Post on 15-Apr-2017
transcript
Hazelcast Deep Dive
Emrah Kocaman
Software Engineer @ Hazelcastemrahkocaman
emrah@hazelcast.com
Hazelcast ArchitectureHazelcast Configuration OptionsHazelcast IMap Configuration DetailsHazelcast ClientsHazelcast SerializationQ/A Session
Agenda
Hazelcast Architecture
Map Set Queue Lock/Sem. Atomics Topic User DefinedMultiMap Ring
Buffer
Web Sessions
Java
C++ .NET
Portable Serialization / Pluggable Serialization
Memcached REST Internal ClientNetwork Protocol
java.util.concurrent.*javax.cache.*Hibernate 2nd Level Cache
Executor ServiceSQL Query Map / Reduce Aggregation
Low-level Services API
Node Engine(Threads, Instances, Eventing, Wait/Notify, Invocation)
Partition Management(Master Partition, Data Affinity, Replicas, Migrations, Partition Groups)
Cluster Management(Multicast, IP List, AWS/OpenStack)
Networking(IPv4, IPv6)
On-heap Storage
Managem
ent Center
(AP
I, JMX
)
Security(C
onnection, Encryption, A
uthentication, Authorization)
WA
N(Topology A
ware P
artition Managem
ent, WA
N R
eplication)
Hazelcast Open Source Hazelcast Enterprise
Predicate Entry Processor
High-Density (HD) Memory Store
Continuous Query
HD Near Cache
Open Client Network Protocol
Hot Restart Store*
* Coming in 3.6
Hazelcast Configuration Options
XML ConfigurationProgrammatic configurationSpring Configuration
XML Configuration
XML Configuration
XML Configuration
XML Configuration`hazelcast.config` system property Working directory Classpath Default from Hazelcast.jar
XML Configuration ClasspathXmlConfig FileSystemXmlConfig InMemoryXmlConfig UrlXmlConfig
Programmatic Configuration
Distributed Map Configuration
Distributed Map Configuration
IT’S DEMO TIME
Hazelcast Clients Same intuitive API Similar configuration approach SSL Support Hazelcast Client ProtocolJava, C/C++, .NET, REST, memcache
IT’S DEMO TIME
Hazelcast Serialization
Map, Cache, Queue, Set, List Executor Service Entry Processors Lock Topic
Where it is used?serialization + deserialization
serialization
serialization
serialization
serialization
deserialization
Optimised TypesByteBooleanCharacterShortIntegerLongFloatDouble
byte[]char[]short[]int[]long[]float[]double[]
StringDateBigIntegerBigDecimalClassEnum
Hazelcast Serialization
Serializable DataSerializable IdentifiedDataSerializable Portable Pluggable
Shopping Cart Item
Shopping Cart
Benchmark – 100,000 times maxOrders = 100 * 1000
maxCartItem= 5
java.io.SerializablePros
StandardDoesn’t require any implementation
ConsTakes more time and cpuOccupies more space
java.io.Serializable - Results
Read Performance31 ops in ms
Write Performance46 ops in ms
Binary object size525 bytes
java.io.ExternalizablePros
Standard Efficient than Serializable in terms of CPU and
Memory
ConsRequires to implement the actual serialization
ShoppingCartItem - implementation
ShoppingCart- implementation
java.io.Externalizable - Results
Read Performance61 ops in ms
Write Performance60 ops in ms
Binary object size235 bytes
DataSerializablePros
Efficient than Serializable in terms of CPU and Memory
ConsHazelcast specificRequires to implement the actual serializationUses Reflection while de-serializing
ShoppingCartItem - implementation
ShoppingCart- implementation
DataSerializable- ResultsRead Performance
64 ops in ms
Write Performance59 ops in ms
Binary object size231 bytes
IdentifiedDataSerializablePros
Efficient than Serializable in terms of CPU and Memory
Doesn’t use Reflection while de-serializing
ConsHazelcast specificRequires to implement the actual serializationRequires to implement a Factory and
configuration
ShoppingCartItem - implementation
ShoppingCart- implementation
IdentifiedDataSerializable – Factory Implementation
IdentifiedDataSerializable- Results
Read Performance68 ops in ms
Write Performance60 ops in ms
Binary object size186 bytes
PortablePros
Efficient than Serializable in terms of CPU and MemoryDoesn’t use Reflection while de-serializingSupports versioning Supports partial de-serialization during Queries
ConsHazelcast specificRequires to implement the actual serializationRequires to implement a Factory and Class DefinitionClass definition is also sent together with Data but
stored only once per class
ShoppingCartItem - implementation
ShoppingCart- implementation
Portable – Factory Implementation
Portable- ResultsRead Performance
65 ops in ms
Write Performance54 ops in ms
Binary object size386 bytes
Pluggable – (ex: Kryo)
ProsDoesn’t require class to implement an interfaceVery convenient and flexible Can be stream based or byte array based
ConsRequires to implement the actual serializationRequires to plug and configure
Stream and ByteArray Serializers
ShoppingCartItem - implementation
ShoppingCart- implementation
ShoppingCart Kryo StreamSerializer
Pluggable Serialization Configuration
Kryo- ResultsRead Performance
60 ops in ms
Write Performance51 ops in ms
Binary object size210 bytes
CompressionCompresses the data.
Can be applied to Serializable and Externalizable only.
Very slow (~1000 times) and CPU consuming.
Can reduce 525 bytes to 26 bytes.
SummarySerializable
R:31 ops/ms, W: 46 ops/ms, Size: 525 bytesExternalizable
R:61 ops/ms, W: 60 ops/ms, Size: 235 bytesDataSerializable
R:64 ops/ms, W: 59 ops/ms, Size: 231 bytesIdentifiedDataSerializable
R:68 ops/ms, W: 60 ops/ms, Size: 186 bytesPortable
R:65 ops/ms, W: 54 ops/ms, Size: 386 bytesKryo
R:60 ops/ms, W: 51 ops/ms, Size: 210 bytes
http://www.zenika.com/formation-hazelcast-essentials.html
30Th November - Free Training