| Skinner, G. and Wong, T., "Stacking Vnodes: A Progress Report", Proc. of the Summer 1993. |
....of storage servers present a unified, fault tolerant virtual disk space to clients at block level. Virtual disk in MVSS is a different generalized abstraction of storage devices. Stackable file system allows extension of functionalities for existing file systems through Vnode Stacking [12] [28], 13] which allows the interposition and composition of vnodes so that file system modules could be layered on top of each other. FiST [29] is a high level language that can generate codes for stackable file systems. Earlier work on stackable file system has been focused on file level ....
G.C. Skinner and T.K. Wong, "`Stacking' vnodes: A Progress Report," Proc. Summer USENIX Technical Conf., pp. 161-174, June 1993.
....(and most other Unix operating systems) share an ancestral design that is not suited for today s high performance distributed file systems. Some of the main issues are coherence, efficiency, and portability. Some of these problems have been noted in previous efforts to extend the vnode interface [40, 42, 46, 28, 21]. Unfortunately, few improvements have been made to the commercial operating systems; and the performance and interface problems are exacerbated by new serverless distributed systems. 6.1 Cache Coherence The vnode interface was originally designed to integrate NFS into Unix based operating ....
Skinner, G. C., and Wong, T. K. Stacking Vnodes: A Progress Report. In Proc. of the 1993 Summer USENIX (June 1993).
....the UNIX kernel [2] In [3] a prototype implementation is described where the functionality of a file system is extended by stacking a new vnode on top of another vnode. The work described in [3] inspired the Ficus layering mechanism [4] as well as further prototyping work within the SunOS system [17]. The Ficus file system [4, 18] makes further improvements Operation Time in microseconds open 127 4KB read 82 4KB write 86 fstat 28 TABLE 3. SunOS 4.1.3 Performance to the vnode interface to provide a layering mechanism. In particular, the Ficus system contains a delegation mechanism that ....
....or user level implementations. New file systems are added statically, not dynamically. In addition, arcane knowledge of the UNIX kernel in general, and vnodes in particular, is needed to introduce new functionality in the system (see, for example, the discussion of inter vnode locking in [17]) Naming and file mounting issues are closely intertwined with vnode composition. There is no naming system as such and no per process view of the naming space. We believe that introducing the notion of more than one cache manager in the system, and identifying the roles of the pagers and ....
Skinner, Glenn C. and Thomas K. Wong. "Stacking Vnodes: A Progress Report." Proceedings of Summer 1993 USENIX Conference (June 1993): 161--174.
....which is supported by specific file system implementations. This architecture is based on work from several sources. The use of the file system name space to address various elements of the system is reminiscent of Plan 9 Inferno [15] The internal implementation is based primarily on the vnodes [10, 16, 17] architecture present in 4.4BSD [11] among others) and the stackable architectures proposed by Heidemann and Popek [7] The most significant design element is the use of a common buffer representation by all I O elements. Figure 2 illustrates the design of the common buffer representation and ....
Skinner, G. and Wong, T., "Stacking Vnodes: A Progress Report", Proc. of the Summer 1993 Conference, USENIX, 1993.
....system architecture supports interposing (stacking) at the level of an individual file, or at the level of an entire file system. 4 Replacing vnodes The vnode interface [16] provides a mechanism to plug new file systems into the operating system kernel. As vnodes were used in different projects [2,9,24,26,28], however, several problems surfaced. A fundamental problem of vnodes is that they combine multiple interfaces into a single interface. They provide both a naming service and data provision. The data provision of vnodes includes file operations, paging operations, and locking. Combining these ....
....and supporting high availability, extensibility, and system evolution. Our approach to file system extensibility is based on Spring: file systems are extended by providing multiple implementations of the file system interfaces and by using interface inheritance. Most other efforts to extend vnodes [9, 24, 26] try to achieve implementation inheritance (e.g. code sharing) While Solaris MC also shares code through implementation inheritance, we consider this feature to be of secondary importance. Watchdogs [2] are a limited form of extending a file system through stacking. Our framework supports the ....
Skinner, Glenn C., and Thomas K. Wong, "Stacking Vnodes: A Progress Report." Proceedings of the Summer 1993 Usenix Conference, 1993.
No context found.
Skinner, G. and Wong, T., "Stacking Vnodes: A Progress Report", Proc. of the Summer 1993.
No context found.
G.C. Skinner, T. Wong, Stacking vnodes: a progress report, in: Proceedings of the 1993.
No context found.
Glenn C. Skinner and Thomas K. Wong, "`Stacking" vnodes: A progress report', USENIX Conference Proceedings. USENIX, June 1993, pp. 161--174.
No context found.
Glenn C. Skinner and Thomas K. Wong, "Stacking Vnodes: A Progress Report." Proceedings of the Summer 1993 Usenix Conference, 1993.
No context found.
Skinner, G. and Wong, T., "Stacking Vnodes: A Progress Report", Proc. of the Summer 1993 Conference, USENIX, 1993.
No context found.
Glenn C. Skinner and Thomas K. Wong, "`Stacking" vnodes: A progress report', USENIX Conference Proceedings. USENIX, June 1993, pp. 161--174.
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