Skip to content

Latest commit

 

History

History
569 lines (432 loc) · 22.7 KB

README.md

File metadata and controls

569 lines (432 loc) · 22.7 KB

We can help you get Chronicle up and running in your organisation, we suggest you invite us in for consultancy, charged on an ad-hoc basis, we can discuss the best options tailored to your individual requirements. - Contact Us

Or you may already be using Chronicle and just want some help - find out more..

Chronicle Queue

Inter Process Communication ( IPC ) with sub millisecond latency and able to store every message: Chronicle

Releases are available on maven central as

<dependency>
 <groupId>net.openhft</groupId>
 <artifactId>chronicle</artifactId>
 <version><!--replace with the latest version, see below--></version>
</dependency>

Click here to get the Latest Version Number

Snapshots are available on OSS sonatype

Contents

Overview

Chronicle is a Java project focused on building a persisted low latency messaging framework for high performance and critical applications.

At first glance Chronicle Queue can be seen as yet another queue implementation but it has major design choices that should be emphasised.

Using non-heap storage options (RandomAccessFile) Chronicle provides a processing environment where applications do not suffer from Garbage Collection. While implementing high performance and memory-intensive applications (you heard the fancy term "bigdata"?) in Java; one of the biggest problems is Garbage Collection. Garbage Collection (GC) may slow down your critical operations non-deterministically at any time. In order to avoid non-determinism and escape from GC delays off-heap memory solutions are ideal. The main idea is to manage your memory manually so it does not suffer from GC. Chronicle behaves like a management interface over off-heap memory so you can build your own solutions over it. Chronicle uses RandomAccessFiles while managing memory and this choice brings lots of possibilities. RandomAccessFiles permit non-sequential, or random, access to a file's contents. To access a file randomly, you open the file, seek a particular location, and read from or write to that file. RandomAccessFiles can be seen as "large" C-type byte arrays that you can access at any random index "directly" using pointers. File portions can be used as ByteBuffers if the portion is mapped into memory.

This memory mapped file is also used for exceptionally fast interprocess communication (IPC) without affecting your system performance. There is no Garbage Collection (GC) as everything is done off heap.

Building Blocks

Chronicle is the main interface for management and can be seen as the Collection class of Chronicle environment. You will reserve a portion of memory and then put/fetch/update records using the Chronicle interface.

Chronicle has three main concepts:

  • Tailer (sequential reads)
  • Excerpt (random reads)
  • Appender (sequential writes).

An Excerpt is the main data container in a Chronicle, each Chronicle is composed of Excerpts. Putting data to a chronicle means starting a new Excerpt, writing data into it and finishing the Excerpt at the end. A Tailer is an Excerpt optimized for sequential reads. An Appender is something like Iterator in Chronicle environment. You add data appending the current chronicle.

Chronicle Queue V3

Current version of Chronicle-Queue (V3) contains IndexedChronicle and VanillaChronicle implementations.

IndexedChronicle

IndexedChronicle is a single writer multiple reader Chronicle.

For each record, IndexedChronicle holds the memory-offset in another index cache for random access. This means IndexedChronicle "knows" where the 3rd object resides in memory. This is why it is named as "Indexed". But this index is just a sequential index, the first object has index 0, the second object has index 1... Indices are not strictly sequential so if there is not enough space in the current chunk, a new chunk is created and the left over space is a padding record with its own index which the Tailer skips.

base-directory /
    name.index
    name.data

VanillaChronicle

Vanilla Chronicle is a designed for more features rather than just speed. It supports:

  • rolling files on a daily, weekly or hourly basis.
  • concurrent writers on the same machine.
  • concurrent readers on the same machine or across multiple machines via TCP replication.
  • zero copy serialization and deserialization.
  • millions of writes/reads per second on commodity hardware.
    (~5 M messages / second for 96 byte messages on a i7-4500 laptop)
  • synchronous persistence as required. (commit to disk before continuing)
  • exact length of entries

The directory structure is as follows.

base-directory /
   {cycle-name} /       - The default format is yyyyMMdd
        index-{n}       - multiple index files from 0 .. {n}
        data-{tid}-{m}  - multiple data files for each thread id (matches the process id) from 0 .. {n}

The index file format is an sequence of 8-byte values which consist of a 16-bit {tid} and the offset in bytes of the start of the record. The data file format has a 4-byte length of record. The length is the inverted bits of the 4-byte value. This is used to avoid seeing regular data as a length and detect corruption. The length always starts of a 4-byte boundary.

Getting Started

Chronicle Construction

Creating an instance of Chronicle is a little more complex than just calling a constructor. To create an instance you have to use the ChronicleQueueBuilder.

String basePath = System.getProperty("java.io.tmpdir") + "/getting-started"
Chronicle chronicle = ChronicleQueueBuilder.indexed(basePath).build();

In this example we have created an IndexedChronicle which creates two RandomAccessFiles one for indexes and one for data having names relatively:

${java.io.tmpdir}/getting-started.index ${java.io.tmpdir}/getting-started.data

Writing

// Obtain an ExcerptAppender
ExcerptAppender appender = chronicle.createAppender();

// Configure the appender to write up to 100 bytes
appender.startExcerpt(100);

// Copy the content of the Object as binary
appender.writeObject("TestMessage");

// Commit
appender.finish();

Reading

// Obtain an ExcerptTailer
ExcerptTailer reader = chronicle.createTailer();

// While until there is a new Excerpt to read
while(!reader.nextIndex());

// Read the objecy
Object ret = reader.readObject();

// Make the reader ready for next read
reader.finish();

Cleanup

Chronicle-Queue stores its data off heap and it is recommended that you call close() once you have finished working with Excerpts and Chronicle-Queue.

appender.close();
reader.close();
chronicle.close();

Replication

Chronicle-Queue supports TCP replication with optional filtering so only the required record or even fields are transmitted. This improves performances and reduce bandwitdh requirements.

Source

A Chronicle-Queue Source is the master source of data

String basePath = System.getProperty("java.io.tmpdir") + "/getting-started-source"

// Create a new Chronicle-Queue source
Chronicle source = ChronicleQueueBuilder
    .indexed(basePath + "/new")
    .source()
    .bindAddress("localhost", 1234)
    .build();

// Wrap an existing Chronicle-Queue
Chronicle chronicle = ChronicleQueueBuilder.indexed(basePath + "/wrap")
Chronicle source = ChronicleQueueBuilder
    .source(chronicle)
    .bindAddress("localhost", 1234)
    .build();

Sink

A Chronicle-Queue sink is a Chronicle-Queue client that stores a copy of data locally (replica).

String basePath = System.getProperty("java.io.tmpdir") + "/getting-started-sink"

// Create a new Chronicle-Queue sink
Chronicle sink = ChronicleQueueBuilder
    .indexed(basePath + "/statefull")
    .sink()
    .connectAddress("localhost", 1234)
    .build();

// Wrap an existing Chronicle-Queue
Chronicle chronicle = ChronicleQueueBuilder.indexed(basePath + "/statefull")
Chronicle sink = ChronicleQueueBuilder
    .sink(chronicle)
    .connectAddress("localhost", 1234)
    .build();

Remote Tailer

A Remote Tailer is a stateless Sink (it operates in memory)

Chronicle chronicle = ChronicleQueueBuilder
    .remoteTailer()
    .connectAddress("localhost", 1234)
    .build();

Remote Appender

A Remote Appender is a Chronicle-Queue implementation which supports append excerpt to a Chronicle-Source. It is not a full implementation of a Chronicle-Queue as you can only create a single ExcerptAppender.

Chronicle chronicle = ChronicleQueueBuilder
    .remoteAppender()
    .connectAddress("localhost", 1234)
    .build();

ExcerptAppender appender.createAppender();
appender.startExcerpt();
appender.writeLong(100)
appender.finish()

The appender can be configured in a fire and forget way (default) or require an ack from the Chronicle Source

Chronicle chronicle = ChronicleQueueBuilder
    .remoteAppender()
    .connectAddress("localhost", 1234)
    .appendRequireAck(true)
    .build();

Advanced usage

Off-Heap Data Structures

An Exceprt provides all the low-level primitives to read/store data to Chronicle-Queue but it is often convenient and faster to think about interfaces/beans and rely on OpenHFT's code generation.

As example, we want to store some events to Chronicle-Queue so we can write an interface like that:

public static interface Event extends Byteable {
    boolean compareAndSwapOwner(int expected, int value);
    int getOwner();
    void setOwner(int meta);

    void setId(long id);
    long getId();

    void setType(long id);
    long getType();

    void setTimestamp(long timestamp);
    long getTimestamp();
}

Now we have the option to automatically generate a concrete class with:

  • DataValueClasses.newDirectInstance(Event.class) which creates a concrete implementation of the given interface backed by an off-heap buffer
  • DataValueClasses.newDirectReference(Event.class) which creates a concrete implementation of the given interface which needs to be supplied with a buffer to write to

Write with Direct Instance

When we write to an object created with newDirectInstance we write to an off-heap buffer owned by the generated class itself so we do not write directly to an Excerpt. Once done we can write to the Excerpt as usual:

final int items = 100;
final String path = System.getProperty("java.io.tmpdir") + "/direct-instance";
final Event event = DataValueClasses.newDirectInstance(Event.class);

try (Chronicle chronicle = ChronicleQueueBuilder.vanilla(path).build()) {
    ExcerptAppender appender = chronicle.createAppender();
    for(int i=0; i<items; i++) {
        event.setOwner(0);
        event.setType(i / 10);
        event.setTimestamp(System.currentTimeMillis());
        event.setId(i);

        appender.startExcerpt(event.maxSize());
        appender.write(event);
        appender.finish();
    }

    appender.close();
}

Write with Direct Reference

An object created with newDirectReference does not hold any buffer so we need to provide one which can be an Excerpt. By doing so we directly write to the Excerpt's buffer.

final int items = 100;
final String path = System.getProperty("java.io.tmpdir") + "/direct-instance";
final Event event = DataValueClasses.newDirectReference(Event.class);

try (Chronicle chronicle = ChronicleQueueBuilder.vanilla(path).build()) {
    ExcerptAppender appender = chronicle.createAppender();
    for(int i=0; i<items; i++) {
        appender.startExcerpt(event.maxSize());

        event.bytes(appender, 0);
        event.setOwner(0);
        event.setType(i / 10);
        event.setTimestamp(System.currentTimeMillis());
        event.setId(i);

        appender.position(event.maxSize());
        appender.finish();
    }

    appender.close();
}

Read with Direct Reference

Reading data from an Excerpt via a class generated via newDirectReference is very efficient as:

  • it does not involve any copy because the de-serialization is performed lazily and only on the needed fields
  • you do not have to deal with data offsets as the code to do so is generated for you by newDirectReference
try (ExcerptTailer tailer = chronicle.createTailer()) {
     final Event event = DataValueClasses.newDirectReference(Event.class);
     while(tailer.nextIndex()) {
         event.bytes(tailer, 0);
         // Do something with event
         tailer.finish();
     }
 } catch(Exception e) {
     e.printStackTrace();
 }

Ordering fields of DataValueClasses

By default when classes generated via DataValueClasses are serialized, their fields will be ordered by field size (smallest first), however sometimes, especially when you add new fields, you may not want new small fields to be serialized towards the top. If you wish to preserve the existing order of your fields, you can do this by using the @Group annotation on each method. The serialization order of the fields is determined by adding @Group to the set() methods. If you wish you can have a number of different methods with the same value in @Group(). Methods with the same value continue to be ordered by size (smallest first).

See test below

import net.openhft.lang.io.ByteBufferBytes;
import net.openhft.lang.io.IByteBufferBytes;
import net.openhft.lang.model.Byteable;
import net.openhft.lang.model.DataValueClasses;
import net.openhft.lang.model.constraints.Group;
import net.openhft.lang.model.constraints.MaxSize;
import org.junit.Assert;
import org.junit.Test;

import java.nio.ByteBuffer;

/**
 * Rob Austin
 */
public class GroupTest {

    @Test
    public void test() {
        IByteBufferBytes byteBufferBytes = ByteBufferBytes.wrap(ByteBuffer.allocate(1024));

        {
            BaseInterface baseInterface = DataValueClasses.newDirectInstance(BaseInterface.class);

            ((Byteable) baseInterface).bytes(byteBufferBytes, 0);

            baseInterface.setInt(1);
            baseInterface.setStr("Hello World");

            Assert.assertEquals(1, baseInterface.getInt());
            Assert.assertEquals("Hello World", baseInterface.getStr());

        }
        {
            ExtInterface extInterface = DataValueClasses.newDirectReference(ExtInterface.class);
            byteBufferBytes.clear();
            ((Byteable) extInterface).bytes(byteBufferBytes, 0);
            extInterface.setInt2(43);

            Assert.assertEquals(1, extInterface.getInt());
            Assert.assertEquals(43, extInterface.getInt2());
            Assert.assertEquals("Hello World", extInterface.getStr());
            extInterface.setInt(2);

            Assert.assertEquals(2, extInterface.getInt());
        }
    }


    @Test
    public void test2() {
        IByteBufferBytes byteBufferBytes = ByteBufferBytes.wrap(ByteBuffer.allocate(1024));

        {
            ExtInterface extInterface = DataValueClasses.newDirectReference(ExtInterface.class);
            byteBufferBytes.clear();
            ((Byteable) extInterface).bytes(byteBufferBytes, 0);

            extInterface.setInt(1);
            extInterface.setInt2(2);
            extInterface.setStr("Hello World");

            Assert.assertEquals(1, extInterface.getInt());
            Assert.assertEquals(1, extInterface.getInt());
            Assert.assertEquals("Hello World", extInterface.getStr());
        }

        {
            BaseInterface baseInterface = DataValueClasses.newDirectReference(BaseInterface.class);
            byteBufferBytes.clear();
            ((Byteable) baseInterface).bytes(byteBufferBytes, 0);

            Assert.assertEquals(1, baseInterface.getInt());
            Assert.assertEquals("Hello World", baseInterface.getStr());
        }
    }

    public interface BaseInterface {
        String getStr();
        void setStr(@MaxSize(15) String str);
        int getInt();
        void setInt(int i);
    }

    public interface ExtInterface extends BaseInterface {
        int getInt2();

        @Group(1)
        void setInt2(int i);
    }
}

Reading the Chronicle after a shutdown

Let's say my Chronicle Reader Thread dies. When the reader thread is up again, how do we ensure that the reader will read from the point where it left off?

Here is a number of solutions. You can:

  • write the results of the message to an output chronicle with meta data like timings and the source index. If you reread the output to reconstruct your state you can also determine which entry was processed, i.e. you want to replay any entries read but for which there was no output.
  • you can record the index of the last entry processed in a ChronicleMap.
  • you can reread all the entries and check via some other means whether it is a duplicate or not.
  • you can mark entries on the input, either for load balancing or timestamp in when the entry was read. The last entry read can be found by finding the last entry without a time stamp. You can use the binary search facility to do this efficiently.

Example with mark entries as processed and recovery in case of failure:

protected static class Reader implements Runnable  {
    private final Chronicle chronicle;
    private final int id;
    private final int type;

    public Reader(final Chronicle chronicle, int id, int type) {
        this.chronicle = chronicle;
        this.id = id;
        this.type = type;
    }

    @Override
    public void run() {
        try (ExcerptTailer tailer = chronicle.createTailer().toStart()) {
            final Event event = DataValueClasses.newDirectReference(Event.class);
            while(tailer.nextIndex()) {
                event.bytes(tailer, 0);
                if(event.getType() == this.type) {
                    // Check if the Reader was interrputed before completion
                    if(event.getOwner() == this.id * 100) {
                        // Do something with the Event
                    }
                    // Try to acquire the Excerpt and mark the event as being processed by this Reader
                    else if (event.compareAndSwapOwner(0, this.id * 100)) {
                        // Do something with the Event

                        // Mark the event as processed by this Reader
                        event.compareAndSwapOwner(this.id * 100, this.id);
                    }
                }

                tailer.finish();
            }
        } catch(Exception e) {
            e.printStackTrace();
        }
    }
}

Non-blocking Remote Client

On a remote client (Sink or Tailer) nextIndex() waits until some data is received from the Source before returning true or false (in case the client receives a heartbeat, nextIndex returns false) but sometimes you do not want this behavior, i.e. you want to monitor multiple chronicles so you can set the number of times the client checks for data before giving up:

Chronicle[] chronicles = new Chronicle[] {
    ChronicleQueueBuilder.remoteTailer()
        .connectAddress("localhost", 1234)
        .readSpinCount(5)
        .build(),
    ChronicleQueueBuilder.remoteTailer()
        .connectAddress("localhost", 1235)
        .readSpinCount(5)
        .build()
 };

 for(Chronicle chronicle : chronicles) {
     if(chronicle.nextIndex()) {
         // do something
     }
 }

If readSpinCount is set to a value greater than zero, the socket channel is configured to be non-blocking and the read operations spins readSpinCount times before giving up if no data is received.

Data filtering

By default a remote client receives every bit stored on the source but that is something you may not want as a client may be interested in some specific data or even fields

final Chronicle highLowSink = sink(sinkHighLowBasePath)
    .withMapping(new new MappingFunction() {
        @Override
        public void apply(Bytes from, Bytes to) {
            //date
            to.writeLong(from.readLong());

            //open which we not send out
            from.readDouble();

            // high
            to.writeDouble(from.readDouble());

            //low
            to.writeDouble(from.readDouble());
        })
    .connectAddress("localhost", port)
    .build();

Full example

Support