Symmetric multi-processor system
View Patent ↗The present invention relates generally to computer operating systems, and more specifically, to operating system calls in a symmetric multiprocessing (SMP) environment. Existing SMP strategies either use a single lock or multiple locks to limit access to critical areas of the operating system to one thread at a time. These strategies suffer from a number of performance problems including slow execution, large software and execution overheads and deadlocking problems. The invention applies a single lock strategy to a micro kernel operating system design which delegates functionality to external processes. The micro kernel has a single critical area, the micro kernel itself, which executes very quickly, while the external processes are protected by proper thread management. As a result, a single lock may be used, overcoming the performance problems of the existing strategies.
1. A method of symmetric multiprocessing in which one or more processors, a first memory storing a micro kernel operating system in a machine executable form and a second memory storing a thread scheduler in a machine executable form are interconnected via a communication network, wherein the micro kernel operating system comprises a critical area and a non-critical area, said method comprising the steps of:
responding to a first thread requiring a call to the critical area of said micro kernel operating system by requesting a global lock of the critical area of the micro kernel operating system for the first thread;
responding to said global lock being available by performing the steps of:
acquiring said global lock from said thread scheduler for the first thread to access the critical area of the micro kernel operating system;
performing said call to the critical area of said micro kernel operating system, wherein the non-critical area of the micro kernel operating system is available for access by a second thread while the global lock is applied to the critical area of the micro kernel operating system for the first thread; and
releasing said global lock.
2. A method of symmetric multiprocessing as claimed in claim 1 , wherein said micro kernel operating system comprises a pre-emptable micro kernel operating system, said method further comprising the steps with said thread scheduler of:
pre-empting any non-critical threads currently executing on said pre-emptable micro kernel operating system prior to executing said first thread on said pre-emptable micro kernel operating system; and
reinstating said pre-empted threads following said step of executing said first thread on said pre-emptable micro kernel operating system.
3. A method of symmetric multiprocessing as claimed in claim 2 wherein said step of performing said call to said micro kernel operating system comprises the steps of:
entering said pre-emptable micro kernel operating system;
executing operating system functions as required by said first thread; and
exiting said pre-emptable micro kernel operating system.
4. The method of claim 1 , where the act of releasing said global lock comprises releasing the global lock of the critical area before completing an operating system call of the non-critical area for the first thread.
5. The method of claim 1 , where the critical area is limited to operating system message-passing capabilities and the non-critical area comprises operating system functionality delegated to external processes.
6. The method of claim 5 , where the message-passing capabilities comprise inter-process control (IPC) message-passing services.
7. The method of claim 1 , where the global lock comprises an only lock used by the micro kernel operating system to manage threads.
8. The method of claim 1 , further comprising:
executing the first thread on a first processor of the one or more processors; and
executing the second thread on a second processor of the one or more processors.
9. A method of symmetric multiprocessing in which one or more processors, a first memory storing a pre-emptable micro kernel operating system in a machine executable form and a second memory storing a thread scheduler in a machine executable form are interconnected via a communication network, wherein the micro kernel operating system comprises a critical area and a non-critical area, said method comprising the steps of:
responding to a first thread requiring a call to the critical area of said micro kernel operating system by requesting a global lock of the critical area of the micro kernel operating system for the first thread;
pre-empting any non-critical threads currently executing on said pre-emptable micro kernel operating system;
acquiring said global lock from said thread scheduler for the first thread to access the critical area of the micro kernel operating system;
entering said pre-emptable micro kernel operating system;
executing operating system functions of the critical area as required by said first thread, wherein the non-critical area of the micro kernel operating system is available for access by a second thread while the global lock is applied to the critical area of the micro kernel operating system for the first thread;
exiting the critical area of said pre-emptable micro kernel operating system;
releasing said global lock; and
reinstating said pre-empted threads.
10. A computer system comprising:
one or more processors;
a first memory storing a pre-emptable operating system in a machine executable form, wherein the operating system comprises a critical area and a non-critical area;
a second memory storing a lock manager in an machine executable form;
a communication network interconnecting said one or more processors, said first memory and said second memory; and
said lock manager configured to:
respond to a first thread requiring access to the critical area of said pre-emptable operating system by requesting a global lock of the critical area of the operating system for the first thread;
pre-empt any non-critical threads currently executing on said operating system;
acquire said global lock for the first thread to access the critical area of the operating system;
execute the first thread on said operating system, wherein the non-critical area of the operating system is available for access by a second thread while the global lock is applied to the critical area of the operating system for the first thread; and
reinstate said pre-empted threads.
11. The computer system of claim 10 , where the lock manager is further configured to release the global lock of the critical area before completing an operating system call of the non-critical area for the first thread.
12. The computer system of claim 10 , where the critical area is limited to operating system message-passing capabilities and the non-critical area comprises operating system functionality delegated to external processes.
13. The computer system of claim 12 , where the message-passing capabilities comprise inter-process control (IPC) message-passing services.
14. The computer system of claim 10 , where the global lock comprises an only lock used by the operating system to manage threads.
15. The computer system of claim 10 , where the one or more processors comprise a first processor to execute the first thread, and where the one or more processors comprise a second processor to execute the second thread.