The impact of operating system structure on memory system performance

Chen, J. Bradley; Bershad, Brian N.

doi:10.1145/168619.168629

Cited by 112 publications

(25 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These transitions between the Mach kernel and UNIX server lower the locality of the system references, thus increasing the cache miss-rate. This is consistent with the findings in [3].…”

Section: Incomplete Traces and Miss-ratessupporting

confidence: 93%

“…In both cases, supervisor code doesn't benefit as much from increased associativity as user code. This is surprising, because one of the findings in [3], based on the SPEC92 benchmarks, is that system instruction references benefit significantly from an increase in associativity. For the SDM benchmarks this does not seem to be the case.…”

Section: Incomplete Traces and Miss-ratesmentioning

confidence: 99%

“…This result differs from those reported in [3f which claims that collision misses are not a major cause of performance degradation. We believe that the higher number of collision misses observed in our simulations are due to the multiprogramming workloads we use -the benchmarks traced in [3] were taken from SPEC92 and consist of CPU-bound single process workloads.…”

Section: Comparison To Other Studiesmentioning

confidence: 99%

See 2 more Smart Citations

The inaccuracy of trace-driven simulation using incomplete multiprogramming trace data

Flanagan

Nelson²,

Archibald³

et al.

Proceedings of MASCOTS '96 - 4th International Workshop on Modeling, Analysis and Simulation of Computer and Telecommunication

View full text Add to dashboard Cite

Trace-driven simulation is commonly used to predict the performance of computer systems. However, existing tracing techniques produce traces inadequate for some studies: they do not usually record operating system references, and they produce relatively short traces. This paper explores the impact of these trace distortions on the performance estimates of unaprocessor memory hierarchies using multiprogramming workloads. We used a hardware monitor to capture traces under a variety of workloads and operating systems. Our monitor captures every reference and can record arbitrarily long traces. We quantify memory hierarchy performance using traces of the SPEC SDMl.1 benchmark suite executing on an a486 CPU. To evaluate variations due t o operating systems, we compare these results under both Mach 3.0 and UNIX Sys V R4. We conclude that for current uniprocessors, long but incomplete traces result in modest errors in estimated performance, but for proposed architectures with large delays to main memory, the errors can be significant.

show abstract

“…These transitions between the Mach kernel and UNIX server lower the locality of the system references, thus increasing the cache miss-rate. This is consistent with the findings in [3].…”

Section: Incomplete Traces and Miss-ratessupporting

confidence: 93%

Section: Incomplete Traces and Miss-ratesmentioning

confidence: 99%

Section: Comparison To Other Studiesmentioning

confidence: 99%

See 1 more Smart Citation

The inaccuracy of trace-driven simulation using incomplete multiprogramming trace data

Flanagan

Nelson²,

Archibald³

et al.

Proceedings of MASCOTS '96 - 4th International Workshop on Modeling, Analysis and Simulation of Computer and Telecommunication

View full text Add to dashboard Cite

show abstract

“…In [3] the authors use machine simulation to investigate the impact of two versions of UNIX on the performance of memory subsystems for a variety of benchmarks and full applications. The results obtained are compared with several commonly held views of memory performance, with several cases differing from the expected behaviour.…”

Section: Related Workmentioning

confidence: 99%

To upgrade or not to upgrade? Catamount vs. Cray Linux Environment

Hammond

Mudalige

Smith

et al. 2010

2010 IEEE International Symposium on Parallel &Amp; Distributed Processing, Workshops and PHD Forum (IPDPSW)

View full text Add to dashboard Cite

Modern supercomputers are growing in diversity and complexity -the arrival of technologies such as multicore processors, general purpose-GPUs and specialised compute accelerators has increased the potential scientific delivery possible from such machines. This is not however without some cost, including significant increases in the sophistication and complexity of supporting operating systems and software libraries. This paper documents the development and application of methods to assess the potential performance of selecting one hardware, operating system (OS) and software stack combination against another. This is of particular interest to supercomputing centres, which routinely examine prospective software/architecture combinations and possible machine upgrades. A case study is presented that assesses the potential performance of a particle transport code on AWE's Cray XT3 8,000-core supercomputer running images of the Catamount and the Cray Linux Environment (CLE) operating systems. This work demonstrates that by running a number of small benchmarks on a test machine and network, and observing factors such as operating system noise, it is possible to speculate as to the performance impact of upgrading from one operating system to another on the system as a whole. This use of performance modelling represents an inexpensive method of examining the likely behaviour of a large supercomputer before and after an operating system upgrade; this method is also attractive if it is desirable to minimise system downtime while exploring software-system upgrades. The results show that benchmark tests run on less than 256 cores would suggest that the impact (overhead) of upgrading the operating system to CLE was less than 10%; model projections suggest that this is not the case at scale.

show abstract

“…Cache coherence partially addresses this issue by cachng frequently accessed data, but is only effective for data which exhibits a high degree of spatial and temporal locality. Traditionally, operating systems have been shown to have very poor locality in comparison to application software [CB93].…”

Section: Introductionmentioning

confidence: 99%

(De-)clustering objects for multiprocessor system software

Parsons¹,

Gamsa²,

Krieger³

et al.

Proceedings of International Workshop on Object Orientation in Operating Systems

View full text Add to dashboard Cite

Designing system software for large-scale shared-memory multiprocessors is challenging because of the level of performance demanded by the application workload and the distributed nature of the system. Adopting an objectoriented approach for our system, we have developed a framework for de-clustering objects, where each object may migrate, replicate, and distribute all or part of its data across the system memory using the policies that will best meet the locality requirements for that data. The mechanism for object invocation hdes the internal structure of an object, allowing a request to be made directly to the most suitable part of the object on a per-processor basis without any knowledge of how the object is de-clustered. Method invocation is very efficient, both within and across address spaces, involving no remote memory accesses in the common case. We describe the design and implementation of this framework in Tornado, our multiprocessor operating system.

show abstract

The impact of operating system structure on memory system performance

Cited by 112 publications

References 24 publications

The inaccuracy of trace-driven simulation using incomplete multiprogramming trace data

The inaccuracy of trace-driven simulation using incomplete multiprogramming trace data

To upgrade or not to upgrade? Catamount vs. Cray Linux Environment

(De-)clustering objects for multiprocessor system software

Contact Info

Product

Resources

About