Micro Kernel on Virtual Machines

本研究的目标是在Java虚拟机上开发一个操作系统，使其独立于底层硬件和CPU体系结构。特别是，我们在Java虚拟机上开发了一个微内核。1998年，我们修改了Java虚拟机，使其可以作为独立的应用程序运行，并用修改后的Java虚拟机替换了BIOS，这样就可以独立于底层CPU体系结构开发微内核。1999年，我们为Java微内核设计了一个进程模型，并尝试了一个简单的文件系统来验证我们的微内核机制。我们还在Java虚拟机上实现了一个TCP/IP协议栈，并测量了它的性能。对于TCP/IP协议栈，BSD使用mbuf， X内核使用图结构，而我们开发了一个适合Java的协议栈。Java没有指针，因此无法在数据结构内部指向。此外，Java限制数据类型之间的强制转换。因此，我们设计了一个适合Java的缓冲结构，并使用监视器进行线程通信。由于Java虚拟机不是本机机器，因此它的性能可能存在问题，但从我们的测量来看，它的性能比我们预期的要好得多。

The goal of this research is to develop an operating system on top of Java virtual machines so that it is independent from underlying hardware and CPU architecture. In particular, we have developed a micro kernel on top of Java virtual machines.In 1998, we have modified the Java virtual machine to be run as a stand-alone application and replaced BIOS by this modified Java virtual machine so that micro kernels can be developed independently from underlying CPU architecture.In 1999, we have designed a process model for the Java micro kernel and tried a simple file system for the purpose of verifying our micro kernel mechanisms.We have also implemented a TCP/IP protocol stack on our Java virtual machine and measured its performance. As for TCP/IP protocol stacks, BSD uses mbuf and X kernel uses graph structure, whereas we have developed a protocol stack which is suitable for Java. Java does not have pointers, so there is no way to point inside data structures. In addition, Java restricts casting between data types. Therefore, we have designed a buffer structure which is suitable for Java and used monitors for thread communication. Since the Java virtual machine is not a native machine, its performance may be a problem, but from our measurements its performance is much better than we have expected.