The class sun.misc.Unsafe allows you to many of the things you shouldn’t be able to do in Java, but are still useful in very specific cases. It is to be avoided 99% of the time, however there are rare occasions where this is the only solution which makes sense.
This post considers how it has been using in OpenHFT and what functionality I would like to see in Java 9. In particular, accessing large amount of memory without impacting the GC can be done this way. Sharing memory between processes, without significant overhead, can only be done this way in Java.
Allocating and freeing off heap memory.
public native long allocateMemory(); public native void freeMemory(long address);
These two method allow you to allocate any size of off heap memory. It is not limited to Integer.MAX_VALUE bytes and you get raw memory where you can apply bounds checking as you need to. e.g. Bytes.writeUTF(String) calculates the length of the encoded string, and checks that the whole string would fit, once, not on each byte.
Java-Lang uses the same internal Cleaner class that DirectByteBuffer uses to ensure the memory is freed. Ideally this wouldn’t be so internal.
Raw access to memory
public native Xxx getXxx(Object, long offset); // intrinsic public native void putXxx(Object, long offset);// intrinsic
In both cases, the Object is null when dealing with off heap memory and the offset is just the address. This allows you to perform RAW memory access using single machine code instructions for the JVMs which threat these as intrinsics. This significantly improves performance for memory access.
The problem with this raw approach is you have to manage the layout of your fields in your data structures yourself. The Java-Lang library addresses this by allowing you to define an interface of getters and setters (even for object types like String and enums) and it will generate the implementation at runtime. i.e. you can access the getter/setters without needing to know the “objects” are off heap.
Thread safe access to memory
public native Xxx getVolatileXxx(Object, long offset); // intrinsic public native void putOrderedXxx(Object, long offset); // intrinsic
These two sets of methods allow you to use a field as volatile with a lazy set. The lazy set is faster for the setting thread but could result in the same thread reading an old value if done too quickly. The solution to this is don’t read a value you just wrote.
These methods are particularly useful when sharing data between processes.
public native boolean compareAndSwapXxxx(Object, long offset, Xxx expected, Xxx setTo) // intrinsic
This method is essential for building locks off heap. In particular it is the most efficient way to share data in a thread safe manner between processes. In tests I have done on a Haswell processor i7-4500, the round trip latency of two processes on the same machine is typically;
|TCP||- 9 micro-seconds.|
|FileLocks||- 5.5 micro-seconds.|
|CAS||- 0.12 micro-seconds.|
|Ordered write||- 0.02 micro-seconds (Half round trip, if this pattern can be used)|
On heap object allocation
public native Object allocateInstance(Class clazz);
When deserializing a class, you want to reconstitute the values in that class the way it was when serialized. This doesn’t work well with the current constructors as noted in JEP 187: Serialization 2.0 A work around for this is to avoid the constructors entirely and create an instance without calling one. This assumes much about trusting the data you have, but it has the advantage of being easy to use and places no assumptions about which constructors you have.
It has often been noted that embedded databases, without a network overhead, can out perform distributed databases in term of latency. I believe the next generation of low latency databases will give the performance of embedded and be shared between processes and give both update and query response times well below a micro-second.
I see no reason these should not be implemented in Java. For Java users, a native interface will perform best as it doesn’t need JNI or translation from a C view of the world to a Java view.
Bulletproof Java Code: A Practical Strategy for Developing Functional, Reliable, and Secure Java Code
Use Java? If you do, you know that Java software can be used to drive application logic of Web services or Web applications. Perhaps you use it for desktop applications? Or, embedded devices? Whatever your use of Java code, functional errors are the enemy!
To combat this enemy, your team might already perform functional testing. Even so, you're taking significant risks if you have not yet implemented a comprehensive team-wide quality management strategy. Such a strategy alleviates reliability, security, and performance problems to ensure that your code is free of functionality errors.Read this article to learn about this simple four-step strategy that is proven to make Java code more reliable, more secure, and easier to maintain.