, hence the additional complexity is involved; however with one small addition in the conguration that ags the node as being non-participant.
Config conf = new Config(); conf.setLiteMember(true);

When a lite member is present in the cluster the other members will be aware of its presence and the fact that it is such a type of node. You will see the appropriate logging in the startup and cluster state logging on the various cluster nodes.
Members [3] { Member [127.0.0.1]:5701 this Member [127.0.0.1]:5702 Member [127.0.0.1]:5703 lite }

Architectural overview
As we have seen there are a number of different types of deployment we could use, which one you choose really depends on our application's make up. Each has a number of trade-offs but most deployments tend to use one of the rst two, with the client and server cluster approach the usual favorite unless we have a mostly compute focused application where the former is a simpler set up. So let's have a look at the various architectural setups we could employ and what situations they are best suited to.

[ 81 ]

For More Information: www.packtpub.com/getting-started-with-hazelcast/book

Typical Deployments

Peer-to-peer cluster
This is the standard example we have been mostly using until now, each node houses both our application itself, and data persistence and processing. It is most useful when we have an application that is primarily focused towards asynchronous or high performance computing, and will be executing lots of tasks on the cluster. The greatest drawback is the inability to scale our application and data capacity separately.

Application Node

Application Node

Application Node

Clients and server cluster

This is a more appropriate setup for the situation where we are mostly storing data in our cluster rather than running tasks. A cluster of server nodes is independently created, scaled, and managed. It is then interacted to via a thin client driver from our application. While this provides good separation between our application and the Hazelcast cluster, it does require more awareness of the classpaths of both our application and the cluster nodes.

Cluster Node

Cluster Node

Cluster Node

Application

Application

Application

[ 82 ]

For More Information: www.packtpub.com/getting-started-with-hazelcast/book

Chapter 7

Hybrid cluster
A middle ground between the two previous strategies, the creation and management of a primary cluster of nodes with a shadow set, holds the application's capabilities but none of the data or computation responsibilities. The only real use case for this strategy is where the client option doesn't provide the required latency, and performance demands from our application due to having to leap frog through other nodes in the cluster to get at our data.
Application Lite Node

Cluster Node

Cluster Node

Cluster Node

Application Lite Node

Summary
We have seen that we have a number of strategies at our deposal for deploying Hazelcast within our architecture. Be it, treating it like a clustered standalone product akin to a traditional data source but with more resilience and scalability. For more complex applications we can directly absorb the capabilities directly into our application, but that does come with some strings attached. But whichever approach we choose for our particular use case, we have easy access to scaling and control at our nger tips. In the next chapter we will look beyond just Hazelcast and the alternative methods of getting access to our held data in the cluster.

[ 83 ]

For More Information: www.packtpub.com/getting-started-with-hazelcast/book

Where to buy this book

You can buy Getting Started with Hazelcast from the Packt Publishing website: http://www.packtpub.com/getting-started-with-hazelcast/book
Free shipping to the US, UK, Europe and selected Asian countries. For more information, please read our shipping policy.

Alternatively, you can buy the book from Amazon, BN.com, Computer Manuals and most internet book retailers.

community experience distilled

P U B L I S H I N G

www.PacktPub.com

For More Information: www.packtpub.com/getting-started-with-hazelcast/book