M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, and A. Gupta, "The Impact of Architectural Trends on Operating System Performance," in Proceedings of the Fifteenth ACM Symposium on Operating System Principles. Copper Mountaint Resort, CO, 1995, pp. 285-298.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....applications executing real world online transaction processing loads. An important issue in database systems is the overhead of processing I O operations. For realistic on line transaction processing (OLTP) database loads, I O can account for a significant percentage of total execution time [26]. Achieving high transaction rates, however, can only be realized by reducing I O overheads on the host CPU and providing more CPU cycles for transaction processing. For this reason, storage architectures strive to reduce costs in the I O path between the database system and the disk subsystem. ....

....TruCluster systems [6] provide unified access to remote disks in a cluster, through a memory channel interconnect [18] which bears many similarities with VI interconnects. Recent work has also examined the use of VI in other storage applications, such as clustered web servers [7] The authors in [26, 21, 1] study the interaction of OLTP workloads and various architectural features, focusing mostly on smaller workloads than ours. Also, work in disk I O [30, 15, 14, 29, 2] has explored many directions in the general area of adding intelligence to the storage devices and providing enhanced features. ....

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, and A. Gupta. The impact of architectural trends on operating system performance. In Symposium on Operating Systems Principles, pages 285--298, 1995.

....In particular, processor speed is not a suitable predictor of NIDS performance, as demonstrated by a nominally slower Pentium 3 system outperforming a Pentium 4 system with higher clock frequency. Memory bandwidth and latency is the most significant contributor to the sustainable throughput [19]. Furthermore, a comparison between two popular operating systems reveals that the more efficient interrupt handling by the Linux kernel leads to non negligible performance improvements. Multiprocessor architectures, on the other hand, offer a relatively small advantage for network intrusion ....

M. Rosenblum, E. Bugnion, S. Herrod, E. Witchel, and A. Gupta, "The Impact of Architectural Trends on Operating System Performance," Proc. 15th ACM Symp. Operating System Principles, ACM Press, New York, N.Y, 1995.

....assign threads to queries; therefore, context switching across concurrent requests is likely to destroy data and instruction locality in the caches. When running OLTP workloads, for instance, most misses occur due to conflicts between threads whose working sets replace each other in the cache [8] [9]. As incoming requests go through different modules of the server code, while the CPU switches execution between concurrent requests. Figure 1 illustrates an example where successive executions of module 2 are interleaved with executions of other modules, forcing module 2 s content to be ....

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, and A. Gupta, "The Impact of Architectural Trends on Operating System Performance," in Proceedings of the Fifteenth ACM Symposium on Operating System Principles. Copper Mountaint Resort, CO, 1995, pp. 285-298.

....address translation options for virtually addressed caches have been outlined, the motivation for their use, and the use of non conventional address translation mechanisms, is now presented. 2.5. 1 Improved translation coverage It has been found that handling TLB misses can consume from 5 40 [17, 22, 23, 34] of execution time, depending on workload. To improve this it is necessary to either improve the TLB hit ratio or the miss penalty in hardware and software, or to find alternative means of providing translation with lower overheads. For example by increasing the TLB coverage, using a multilevel ....

....and SPEC benchmarks was added utime, rename, rmdir, umask, gettimeofday, munmap, llseek. A limitation of the above approach was that the kernel impact on the caches and TLBs was ignored, with the exception of the TLB refill handler, in some the cases the most time consuming OS component [34], which was simulated. The actual kernel overhead would depend on the specific kernel used, and would introduce additional parameters, together with potential kernel modifications, resulting in greater complexity Benchmark name Number of instruction references to skip go 1000M m88ksim N A gcc ....

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, and A. Gupta. The impact of architectural trends on operating system performance. In Proceedings of the fifteenth ACM symposium on Operating systems principles, pages 285--298. ACM Press, 1995.

....section we describe the infrastructure and methodology used in our simulation based experiments. 3. 1 SMT and superscalar simulation methodology Our simulator is an enhanced version of the Alpha based SMT simulator used in [25] It combines an executiondriven processor simulator with the SimOS [26] machine simulation framework. The processor simulator models the pipeline and the memory system in great detail. Table 1 lists the base processor parameters; see [23] for a more complete list) The SMT pipeline consists of 9 stages, with 2 stages each dedicated to reading and writing the large ....

ROSENBLUM, M., BUGNION, E., HERROD, S., WITCHEL, E., AND GUPTA, A. The impact of architectural trends on operating system performance. In Symposium on Operating Systems Principles (December 1995).

....in between) DiskSim was originally part of a larger, system level model [5, 2] that modeled each request s interactions with executing processes, but has been separated out for public dissemination. As a result, it can be integrated into full system simulators (e.g. simulators like SimOS [14]) with little difficulty. 1.2 What DiskSim Does Not Do DiskSim, by itself, simulates and reports on only the performance related aspects of the storage subsystem. It does not model the behavior of the other computer system components or interactions between them and the storage sub system. 2 ....

Mendel Rosenblum, Edouard Bugnion, Stephen Alan Herrod, Emmett Witchel, and Anoop Gupta. The impact of architectural trends on operating system performance. ACM Symposium on Operating System Principles (Copper Mountain Resort, CO, 3 6 December 1995.

....in all I O operations, its performance directly affects I O performance. Although the performance of the central processing unit (CPU) is generally improving at a faster rate [44] than I O device performance [57] operating systems do not benefit from this trend as much as applications [74][88]. Because operating system code is executed relatively infrequently, it incurs many cache and translation lookaside buffer (TLB) misses and is hence more limited by main memory performance. OS code also does not benefit from dynamic instruction scheduling as much as application software because ....

....processor cycles and displaces cache and TLB entries that later need to be reloaded by the application. Performance of copy operations is largely determined by main memory performance, increases in CPU compute performance generally do not translate into proportionally improved copy bandwidth [88]. The amount of overhead scales with the request size and limits the achievable I O performance. Copy on write [86] is an optimization that attempts to avoid copying of output data and uses the application buffer for I O. Rather than copying the data, the kernel marks the pages that contain the ....

M. Rosenblum et al., "The Impact of Architectural Trends on Operating System Performance," Proc. 15ths ACM Symp. Operating System Principles (SOSP-15), ACM Press, New York, N.Y., 1995, pp. 285-298.

....of user and kernel activity across all major hardware resources. We also quantify the effect of kernel activity on overall machine performance. SimOS makes possible architectural studies that can accurately measure all OS activity. One of the first SimOS based studies, by Rosenblum et al. [65], examined the performance of successive generations of superscalars and multiprocessors executing program development and database workloads. Their investigation focused on cache performance and overall memory performance of different portions of the operating system as different ....

ROSENBLUM, M., BUGNION, E., HERROD, S., WITCHEL, E., AND GUPTA, A. The impact of architectural trends on operating system performance. In Proceedings of the Symposium on Operating Systems Principles (December 1995). 122

....only a ected if they interact with the untrusted network because this attack would be blocked at the perimeter of the network. This makes network DoS attacks much more lethal. One alternate possibility for implementing a distributed rewall is to run the host operating system in a virtual machine [37]. The virtual machine may be managed by the administrators, while the operating system is under control of the end user. However, the virtual machine may require supporting several di erent processor architectures. In addition, virtual machines have historically provided virtual interfaces to all ....

Mendel Rosenblum, Edouard Bugnion, Stephen Alan Herrod, Emmett Witchel, and Anoop Gupta. The impact of architectural trends on operating system performance. ACM Symposium on Operating System Principles (Copper Mountain Resort, CO, 3-6 December 1995.

....is because the applications are in general multi threaded and exercise the I O subsystem much more extensively. This trend is likely to continue in the near future and it is very important to consider the operating system not only for complete system evaluations as other studies have pointed out [21], but also when attempting to optimize the hardware and or the application [32] A detailed characterization of the interactions between the application OS hardware can have considerable ramifications in the design of each system component, and this paper takes a step in that direction. We focus ....

....commercial JDK [13] from Sun Microsystems simulated on top of IRIX 5.3. Vortex is a database manipulation code and gcc is a compiler code from the SPECint95 benchmark suite. Pmake is a program development workload that is a variant of the compiled phase of the modified Andrew benchmark employed in [21]. The sendmail benchmark forwards email messages to local accounts on the system using the Simple Mail Transport Protocol (SMTP) The sizes of the messages vary from 1K to 1.5M. We also use three benchmarks that use PostgreSQL [19] 26] a relational database management system (DBMS) The database ....

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, A. Gupta, The Impact of Architectural Trends on Operating System Performance, In Proceedings of the 15th ACM Symposium on Operating System Principles, pages 285-298, 1995.

....real applications. a real OS and runs real applications, a complete machine simulator enables the same types of experiments as storage emulation. Further, by manipulating simulator parameters, the effects of new storage devices on hypothetical machines (e.g. with 10 GHz CPUs) can be evaluated [20, 22]. Unfortunately, substantial effort is required to build and maintain a complete machine simulator, both in terms of correctly executing programs and correctly accounting for time. For example, the SimOS machine simulator required extensive effort to create and validate; just a few years later, ....

Mendel Rosenblum, Edouard Bugnion, Stephen Alan Herrod, Emmett Witchel, and Anoop Gupta. The impact of architectural trends on operating system performance. ACM Symposium on Operating System Principles (Copper Mountain Resort, CO, 3 6 December 1995.

....decision support, and e business applications are driving the development of powerful server systems. From an execution point of view, commercial workloads are different from technical workloads and present more vigorous demands on the memory and storage sub systems [1, 2, 3] In fact, studies [4] that analyzed transaction processing workloads indicate that systems spend a significant fraction of the execution time waiting for I O devices to access the data. Once the data is brought to main memory, the processor uses a substantial amount of the remaining execution time to handle memory ....

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, and A. Gupta, "The impact of architectural trends on operating system performance," in Proceedings of the Fifteenth ACM Symposium on Operating Systems Principles, (Copper Mountain, CO), pp. 285-298, ACM Press, Dec. 1995.

....a commercial JDK from Sun Microsystems simulated on top of IRIX 5.3. Vortex is a database manipulation code and gcc is a compiler code from the SPECint95 benchmark suite. Pmake is a program development workload that is a variant of the compiled phase of the modified Andrew benchmark employed in [Rose95a]. The sendmail benchmark forwards email messages to local accounts on the system using the Simple Mail Transport Protocol (SMTP) The sizes of the messages vary from 1K to 1.5M. We also use three benchmarks that use PostgreSQL [Psql1, Psql2] a relational database management system (DBMS) The ....

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, A. Gupta, The Impact of Architectural Trends on Operating System Performance, In Proceedings of the 15th ACM Symposium on Operating System Principles, pages 285-298, 1995.

No context found.

Mendel Rosenblum, Edouard Bugnion, Stephen A. Herrod, Emmett Witchel, and Anoop Gupta. The impact of architectural trends on operating system performance. In SOSP, 1995.

No context found.

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, and A. Gupta, "The Impact of Architectural Trends on Operating System Performance," in Proceedings of the Fifteenth ACM Symposium on Operating System Principles. Copper Mountaint Resort, CO, 1995, pp. 285-298.

No context found.

M. Rosenblum, et al. "The impact of architectural trends on operating system performance." In Proc. of the 15th ACM SOSP, pages 285--298, December 1995.

No context found.

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, and A. Gupta. The impact of architectural trends on operating system performance. In Proceedings of the fifteenth ACM symposium on Operating systems principles, pages 285--298. ACM Press, 1995.

No context found.

M. Rosenblum, E. Bugnion, S.A. Herrod, E. Witchel, and A Gupta. The impact of architectural trends on operating system performance. In Proc. 15th ACM symposium on Operating systems principles, pages 285--298, Copper Mountain, CO, December 1995.

No context found.

Mendel Rosenblum, Edouard Bugnion, Stephen Alan Herrod, Emmett Witchel, and Anoop Gupta. The impact of architectural trends on operating system performance. ACM Symposium on Operating System Principles (Copper Mountain Resort, CO, 3--6 December 1995.

No context found.

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, and A. Gupta. The impact of architectural trends on operating system performance. In Proceedings of the 15th Symposium on Operating Systems Principles, Copper Mountain, CO, 1995.

No context found.

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, and A. Gupta. The impact of architectural trends on operating system performance. In Symposium on Operating Systems Principles, pages 285--298, 1995.

No context found.

Mendel Rosenblum, Edouard Bugnion, Stephen A. Herrod, Emmett Witchel, and Anoop Gupta. The Impact of Architectural Trends on Operating System Performance. In The 15th ACM Symposium on Operating Systems Principles.

No context found.

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, and A. Gupta. The impact of architectural trends on operating system performance. In 15 Symposium on Operating Systems Principles, Dec 1995.

No context found.

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, and A. Gupta. The impact of architectural trends on operating system performance. In 15th ACM Symposium on Operating System Principles, pages 285--298, December 1995.

No context found.

M. Rosenblum, E. Bugnion, S. A. Herrod, E. Witchel, and A. Gupta. "The Impact of Architectural Trends on Operating System Performance." In Proc. SOSP-15, pp.285-298, 1995.

First 50 documents  Next 50

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