This library is an ultra low latency, high throughput, persisted, messaging and event driven in memory database.  The typical latency is as low at 16 nano-seconds and supports throughputs of 5-20 million messages/record updates per second.

It uses almost no heap, trivial GC impact, can be much larger than your physical memory size (only limited by the size of your disk) and can be shared *between processes* with better than 1/10th latency of using Sockets over loopback.
It can change the way you design your system because it allows you to have independent processes which can be running or not at the same time (as no messages are lost)  This is useful for restarting services and testing your services from canned data. e.g. like sub-microsecond durable messaging.
You can attach any number of readers, including tools to see the exact state of the data externally. e.g. I use; od -t cx1 {file}  to see the current state.

===================================
Version History
===================================

Version 0.2 - Add support for a 32-bit unsigned index. IntIndexedChronicle. This is slightly slower on a 64-bit JVM, but more compact. Useful if you don't need more than 4 GB of data.

Version 0.1 - Can read/write all basic data types. 26 M/second (max) multi-threaded.

It uses memory mapped file to store "excerpts" of a "chronicle"  Initially it only supports an indexed array of data.

===================================
Throughput Test - IndexedChronicleThroughputMain
===================================
https://github.com/peter-lawrey/Java-Chronicle/blob/master/testing/src/main/java/vanilla/java/chronicle/impl/IndexedChronicleLatencyMain.java

On a 4.6 GHz, i7-2600, 16 GB of memory, Fast SSD drive. Centos 5.7.

The average RTT latency was 175 ns. The 50/99 / 99.9/99.99%tile latencies were 16/19 / 287/361 - ByteBuffer (tmpfs)
The average RTT latency was 172 ns. The 50/99 / 99.9/99.99%tile latencies were 16/19 / 278/352 - Using Unsafe (tmpfs)

The average RTT latency was 180 ns. The 50/99 / 99.9/99.99%tile latencies were 16/19 / 311/1,911 - ByteBuffer (ext4)
The average RTT latency was 178 ns. The 50/99 / 99.9/99.99%tile latencies were 16/19 / 310/1,909- Using Unsafe (ext4)

https://github.com/peter-lawrey/Java-Chronicle/blob/master/testing/src/main/java/vanilla/java/chronicle/impl/IndexedChronicleThroughputMain.java

On a 4.6 GHz, i7-2600, 16 GB of memory, Fast SSD drive. Centos 5.7.

 Took 12.416 seconds to write/read 200,000,000 entries, rate was 16.1 M entries/sec - ByteBuffer (tmpfs)
 Took 9.185 seconds to write/read 200,000,000 entries, rate was 21.8 M entries/sec - Using Unsafe (tmpfs)

 Took 25.693 seconds to write/read 400,000,000 entries, rate was 15.6 M entries/sec - ByteBuffer (ext4)
 Took 19.522 seconds to write/read 400,000,000 entries, rate was 20.5 M entries/sec - Using Unsafe (ext4)

 Took 71.458 seconds to write/read 1,000,000,000 entries, rate was 14.0 M entries/sec - Using Unsafe (ext4)
 Took 141.424 seconds to write/read 2,000,000,000 entries, rate was 14.1 M entries/sec - Using Unsafe (ext4)

 Note: in the last test, it is using 112 GB! of dense virtual memory in Java without showing a dramatic slow down or performance hit.

 The 14.1 M entries/sec is close to the maximum write speed of the SSD as each entry is an average of 28 bytes (with the index) => ~ 400 MB/s

 ===================================
 More compact Index for less than 4 GB of data
 ===================================
 https://github.com/peter-lawrey/Java-Chronicle/blob/master/testing/src/main/java/vanilla/java/chronicle/impl/IntIndexedChronicleThroughputMain.java

on a 4.6 GHz, i7-2600
Took 6.325 seconds to write/read 100,000,000 entries, rate was 15.8 M entries/sec - ByteBuffer (tmpfs)
Took 4.590 seconds to write/read 100,000,000 entries, rate was 21.8 M entries/sec - Using Unsafe (tmpfs)

Took 7.352 seconds to write/read 100,000,000 entries, rate was 13.6 M entries/sec - ByteBuffer (ext4)
Took 5.283 seconds to write/read 100,000,000 entries, rate was 18.9 M entries/sec - Using Unsafe (ext4)