Ben Gamsa, Orran Krieger, Jonathan Appavoo, and Michael Stumm. Tornado: Maximizing locality and concurrency in a shared memory multiprocessor operating system. In Proceedings to the 3rd Symposium on Operating Systems Design and Implementation. USENIX Association, February 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....does not provide fault containment. The Virtual Clusters resource manager does not rely on the cooperation of the various instances of the operating system, and it also supports fault containment. The K42 project [31] at IBM research, which is closely tied with the Tornado operating system [26] from Toronto University, was designed from scratch with scalability as the design goal. It is one of the most scalable operating systems. However, this approach requires a huge development effort to support all of the existing commercial applications on a brand new operating system. By using a ....

Benjamin Gamsa, Orran Krieger, Jonathan Appavoo, and Michael Stumm. Tornado: Maximizing locality and concurrency in a shared memory multiprocessor operating system. In Proceedings of the 3rd Symposium on Operating Systems Design and Implementation (OSDI), pages 87--100, February 1999.

....in Tornado by adopting an object oriented approach where each virtual and physical resource in the system is represented by an independent object so that accesses on di#erent processors to di#erent objects do not interfere with each other. Details of the Tornado operating system can be found in [12, 11]. The natural outgrowth of the Hurricane experience was to build an operating system in which each data structure could specify its own clustering size. Tornado serves as the operating system for the NUMAchine multiprocessor [41] In Tornado, unlike Hurricane, there is only one operating 8 system ....

Ben Gamsa. Tornado: Maximizing Locality and Concurrency in a Shared-Memory Multiprocessor Operating System. PhD thesis, University of Toronto, 1999.

....locks that need to be acquired. Some building blocks, such as those used for a shared file or the Process building block used for a parallel program, may be widely shared across a large multiprocessor. To implement these widely shared building blocks efficiently, the concept of clustered objects [1, 12, 29] has been developed. A clustered object building block is one that can be partitioned into representative (rep) objects, where independent requests on different processors are (in the common case) handled by different representatives of the object. A clustered object is like any other building ....

....K42 Operating System K42 is an operating system designed from the ground up and targeted at multiprocessor shared memory machines. The project is being conducted at IBM Watson Research with collaboration from University of Toronto and other universities. K42 was heavily influenced by Tornado [1, 2, 11, 12, 29], developed at the University of Toronto. Achieving good performance for shared memory multiprocessor programs has received considerable attention [5, 12, 15, 19, 29] K42 s overall structure, algorithms, and data structures have been designed with the purpose of achieving good multiprocessor ....

[Article contains additional citation context not shown here]

B. Gamsa, O. Krieger, J. Appavoo, and M. Stumm. Tornado: Maximizing Locality and Concurrency in a Shared Memory Multiprocessor Operating System. In Symp. on Operating Systems Design and Implementation, pages 87--100, 1999.

....K42 Operating System K42 is an operating system designed from the ground up and targeted at multiprocessor shared memory machines. The project is being conducted at IBM Watson Research with collaboration from University of Toronto and other universities. K42 was heavily influenced by Tornado [1, 2, 11, 12, 29], developed at the University of Toronto. Achieving good performance for shared memory multiprocessor programs has received considerable attention [5, 12, 15, 19, 29] K42 s overall structure, algorithms, and data structures have been designed with the purpose of achieving good multiprocessor ....

Ben Gamsa. Tornado: Maximizing Locality and Concurrency in a Shared Memory Multiprocessor Operating System. PhD thesis, Dept. of Computer Science, University of Toronto, 1999.

....name caching and the FCM provided data le block caching, all of which were not available in the server level experiments. This con guration enabled the scalability of the IPC, VFS, and FCM components of K42 to exercised by the experiment. FCM and IPC scalability were demonstrated by Gamsa et al. [23] on the Tornado Operating System, which is closely related to K42. Similar to previous experiments, each thread executed on its own processor and disk. The workload was executed twice to ensure the FCM and VFS facilities had cached as much data as possible. The rst run was used to warm up the ....

Ben Gamsa, Orran Krieger, Jonathan Appavoo, and Michael Stumm. Tornado: Maximizing locality and concurrency in a shared memory multiprocessor operating system. In Proceedings to the 3rd Symposium on Operating Systems Design and Implementation. USENIX Association, February 1999.

....on large scale, shared memory, NUMA multiprocessor computers. The operating system is a product of research at the University of Toronto in collaboration with the K42 group at the IBM Thomas J. Watson Research Center. The core of K42 is based on the University of Toronto s Tornado Operating System [22]. K42 is the university s third generation of research on scalable operating systems. Hurricane OS Hector Multiprocessor was the rst generation and Tornado OS NUMAchine Multiprocessor [24] was the second generation. K42 runs on 64 bit processors that include the 64 bit PowerPC, MIPS, and AMD ....

....process of experimentation. In general, we have found it important to apply the principles of maximizing locality and Error bars were removed from the 12 processor con guration. concurrency that have been developed in about 10 years of multiprocessor software research by our research group [22]. In addition, the constant average queue length principle outlined by Peacock et al. 56, p. 86] is a fundamental pre requisite to scalability. More speci cally, these principles led to the use of techniques such as padding critical data structures to avoid false sharing of secondary cache ....

[Article contains additional citation context not shown here]

Ben Gamsa. Tornado: Maximizing Locality and Concurrency in a Shared-Memory Multiprocessor Operating System. PhD thesis, University of Toronto, Department of Computer Science, 1999.

....We therefore propose to further optimize SmartApps by interfacing them with an existing customizable OS. While there have been several proposals of modular, customizable OSs, we plan to use the K42 [4] experimental OS from IBM, which represents a commercial grade development of the TORNADO system [21, 5]. Instead of allowing users to actually alter or rewrite parts of the OS and thus raise security issues, the K42 system allows the selective and parametrized use of OS modules (objects) Additional modules can be written if necessary but no direct user access is allowed to them. This approach will ....

J. Appavo B. Gamsa, O. Krieger and M. Stumm. Tornado: Maximizing locality and concurrency in a shared memory multiprocessor operating system. In Proc. of OSDI, 1999.

....applications running on the system, requiring the entire machine to be rebooted. The solutions that have been proposed to date are either based on hardware partitioning [4] 21] 25] 28] or require developing new operating systems with improved scalability and fault containment characteristics [3][8][10] 22] Unfortunately, both of these approaches suffer from serious drawbacks. Hardware partitioning limits the flexibility with which allocation and sharing of resources in a large system can be adapted to dynamically changing load requirements. Since partitioning effectively turns the system ....

....a significant software development effort. Furthermore, adding support for fault containment is a daunting task in practice, since the base operating system is inherently vulnerable to faults. New operating system developments have been proposed to address the requirements of scalability (Tornado [8] and K42 [10] and fault containment (Hive [3] While these approaches tackle the problem at the basic level, they require a very significant development time and cost before reaching commercial maturity. Compared to these approaches, Cellular Disco is about two orders of magnitude simpler, while ....

Ben Gamsa, Orran Krieger, Jonathan Appavoo, and Michael Stumm. Tornado: Maximizing Locality and Concurrency in a Shared Memory Multiprocessor Operating System. In Proceedings of the 3rd Symposium on Operating Systems Design and Implementation (OSDI), pp. 87-100. February 1999.

No context found.

Ben Gamsa, Orran Krieger, Jonathan Appavoo, and Michael Stumm. Tornado: maximizing locality and concurrency in a shared memory multiprocessor operating system. In Symposium on Operating Systems Design and Implementation, pages 87--100, 22-25 February 1999.

No context found.

B. Gamsa, O. Krieger, J. Appavoo, and M. Stumm. Tornado: maximizing locality and concurrency in a shared memory multiprocessor operating system. Symposium on Operating Systems Design and Implementation, pages 87--100, February 1999.

No context found.

Ben Gamsa, Orran Krieger, Jonathan Appavoo, and Michael Stumm. Tornado: maximizing locality and concurrency in a shared memory multiprocessor operating system. In Symposium on Operating Systems Design and Implementation, pages 87-- 100, 22-25 February 1999.

....for thread creation destruction. The systems on which the tests were run are: SGI Origin 2000 running IRIX 6.4, Convex SPP 1600 running SPP UX 4.2, IBM 7012 G30 PowerPC 604 running AIX 4.2.0.0, Sun 450 UltraSparc II running Solaris 2.5.1. Finally, figure 1. 1 are results gathered by Gamsa et al.[12] of simple micro benchmarks run on a number of commercial SMP operating systems. The micro benchmarks are of three separate tests: in core page faults, file stat and thread creation, each with n worker threads performing the operation being tested: Page Fault Each worker thread accessed a set of ....

....in Tornado by adopting an object oriented approach where each virtual and physical resource in the system is represented by an independent object so that accesses on di#erent processors to di#erent objects do not interfere with each other. Details of the Tornado operating system can be found in [12, 11]. The natural outgrowth of the Hurricane experience was to build an operating system in which each data structure could specify its own clustering size. Tornado serves as the operating system for the NUMAchine multiprocessor [41] In Tornado, unlike Hurricane, there is only one operating 8 system ....

Ben Gamsa, Orran Krieger, Jonathan Appavoo, and Michael Stumm. Tornado: Maximizing locality and concurrency in a shared memory multiprocessor operating systems. Submitted to 3rd Symposium on Operating Systems Design and Implementation, February 22-25, 1999, New Orleans, LA.

....to no longer be in use once the threads in the first epoch have ended. At this point, the old chain is quiescent and can be modified at will (including being deleted) We have utilized this style of synchronization, termed Read Copy Update(RCU) to implement a semiautomatic garbage collector [21] and hot swapping in K42. Others have used it in PTX [37] and in Linux to implement a number of optimization such as: lock free module loading and unloading [36, 44, 45] The key to leveraging RCU techniques is being able to divide the work of the system into short lived requests that have a ....

....are made through this reference. K42 can perform a hot swap or interposition on this component by updating the entry in the appropriate tables. Although this incurs an extra pointer dereference per component call, the object translation table has other benefits (e.g. improved SMP scalability [21]) that outweigh this overhead. The remainder of this section describes how K42 utilizes the four system support features described above to provide interposition and hot swapping. 4.3.1 Interposition Object interposition interposes additional functionality around all function calls to an ....

[Article contains additional citation context not shown here]

B. Gamsa, O. Krieger, J. Appavoo, and M. Stumm. Tornado: maximizing locality and concurrency in a shared memory multiprocessor operating system. Symposium on Operating Systems Design and Implementation, pages 87--100, February 1999.

....module cleanup until all racing uses of that module have finished. This eliminates destructive races. More information is available with the patch [Soni01a] 6 Advanced Infrastructure for ReadCopy Update This section describes the design used in the Tornado and K42 research operating systems [Gamsa99] and a more complex but higher performance design for non preemptive Linux kernels. Most, and perhaps all, of the optimizations found in this last implementation can easily be applied to the implementation described in Section 4. 6.1 Tornado K42 Design for ReadCopy Update The K42 and Tornado ....

....management: per process system call tables as well as the multi processor trace data structures used to support user level debugging of multi threaded processes. 8. LAN drivers: resolve races between shutting down a LAN device and packets being received by that device. The Tornado and K42 [Gamsa99] research operating systems independently developed a form of readcopy update, which is used as follows: 1. To provide existence guarantees throughout these operating systems. These existence guarantees simplify handling of races between use of a data structure and its deletion. 2. To identify ....

B. Gamsa, O. Kreiger, J. Appavoo, and M. Stumm. Tornado: maximizing locality and concurrency in a shared memory multiprocessor operating system, Proceedings of the 3rd Symposium on Operating System Design and Implementation, New Orleans, LA, February, 1999.

No context found.

B. Gamsa, O. Krieger, J. Appavoo, and M. Stumm. Tornado: Maximizing Locality and Concurrency in a Shared Memory Multiprocessor Operating System. In Symp. on Operating Systems Design and Implementation, 1999.

....Ring is undergoing testing, and upon completion will link together the Local Rings to form a 48 processor system. Updates on the hardware status, with photographs, can be found at http: www.eecg.toronto.edu parallel. We have developed a custom parallel POSIX compliant operating system, Tornado [3], to take advantage of NUMAchine s inherent clustering and provide usercustomizable support in the kernel for parallel file systems. Tornado boots and runs parallel programs such as the SPLASH 2 suite and numerous X11 programs. The systems group is also working closely with IBM research, and has ....

B. Gamsa, O. Krieger, J. Appavoo, and M. Stumm. Tornado: Maximizing locality and concurrency in a shared memory multiprocessor operating system. In Proc. of the 3rd Symposium on Operating Systems Design and Implementation, pages 87--100, 1999.

No context found.

Ben Gamsa, Orran Krieger, Jonathan Appavoo, and Michael Stumm. Tornado: Maximizing locality and concurrency in a shared memory multiprocessor operating system. In Proceedings to the 3rd Symposium on Operating Systems Design and Implementation. USENIX Association, February 1999.

No context found.

Ben Gamsa. Tornado: Maximizing Locality and Concurrency in a Shared-Memory Multiprocessor Operating System. PhD thesis, University of Toronto, Department of Computer Science, 1999.

No context found.

J. Appavo B. Gamsa, O. Krieger and M. Stumm. Tornado: Maximizing locality and concurrency in a shared memory multiprocessor operating system. In Proc. of OSDI, 1999.

No context found.

B. Gamsa, O. Krieger, J. Appavoo, and M. Stumm. "Tornado: Maximizing Locality and Concurrency in a Shared Memory Multiprocessor Operating System." In Proceedings of the 3rd Symposium on Operating Systems Design and Implementation (OSDI), pp. 87-100, February 1999.

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