D. R. Cheriton and K. J. Duda. A caching model of operating system kernel functionality. In Proc. OSDI, pages 179--193, Monterey, CA, Nov. 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....from a centralized implementation (with a single shared rep) to a distributed one (with a rep per processor) 1.2 Dynamic Customization in K42 Building blocks provide tremendous flexibility in allowing K42 to be customized for an application. As other work in customizable systems demonstrates [3, 4, 6, 8, 9, 26], this flexibility can translate into significant performance gains. Often times though, when a resource is first accessed, it is not clear, especially from the operating system s perspective, what its request pattern will be. Also, an application s use of operating system resources may change ....

....and results, and describes possible future work. Related Work A number of research operating system projects have explored customizability. Recent research in extensible and customizable operating systems includes SPIN [3, 22] VINO [26, 27] Exokernel [8] Fluke [9] L4 [18] Cache Kernel [6], Choices [4] Nemesis [17] Scout [20] and K42 [2] Many of the above operating systems achieve customizability by extensibility. The Exokernel, Fluke, L4, and Cache Kernel allow for customizability by having the kernel redirect hardware events to external address spaces where they can be ....

[Article contains additional citation context not shown here]

K.J. Duda and D.R. Cheriton. A caching model of operating system kernel functionality. In Proc. 1st Symp. on Operating Systems Design and Implementation, pages 179--193, 1994.

....by executing a trap instruction, which transfers execution to kernel code. Kernel code is trusted and has full access to all memory, all I O ports, and all instructions. Protection is the main function of the user kernel boundary, which is also often colloquially referred to as the red line [20]. However, the red line serves several other functions: it enforces resource control and provides safe termination. The user kernel boundary is necessary to enforce resource control. A process running in user mode is subject to policy limitations on the resources that it can consume. The kernel, ....

Cheriton, D. R., and Duda, K. J. A caching model of operating system kernel functionality. In Proceedings of the First Symposium on Operating Systems Design and Implementation (Monterey, CA, November 1994), pp. 179--193.

....Bindings Frame Buffer Translation Lookaside Buffer Network Memory Disk World Wide Web TCP IP POSIX NCSA Mosaic Distributed Shared Memory Barnes Hut VM IPC Traps Exokernel Library Operating system API Applications Figure 3.2: Architecture of the Exokernel 3.3.2. 3 V V [Cheriton,94] from Stanford University builds on previous work on the V distributed operating system which dates back to the early 1980 s. V , like the other library extensible operating systems covered, proposes a split level operating system structure consisting of a minimal kernel (in this case, the Cache ....

Cheriton, D.R., Duda, K.J., "A Caching Model of Operating System Kernel Functionality". Proceedings of the 1st USENIX Symposium on Operating System Design and Implementation (OSDI '94), Monterey CA, U.S.A., November 1994.

.... Systems As mentioned earlier, MIT s Exokernel [Engler95, Kaashoek97] and the University of Cambridge s Nemesis [Leslie96] are the best known examples of vertical operating systems, but other groups have also designed operating systems which address the same goals e.g. Stanford s Cache kernel [Cheriton94], Columbia s Synthesis OS [Massalin92] and the University of Washington s SPIN OS [Bershad95] Exokernel and Nemesis both adhere closely to the vertical OS model described earlier, providing lowlevel OS primitives upon which applications can construct their own resource management policies. ....

D.R. Cheriton and K.J. Duda; \A Caching Model of Operating System Kernel Functionality", 1st Symposium on Operating System Design and Implementation (November 1994).

....we propose is simple: complete elimination of operating system abstractions by lowering the operating system interface to the hardware level. 1 Introduction Throughout the history of computer science there has been a fairly constant opinion that current operating systems are inadequate [4, 7, 9, 11, 15, 18]. The literature is rife with specific examples that describe the cost of the inappropriate, inefficient abstractions peddled by operating systems [2, 4, 12, 13, 18, 23, 24] This situation has persisted for the last three decades, and has survived numerous assaults (object oriented operating ....

....mud, applications can craft their own abstractions, chosen for appropriateness and efficiency, rather than make do with abstractions force fed under duress. With microkernels, applications can have even fewer options than with monolithic ones. Two current OS research efforts, the Cache Kernel [7] and Aegis [9, 10] adhere closely to our precepts for a model operating system. Further experience is needed to see if a low level kernel interface is indeed the panacea that can cure current operating system troubles. The open operating system for a single user machine is motived by similar ....

D. Cheriton and K. Duda. A caching model of operating system kernel functionality. In Proceedings of the Sixth SIGOPS European Workshop, September 1994.

....be eliminated by redirecting call sites to the new function. This technique is explored in Chapter 8. Dynamic kernel instrumentation may also be used to change kernel functionality, such as installing a process specific version of a kernel resource management routine. Extensible operating systems [10, 11, 21, 35, 36, 47] have focused on this subject. It is worthwhile to note that dynamic instrumentation can provide a similar operation in a commodity kernel by splicing the following code at the appropriate kernel policy function: If current pid equals some pid then jump to customized version, else fall through . ....

....locations where kernel code can be inserted, can be almost any machine code instruction within the kernel. Runs on a commodity kernel. This enables instrumentation under real world workloads. It is worthwhile to note that much recent operating system research has taken place on custom kernels [10, 11, 21, 34, 35, 36, 37, 47, 59, 61, 62, 68, 73, 74, 75, 81, 82, 83, 84]; this dissertation shows that run time instrumentation is feasible on a commodity kernel. Runs on an unmodified kernel. This contribution is important, because requiring a modified or somehow customized kernel, even an otherwise commodity one, would likely preclude an instrumentation tool s ....

[Article contains additional citation context not shown here]

D.R. Cheriton and K.J. Duda. A Caching Model of Operating System Kernel Functionality. 1st USENIX Symposium on Operating System Design and Implementation (OSDI), Monterey, CA, November 1994.

....provided by the kernel ) 6. How fair is it (Could one application steal the shared resources ) 7. How programmable are the extensions (Are the interfaces well defined Is there high level language support ) 1. 2 The Outline The three papers selected for my area exam are: Cache Kernel [CD94] from Stanford University. SPIN [BCE 94] from Univ. of Washington, Seattle. Scout [MMO 94] from Univ. of Arizona, Tucson. Both Cache Kernel and SPIN provide application specific customization of the kernel at runtime, but have different underlying philosophy: the Cache Kernel allows ....

....system (Section 3.4.4) the default system servers should be merged into the kernel in their entirety using software faultisolation techniques to avoid IPC overhead. 5. 4 Loading Code into the Kernel Some researchers find the idea of downloading application code into the kernel to be risky [CD94] The idea is surely unconventional, but I consider it an inevitable development. The purpose of the hardware enacted firewall between the kernel and the user mode is to disallow potentially unsafe code from accessing physical resources. Unfortunately, the trap doors provided by the hardware ....

D. Cheriton and K. Duda. A caching model of operating system kernel functionality. Proceedings of the Sixth SIGOPS European Workshop, September 1994.

....where their tasks should execute [21] The ability to get specific finegrained amounts of resources, under process control, is lacking in most popular operating systems. However, new research systems are beginning to address this problem, including Exokernel [47] L5 [91] SPIN [22] Cache Kernel [31], Fluke[54] and Scout[102] Our approach differs either in the level of abstraction provided, or the granularity of resource control, or both. G.2.3.3 Network Communication Support Agents need to communicate with the client they left behind, with servers they visit, and, of course, with other ....

D. Cheriton and K. Duda, "A caching model of operating system kernel functionality," Proc. 1st USENIX Symp. on Operating Systems Design and Implementation(OSDI), November 1994, pp. 179--194.

....valid on remote nodes. The middleware sees the whole network as a source of available resources which creates opportunities for optimizing the implementation of the CORBA middleware. The system, in its turn, uses the local machine as a cache for remote resources, in the spirit of the Cache Kernel [CD94] 4.1 O architecture We developed the O microkernel as an object oriented, adaptable system. It is organized as a set of (nested) resource containers that export hardware resource units. Most of the kernel code comprises a framework that supplies di erent resource unit allocators and a ....

D. Cheriton and K. Duda. A caching model of operating system kernel functionality. In Proceedings of the First Symposium on Operating Systems Design and Implementation, pages 179-193, November 1994.

....general area of adaptive or extensible operating systems. While not using full reflection, these projects have contributed a number of interesting ideas. Examples of adaptive operating systems are Spin [Bershad95] Synthetix [Cowan96] Exokernel Aegis [Engler95] Spring [Mitchell94] Cache Kernel [Cheriton94], Choices [Campbell95] Kea [Veitch96] KTK [Gheith94] Scout [Montz94] and the more architectural work of Chris Maeda on service decomposition [Maeda93, Maeda94] There has also been some related work on incremental specialisation [Druschel93, Pu93, Pu95] Other notable work on the use of ....

Cheriton, D.R., K.J. Duda, "A Caching Model of Operating System Kernel Functionality". Proceedings of the 1st USENIX Symposium on Operating System Design and Implementation (OSDI '94), Monterey CA, U.S.A., November 1994.

....system. Since all of the typical services and abstractions are implemented within user level libraries, the term library operating system has emerged to describe systems of this nature. Two examples of systems supporting the library operating system concept are Aegis [21 23] and the Cache Kernel [11]. The approaches taken by these systems are quite different and will be described separately. 4.3.1 The Exo Kernel Approach The philosophy behind exo kernels, as typified by Aegis, is that abstracting over the hardware resources is the wrong approach for an operating system to take. No matter ....

....Kernel will schedule and dispatch the set of cached thread objects in round robin fashion, effectively sharing the CPU resources among the active threads. When a cached thread blocks, its descriptor is unloaded from the kernel and returned to the application that owns it. According to Cheriton [11], the approach taken by the Cache Kernel offers three benefits. First, the low level caching interface allows applications to control resource management according to any desired policy. Second, the caching approach prevents applications from exhausting the supply of object descriptors as may ....

D.R. Cheriton and K.J. Duda. "A Caching Model of Operating System Kernel Functionality", in Proceedings of the First Symposium on Operating System Design and Implementation, Monterey, California, pp. 179-194, 1994.

....mode and avoid the overhead of system calls. The problem may also be addressed by exposing kernel maintained data structures and thus obviate the need for (some) system calls. The Cache Kernel Approach Rather than exporting an interface to the hardware as do exo kernels, the Cache Kernel [17] supports a small number of primitive abstractions for which all policy decisions are implemented by library operating systems at user level. The interface to the Cache Kernel supports three types of object: address spaces, threads, and descriptors. Each object is represented by a descriptor that ....

....Kernel will schedule and dispatch the set of cached thread objects in round robin fashion, effectively sharing the CPU resources among the active threads. When a cached thread blocks, its descriptor is unloaded from the kernel and returned to the application that owns it. According to Cheriton [17], the approach taken by the Cache Kernel offers three benefits. First, the low level caching interface allows applications to control resource management according to any desired policy. Second, the caching approach prevents applications from exhausting the supply of object descriptors as may ....

D.R. Cheriton and K.J. Duda. "A Caching Model of Operating System Kernel Functionality", in Proceedings of the First Symposium on Operating System Design and Implementation, Monterey, California, pp. 179-194, 1994.

....the server approach and apply it to reorganize the entire operating system. A modest number of basic abstractions are implemented in a small ( micro ) kernel. The bulk of the operating system is moved into separate servers. Examples include Mach [1] and Chorus [45] Exokernel [13] and Cache Kernel [9] attempt to push this organization as far as possible. In a micro kernel, the file system is usually implemented by one or more servers. In the process of reorganization, many micro kernels and their file systems have also acquired an objectoriented flavor. An example is Mach 3.0 [17, 46] While ....

D. R. Cheriton and K. J. Duda. A Caching Model of Operating System Kernel Functionality. In Proceedings of the First USENIX Symposium on Operating Systems Design and Implementation, pages 179--193. USENIX Association, Nov. 1994.

....to the system manager. And, of course, all these parameters should be properly documented. No system has unlimited resources, and a system manager who has the ability to change limits also has the responsibility to live within the available means. The cache kernel model adopted by the V system [4] may ease this problem, by virtualizing as many kernel resources as possible. Even with this design it may be necessary to adjust the size of kernel resource caches, to optimize the cache miss costs for a specific application. 7.2. Resource introspection When an IIS system is dedicated to a ....

David R. Cheriton and Kenneth J. Duda. A Caching Model of Operating System Kernel Functionality. In Proc. First USENIX Symposium on Operating Systems Design and Implementation, pages 179-193. Monterey, CA, November, 1994.

....when other resources are shared, and study the performance of our system for real applications. 8 Related work A number of papers have been published on performance issues in shared memory multiprocessor operating systems, mostly in the context of resolving specific problems in a specific system [5, 6, 9, 22, 26, 28] . These systems were mostly uniprocessor or small scale multiprocessor systems trying to scale up to larger systems. Other workon locality issues in operating system structure were mostly either done in the context of earlier non cachecoherent NUMA systems [8] or, as in the case of Plan 9, were ....

D. R. Cheriton and K. J. Duda. A caching model of operating system kernel functionality. In Proc. Symp. on Operating Systems Design and Implementation (OSDI), pages 179--193, 1994.

....example, the application cannot replace the existing VM abstractions; it can only specialize existing implementations. In addition, this approach makes the kernel more complicated, while AVM simplifies the kernel. Many of the ideas we discuss here could be implemented on top of the Caching Kernel [6]. However, the Caching Kernel is implemented on the Motorola 68040, which has hardware page tables, limiting the flexibility with which page table structures can be experimented. Additionally, it is unclear to what degree applications can control page attributes (e.g. specific page requests or ....

D. Cheriton and K. Duda. A caching model of operating system kernel functionality. In Proceedings of the Sixth SIGOPS European Workshop, September 1994.

....the time needed to run non application code 3. Proposed Architecture Currently, there are many ongoing research projects trying to solve the problems that have arisen with the first generation microkernels. Among the most impressive ones could be listed Exokernel ( 5] 8] Cache Kernel ( 2] [6]) L3 L4 ( 1] 7] Fluke ( 11] and SPIN ( 4] 9] They all take completely different approach to the problem. From the perspective of the abstraction provided, they can classified into three groups: no abstraction systems (Exokernel) minimal abstraction systems (L3, Fluke, Cache ....

Cheriton, D.R., Duda, K.J.: A Caching Model of Operating System Kernel Functionality. Operating System Review, January 1995, pp. 83-86

....and the time needed to run non application code 3. Proposed Architecture Currently, there are many ongoing research projects trying to solve the problems that have arisen with the first generation microkernels. Among the most impressive ones could be listed Exokernel ( 5] 8] Cache Kernel ([2], 6] L3 L4 ( 1] 7] Fluke ( 11] and SPIN ( 4] 9] They all take completely different approach to the problem. From the perspective of the abstraction provided, they can classified into three groups: no abstraction systems (Exokernel) minimal abstraction systems (L3, Fluke, Cache ....

Cheriton, D.R., Duda, K.J.: A Caching Model of Operating System Kernel Functionality. USENIX Symposium on OSDI, 1994, pp. 83-86

....system s #Assenmacher et al. 1993# # kernel is a further example of a small kernel which delegates as much as possible to user space. Besides its two basic concepts protection domain and virtual processor, the Panda kernel handles only interrupts and exceptions. Cache Kernel. The Cache kernel #Cheriton and Duda 1994# is also a small and hardware dependent # kernel. In contrast to the Exokernel, it relies on a small #xed #non extensible# virtual machine. It caches kernels, threads, address spaces and mappings. The term caching refers to the fact that the # kernel never handles the complete set of e.g. all ....

Cheriton, D. R. and Duda, K. J. 1994. A caching model of operating system kernel functionality.In1st USENIX Symposium on Operating Systems Design and Implementation #OSDI#, Monterey, CA, pp. 179#194.

....are similar to those provided by the EROS space bank and the meter mechanism of KeyKOS. Like L3, Fluke s persistence is implemented by a user level pager, but its performance is unreported. Also like L3, Fluke s persistence mechanism lacks a consistency check. Cache Kernel: The Cache Kernel [6] uses a caching approach in some ways similar to EROS. Where EROS writes operating system objects (Processes) back to protected structures (Nodes) the Cache Kernel writes this state back to untrusted application kernels. Hydra and CAL TSS: Hydra [56] and CAL TSS are early software capability ....

D. R. Cheriton and K. J. Duda. A caching model of operating system kernel functionality. In Proc. USENIX Symposium on Operating Systems Design and Implementation, pages 179--193. USENIX Association, Nov. 1994.

....the synthesis kernel is written in assembly, and uses ad hoc runtime code generation macros to specialize system calls. Furthermore, the Aegis system will use runtime code generation to efficiently execute application kernel extensions. Concurrently with our work the SPIN [5] and Cache Kernel [8] projects are also investigating adaptable kernels that allow applications to make policy decisions efficiently. The SPIN system encapsulates policies in spindles that can be dynamically loaded into the kernel. To ensure safety, spindles will be written in a pointersecure language and will be ....

....allows applications to implement services that were formerly enforced by the operating system (e.g. page table management) with a corresponding improvement in flexibility and performance. Additionally, the Cache Kernel has an alarmist view about downloading application code into the kernel [8]; this restricts the efficiency and flexibility of their approach. Techniques for certifying the safety of code include inspection and sandboxing. Deutsch showed how to safely introduce user written instrumentation code into a running kernel by bounding loads, stores, jumps and runtime [11] ....

D. Cheriton and K. Duda. A caching model of operating system kernel functionality. Proceedings of the Sixth SIGOPS European Workshop, September 1994.

....be retrieved approach is complex in calculating the I O cost and needs to be well tuned when the system configuration is changed. The system performance is not addressed either, though the application access pattern knowledge can be used to improve the system performance. The recent Cache Kernel [8] argues the flexibility of existing micro kernel operating systems. User applications can have their own specific Application Kernel (AK) to meet their specific requirements. The Cache kernel and the Application kernel communicate via the memory mapping IPC in loading and unloading the kernel ....

D. R. Cheriton and K. J. Duda, 'A Caching Model of Operating System Kernel Functionality', Proceedings of the First Symposium on Operating Systems Design and Implementation, Monterey, California, USA, November 1994, pp. 179-194.

....to accommodate every application must anticipate all possible needs. The implementation of such an interface would need to resolve all tradeoffs and anticipate all ways the interface could be used. Experience suggests that such anticipation is infeasible and that the cost of mistakes is high [3, 9, 16, 25, 43, 79]. The exokernel architecture attacks this problem by giving untrusted application code as much safe control over resources as possible, thereby allowing orders of magnitude more programmers to innovate and use innovations, without compromising system integrity. It does so by dividing ....

....policy, albeit with one less layer of indirection. For instance, a page table is a very detailed policy that controls how to translate, store and delete mappings and what actions to take on invalid addresses and accesses. 1 Some newer microkernels push the kernel interface closer to the hardware [16, 42, 73], obtaining better performance and robustness than previous microkernels and allowing for a greater degree of flexibility, since shared monolithic servers can be broken into several servers. Techniques to reduce the cost of shared servers by improving IPC performance, moving code from servers into ....

[Article contains additional citation context not shown here]

D. Cheriton and K. Duda. A caching model of operating system kernel functionality. In Proceedings of the First Symposium on Operating Systems Design and Implementation, pages 179--193, November 1994.

No context found.

D. R. Cheriton and K. J. Duda. A caching model of operating system kernel functionality. In Proc. OSDI, pages 179--193, Monterey, CA, Nov. 1994.

No context found.

Cheriton, D.R., Duda, K.J., "A Caching Model of Operating System Kernel Functionality". Proceedings of the 1 st USENIX Symposium on Operating System Design and Implementation (OSDI '94), Monterey CA, U.S.A., November 1994.

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