Thomas E. Anderson, Brian N. Bershad, Edward D. Lazowska, and Henry M. Levy. Scheduler Activations: E#ective Kernel Support for the User-Level Management of Parallelism. ACM Transactions on Computer Systems, 10(1):53--79, February 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....forwards system calls to the underlying operating system. Therefore, both processes block when either does a blocking system call. So long as the operating system exports some mechanism for dealing with blocking system calls i.e. non blocking versions of the same calls, or scheduler activations [ABLL91] the runtime system can execute blocking calls in a non blocking way. If the SEA is being done by the operating system, then the translator can be informed that one kernel thread is blocked, and the translator will continue running only the other thread. 4.5 Implementation status of SSMT Most ....

T. E. Anderson, B. N. Bershad, E. Lazowska, and H. M. Levy. Scheduler activations: Eective kernel support for the user-level management of parallelism. In Proc. Thirteenth ACM Symp. on Operating System Principles, page 95, Paci c Grove, CA, Oct 1991.

....has about the internal state of the OS, the more one can optimize system services and applications. For example, in Scheduler Activations, one key piece of state information passed from the kernel to the user level scheduling library is the number of processors on which the application is running [2]; with this information, a threads library can do a better job of scheduling amongst the currently active threads. Thus, we summarize ways in which one can determine the internal state of the OS when no such interface exists. Acquire algorithmic knowledge. Developers interfacing with the OS often ....

T. Anderson, B. Bershad, E. Lazowska, and H. Levy. Scheduler Activations: E#ective Kernel Support for the User-Level Management of Parallelism. In ACM Transactions on Computer Systems, pages 53--79, February 1992.

....to reduce the number of user kernel interactions required for certain service requests. In addition, solutions have been created for better integration of certain kernel and user level actions taken on behalf of requests, thereby reducing request latencies. Examples include scheduler activations [6] to exploit kernel level knowledge of request state to guide user level scheduling decisions, upcalls [7, 8] to integrate kernel with user level actions for protocol processing, and real time variants of the latter [9, 10, 11] to address the speci c needs of multi media and real time ....

T. E. Anderson, B. N. Bershad, E. D. Lazowska, and H. M. Levy, \Scheduler Activations: Eective Kernel Support for the User-Level Management of Parallelism," in Proceedings of the 13 th ACM Symposium on Operating System Principles, pp. 95-109, December 1991.

....for events such as kernel preemption and the actions such as resuming a preempted thread on another processor occur by setting a variable that indicates a scheduling change and saving the processor state in memory shared by the kernel and user. This interface di ers from Scheduler Activations[1] or Scheduler Conscious Synchronization[7] in that kernel intervention is not necessary to restart preempted threads. This exibility allows lightweight virtualization of synchronization operations such as spin locks in user space. The risk of busy waiting on a lock owned by a thread blocked in ....

T. Anderson and et. al. Scheduler activations: Eective kernel support for the user-level management of parallelism. ACM Transactions on Computer Systems, 10:53-79, 1992.

....whenever it (the kernel) makes a scheduling decision that a ects one of the kernel level threads utilized in the hybrid mapping. Based on this noti cation, the user level scheduler can adjust the mapping to avoid performance degradation. This modi cation was rst suggested by Anderson et al.[2] They describe a mechanism called scheduler activations in which a user level thread scheduler effectively registers itself with the kernel in order to obtain up calls from the kernel whenever it (the kernel) makes a scheduling decision a ecting a kernel thread associated with that process. An ....

T. Anderson, B. Bershad, E. Lazowska, and H. Levy. Scheduler Activations: Eective Kernel Support for the User-Level Management of Parallelism. ACM Transactions on Computer Systems, 10(1):53-79, February 1992.

....by providing novel kernel level functionality to improve I O performance. Other work has looked at improving threaded and event driven programming models. Banga et al. 2] suggest changes to the UNIX system call interface to better support event driven applications, and scheduler activations [1] can be used to reduce the cost of kernel based thread implementations. The Click modular packet router [19] uses a software architecture which is similar to our framework; packet processing stages are implemented by separate code modules with their own private state. Click modules communicate ....

T. Anderson, B. Bershad, E. Lazowska, and H. Levy. Scheduler Activations: E#ective Kernel Support for the User-Level Management of Parallelism. ACM Transactions on Computer Systems, 10(1):53--79, February 1992.

....using collections of lightweight threads. Both kernel level and user level thread packages have become a common part of computing environments. User level packages can o er substantially better performance than kernel facilities, because thread operations do not require expensive system calls [2, 15, 23]. A user level thread package can also provide exibility in choosing scheduling mechanisms and synchronization primitives, especially if the package is implemented within a powerful higher level language. Wand [35] enumerated three essential mechanisms for language based multithreaded ....

....operating system as well as its architecture. Other important areas to address include concurrent garbage collection, compiler level support for di erent memory consistency models, and improved I O interfaces and processor utilization using an operating system based on scheduler activations [2]. Finally, some progress has recently been made in the exploration of a semantics for a multiprocess SML [8, 31] We hope to to use these results to give a more formal semantics for MP, which would reinforce the advantages of using Standard ML as a base language for multiprocessing. ....

T. E. Anderson, B. N. Bershad, E. D. Lazowska, and H. M. Levy. Scheduler activations: Eective kernel support for the user-level management of parallelism. In Proc. 13th ACM Symposium on Operating Systems Principles, pages 95-109, Oct. 1991.

....from these systems in practice. Other issues: Finally, there are many issues that are not addressed by any of the systems under consideration. For instance a process fork introduces a number of problems for user space communication like dealing with copy on write, and connection inheritance [1]. Handling these problems may require changes to the underlying operating system, so they are not solved yet. Other issues, such as fault containment, fault tolerance, fairness, real time guarantees, quality of service [7] mapping and routing, general purpose flow control, system behavior under ....

T. Anderson, B. Bershad, E. Lazowska, and H. Levy. Scheduler activations: E#ective kernel support for the user-level management of parallelism. In Proceedings of the Thirteenth Symposium on Operating Systems Principles, pages 95--109, Oct. 1991.

....back to pool 1 so that thread A can release the lock. These problems could be avoided if the OS scheduler reported its scheduling decisions to the application. One mechanism to achieve this cooperation is based on the concept of activations, which was rst proposed in an article by Anderson et al.[1] Its authors implemented this mechanism with the FastThread library on the Topaz system. However, this system is no longer running, and the sources were never released. All the terms (activation, upcall, etc. used in this paper come from this article. This mechanism enables the kernel to notify ....

T. Anderson, B. Bershad, E. Lazowska, and H. Levy. Scheduler Activations: Eective Kernel Support for the User-Level Management of Parallelism. ACM Transactions on Computer Systems, 10(1):5379, February 1992.

.... 41] mechanisms for obtaining consistency, 43, 53] atomicity and crash recovery mechanisms in transactions [36] and in laying out data and building query plans in distributed database systems [57, 70] Finally, conventional operating system research into OS structure [19] thread subsystems [4], synchronization [45] and queuing in distributed systems [11, 13] is applicable to the design of the event and thread model in our underlying segment I O layer. 6 Summary In this work, we have proposed to design, build, and evaluate highly available scalable distributed data structures (SDDS s) ....

Thomas E. Anderson, Brian N. Bershad, Edward D. Lazowska, and Henry M. Levy. Scheduler Activations: Eective Kernel Support for the UserLevel Management of Parallelism. ACM Transactions on Computer Systems, 10(1):53-79, February 1992.

....In e#ect, these systems provide a metalevel interface that allows the operating system to be customized. Integrating the MetaBETA runtime system tightly with the operating system can provide more e#cient use of the system resources, e.g. combining user level and system level thread scheduling [Anderson et al. 92] and combining garbage collection and virtual memory management. 6.3 Applicational Implementing distribution and persistence. The experience from designing and implementing the current version of object persistence and distribution has been a major factor in the design of both the metalevel ....

T. E. Anderson, B. N. Bershad, E. D. Lazowska, and H. M. Levy. Scheduler activations: E#ective kernel support for the user-level management of parallelism. ACM Transactions on Computer Systems, 10(1):53--79, February 1992.

No context found.

Thomas E. Anderson, Brian N. Bershad, Edward D. Lazowska, and Henry M. Levy. Scheduler Activations: E#ective Kernel Support for the User-Level Management of Parallelism. ACM Transactions on Computer Systems, 10(1):53--79, February 1992.

No context found.

T. Anderson, B. Bershad, E. Lazowska, and H. Levy. Scheduler Activations: Eective Kernel Support for the User-Level Management of Parallelism. ACM Transactions on Computer Systems, 10(1):5379, February 1992.

No context found.

T. Anderson, B. Bershad, E. Lazowska and H. Levy,Scheduler Activations: E#ective Kernel Support for the User-Level Management of Parallelism. Proc. 13th ACM SOSP, October 1991

No context found.

Anderson, T., Bershad, B., Lazowska, E. and Levy, H. Scheduler Activations: Eective Kernel Support for the User-Level Management of Parallelism, Technical Report 90-04-02, University of Washington, April 1990.

No context found.

T. Anderson, B. Bershad, E. Lazowska and H. Levy. Scheduler activations: e#ective kernel support for the user-level management of parallelism. In Proceedings of the 13th ACM Symposium on Operating Systems Principles. October 1991.

No context found.

T. Anderson, B. Bershad, E. Lazowska, and H. Levy. Scheduler Activations: E#ective Kernel Support for the User-Level Management of Parallelism. ACM Transactions on Computer Systems, 10(1):53--79, February 1992.

No context found.

Thomas E. Anderson, Brian N. Bershad, Edward D. Lazowska, and Henry M. Levy. Scheduler activations: E#ective kernel support for the user-level management of parallelism. ACM Transactions on Computer Systems, 10(1):53--79, February 1992.

No context found.

T. Anderson, B. Bershad, E. Lazowska, and H. Levy. Scheduler Activations: Eective Kernel Support for the User-Level Management of Parallelism. ACM Transactions on Computer Systems, 10(1):5379, February 1992.

No context found.

T. Anderson, B. Bershad, E. Lazowska, and H. Levy. Scheduler Activations: E#ective Kernel Support for the User-Level Management of Parallelism. ACM Transactions on Computer Systems, 10(1):53--79, February 1992.

No context found.

Anderson T., Bershad B., Lazowska E., Levy H. Scheduler Activations: E#ective Kernel Support for the User-Level Management of Parallelism. ACM Transactions on Computer Systems. February 1992.

No context found.

T. Anderson, B. Bershad, E. Lazowska, and H. Levy. Scheduler Activations: E#ective Kernel Support for the User-Level Management of Parallelism. ACM Transactions on Computer Systems, 10(1):53--79, February 1992.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC