The duality of memory and communication in the implementation of a multiprocessor operating system

Young, Michael W.; Tevanian, Avadis; Rashid, Richard F.; Golub, David; Eppinger, Jeffrey L.

doi:10.1145/37499.37507

Cited by 93 publications

(23 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…An object is similar to a closure: it implements an interface and consists of a program and state. These objects extend the objects exposed by the underlying microkernel, Mach [39], to include standard system functionality and to dictate system policy. System services are made available exclusively through objects.…”

Section: The Gnu Hurd's Architecturementioning

confidence: 99%

A critique of the GNU hurd multi-server operating system

Walfield¹,

Brinkmann²

2007

SIGOPS Oper. Syst. Rev.

View full text Add to dashboard Cite

The GNU Hurd's design was motivated by a desire to rectify a number of observed shortcomings in Unix. Foremost among these is that many policies that limit users exist simply as remnants of the design of the system's mechanisms and their implementation. To increase extensibility and integration, the Hurd adopts an objectbased architecture and defines interfaces, which, in particular those for the composition of and access to name spaces, are virtualizable.This paper is first a presentation of the Hurd's design goals and a characterization of its architecture primarily as it represents a departure from Unix's. We then critique the architecture and assess it in terms of the user environment of today focusing on security. Then follows an evaluation of Mach, the microkernel on which the Hurd is built, emphasizing the design constraints which Mach imposes as well as a number of deficiencies its design presents for multi-server like systems. Finally, we reflect on the properties such a system appears to require.

show abstract

Section: The Gnu Hurd's Architecturementioning

confidence: 99%

A critique of the GNU hurd multi-server operating system

Walfield¹,

Brinkmann²

2007

SIGOPS Oper. Syst. Rev.

View full text Add to dashboard Cite

show abstract

“…It also allows the allocation and deallocation of areas in virtual memory so that spaces for the volatile and stable areas can be dynamically allocated, enlarged, and deallocated as required. Furthermore, Mach provides an external pager interface [55]. By writing an external pager, we can ensure that the disk backing store for the stable area survives a crash, and we can also control paging to the backing store and coordinate it with writing to the log.…”

Section: Porting Argus To Machmentioning

confidence: 99%

“…Stable heap management will make it easier to write reliable programs and could be useful in programming languages for reliable distributed computing [14,29], programming languages with persistent storage [1,3], and object-oriented database systems [11,33,52,55].…”

Section: Introductionmentioning

confidence: 99%

Atomic incremental garbage collection and recovery for a large stable heap

Kolodner

Weihl

1993

Proceedings of the 1993 ACM SIGMOD International Conference on Management of Data

View full text Add to dashboard Cite

A stable heap is storage that is managed automatically using garbage collection, manipulated using atomic transactions, and accessed using a uniform storage model. These features enhance reliability and simplify programming by preventing errors due to explicit deallocation, by masking failures and concurrency using transactions, and by eliminating the distinction between accessing temporary storage and permanent storage. Stable heap management is useful for programming languages for reliable distributed computing, programming languages with persistent storage, and object-oriented database systems.Many applications that could benefit from a stable heap (e.g., computer-aided design, computer-aided software engineering, and office information systems) require large amounts of storage, timely responses for transactions, and high availability. We present garbage collection and recovery algorithms for a stable heap implementation that meet these goals and are appropriate for stock hardware. The collector is incremental: it does not attempt to collect the whole heap at once. The collector is also atomic: it is coordinated with the recovery system to prevent problems when it moves and modifies objects. The time for recovery is independent of heap size, and can be shortened using checkpoints.An object in a stable heap is stable and survives crashes if it is reachable from a root that is designated stable; other objects are volatile (e.g., objects local to procedure invocations) and do not need to survive crashes. By tracking objects as they become stable and dividing the heap into a stable area and a volatile area, our algorithms incur the costs of crash recovery and atomic garbage collection only for stable objects. Our tracking algorithm is concurrent; multiple transactions can invoke the tracking algorithm at the same time.We present a formal specification for a stable heap and a formal description of the atomic incremental garbage collector and recovery system used to implement it. We demonstrate the correctness of the algorithms by exhibiting the abstraction function and the invariants they maintain. We also describe a prototype that we implemented to show the feasibility of our algorithms.

show abstract

“…hypercubes) and network shared memory imin a sequence of accesses a page should be replicated into plementations [4,9,21] do not support this functionality, other clusters, without look-ahead. Under our model all clusters are uniformly equidistant; if a page is not resident locally, the cost to access it does not depend on the cluster in which 3 Basic Problem it is accessed.…”

Section: Basic Algorithmsmentioning

confidence: 99%

Competitive Management of Distributed Shared Memory

Black¹,

Gupta²,

Weber³

1988

View full text Add to dashboard Cite

H \Abstract useful for this work because they explicitly address the conMstant factors ignored by standard complexity analysis. and This paper presents and analyzes algorithms for managing because they are well-suited to the analysis of resource manthe distributed shared memory present in non-uniform memagement problems. Previous work has developed competitive ory access multiprocessors and related systems. The cornalgorithms for the related problems of optimizing the use of < petitive properties of these algorithms guarantee that their snoopy caches [8].performance is within a small constant factor of optimal even The performance results for these algorithms are based on though they make no use of any information about memory trace-driven simulations of several production applications reference patterns. Both hardware and software implementafrom UMA multiprocessors. These results show that the protion concerns are covered. A case study of the Mach operating posed algorithms attain total speedups of 5 to 10 over random system indicates that integration of these algorithms into opassignment of pages. This indicates that significant locality erating systems does not pose major problems. On the other (both code and data) may exist in a large class of multiproceshand. hardware support is required to obtain the full funcsor applications, and that this locality can be detected and tionality of the algorithms. We also sketch possible algorithm exploited automatically. As a result such applications may extensions to additional hardware architectures and softwae not require extensive design changes or modifications for use programming models. 51:7 -.. "on NUMA multiprocessors; no such changes or modifications Trace driven simulations are used to evaluate our approach were made to our applications. and compare it to other alternatives. Speedups of 5 to 10 over This paper cctcentrates on the application aspects of our random assignment of pages on production applications are work. Proofs o: -e competitive properties of the algorithms achieved without modifying the applications for non-uniform can be found in [2]. The next section presents a basic model memory access (NUMA) architectures. We compare our prothat covers the systems to which our algorithms are appliposed hardware support with the more aggressive approach cable. This is followed by an introduction to competitive of fully-consistent caches. An additional factor of 2 to 3 in algorithms. Section 4 breaks down the basic problem and performance can be obtained from the cache approach, but at presents our competitive algorithms for solving it. Sections 5 the cost of much more hardware. These results indicate that and 6 continue with a discussion of implementation concerns the algorithms and their hardware support may represent a including the difficulties imposed by most current hardware. viable cost/performance tradeoff.Section 7 presents our performance results from trace driven simulations. Sections 8, 9, and 10 briefly discuss extensions of this work. Sections 11...

show abstract

The duality of memory and communication in the implementation of a multiprocessor operating system

Cited by 93 publications

References 11 publications

A critique of the GNU hurd multi-server operating system

A critique of the GNU hurd multi-server operating system

Atomic incremental garbage collection and recovery for a large stable heap

Competitive Management of Distributed Shared Memory

Contact Info

Product

Resources

About