R. Rashid Jr., A. Tevanian, M. Young, M. Young, D. Golub, R. Baron, D. Black, W. Bolosky, and J. Chew. Machine-independent virtual memory management for paged uniprocessor and multiprocessor architectures. In Proceedings of the 2nd Symposium on Architectural Support for Programming Languages and Operating Systems (ASPLOS), October 1987.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....level of indirection to virtualize memory. While the operating system inside a virtual machine allocates physical memory to satisfy the needs of applications, Cellular Disco allocates machine memory as needed to back the physical memory requirements of each virtual machine. Similar to Mach [48], Cellular Disco maintains a pmap data structure for each virtual machine to map its physical addresses to machine addresses. The pmap is an array that has an entry for each physical address the virtual machine is allowed to access. The pmap is stored in the VMDATA data structure, which also ....

R. Rashid, A. Tevanian, M Young, D. Golub, R. Baron, D. Balck, W. J. Bolosky, and J. Chew. Machine-independent virtual memory management for paged uniprocessor and multiprocessor architectures. IEEE Transactions on Computer, 37(8):896--908, August 1988.

....all kernel code, including safe sync, runs in virtual addressing mode with paging enabled [33] which uses virtual addresses to access code and data. Because safe sync accesses virtual addresses, it depends on the FreeBSD virtual memory code and data (such as the doubly linked address map entries [56]) To fix this dependency, we must configure the processor to use physical addresses during safe sync. Second, safe sync uses the low level kernel device drivers to write data to disk. The FreeBSD disk device drivers are quite complex and there is no simple disk device driver routine to ....

R.F. RAshid Jr., A. Tevanian, M. Young, D. Golub, R. Baron, D. Black Jr., W.J. Bolosky, and J. Chew, "Machine-Independent Virtual Memory Management for Paged Uniprocessor and Multiprocessor Architectures," IEEE Trans. Computers, vol. 37, no. 8, pp. 896-908, Aug. 1988.

....as a whole, which does allow file system functions such as layout to be tailored to an application s needs, but is a very heavy weight mechanism for doing so. Linux provides loadable kernel modules that allow entire file systems to be loaded in this fashion. Mach provides for external pagers [11], which are user level daemons that move virtual memory pages between memory and disk. This mechanism could also be exploited by an application to affect layout policies, but is also a heavy weight solution. The Xok exokernel [8] supports user level library file systems (libFSes) The underlying ....

R. Rashid, A. Tevanian, M Young, D. Golub, R. Baron, D. Balck, W. J. Bolosky, and J. Chew. Machine-independent virtual memory management for paged uniprocessor and multiprocessor architectures. IEEE Transactions on Computers, 37(8):896--908, August 1988.

....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 output buffer as read only by the application. If the application modifies the output buffer before the I O operation ....

R.F. Rashid et al., "Machine-Independent Virtual Memory Management for Paged Uniprocessor and Multiprocessor Architectures," IEEE Transactions on Computers, vol. 37, no. 8, August 1988, pp. 896-908.

....the needs of real time systems, let al..one small memory embedded ones. In microkernels designed for general purpose computing such as Mach [14] L3 [25] and SPIN [15] researchers have focused on optimizing kernel services such as thread management [26, 27] IPC [28] and virtual memory management [29]. Virtual memory is not a concern in our target applications. Thread management and IPC are important but not for the same reasons as for general purpose computing. The sources of OS overhead are different for embedded real time systems than for general purpose computing systems and this ....

R. Rashid, A. Tevanian, M. Young, D. Golub, R. Baron, D. Black, W. Bolosky, and J. Chew, "Machine-independent virtual memory management for paged uniprocessor and multiprocessor architectures," IEEE Transactions on Computers, vol. 37, no. 8, pp. 896--908, August 1988.

....flat address space In the first model all data resides in a single fiat address space with no structure is imposed upon it. This provides an addressing environment similar to that provided to a conventional Unix process. This model is used to implement the Napier88 persistent programming system [32]. The construction of very large stores using this technique was, until recently, not feasible on conventional architectures due to address size limitations. However, the advent of machines such as the DEC Alpha [41] and the MIPS R4000 [18] which (logically) support a 64 bit address has created ....

....are also responsible for maintaining the data when it is not RAM resident. Rather than being part of the kernel, managers are ordinary programs which reside and execute within their own containers; their state is therefore resilient. The concept of a manager is similar to the Mach external pager [32, 44] which has been successfully used to implement a coherent distributed persistent address space [20] In common with Mach and more recent systems [13, 19] managers are responsible for: provision of the pages of data stored in the container, responding to access faults, and receiving data ....

Rashid, R., Tevanian, A., Young, M., Golub, D., Baron, R., Black, D., Bolosky, W. and Chew, J. "Machine-Independent Virtual Memory Management for Paged Uniprocessor and Multiprocessor Architectures", Proceedings of the Second International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS II), Palo Alto, ACM Order Number 556870, pp. 31- 39, 1987.

....are also responsible for maintaining the data when it is not RAM resident. Rather than being part of the kernel, managers are ordinary programs which reside and execute within their own containers; their state is therefore resilient. The concept of a manager is similar to the Mach external pager [31] which has been successfully used to implement a coherent distributed persistent address space [19] In common with Mach, managers are responsible for: provision of the pages of data stored in the container, responding to access faults, and receiving data removed from physical memory by the ....

Rashid, R., Tevanian, A., Young, M., Golub, D., Baron, R., Black, D., Bolosky, W. and Chew, J. "Machine-Independent Virtual Memory Management for Paged Uniprocessor and Multiprocessor Architectures", Proceedings of the Second International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS II), Palo Alto, pp.31-39, 1987.

....call opens it again and restores the file descriptor for it. 3.3 Restoring Memory Performance penalties of the process cleaning are mainly due to copying memory for saving and restoring the state of a process. To reduce the amount of copied memory, Compacto uses a technique known as copy on write [4, 15]. Furthermore, the users can select an implementation strategy for restoring a memory image. As shown in Figure 1, a typical server running on Compacto saves its state only once and repeatedly restores the saved state whenever it finishes handling a request. The user can select an implementation ....

R. Rashid, A. Tevanian, M. Young, D. Golub, and R. Baron. MachineIndependent Virtual Memory Management for Paged Uniprocessor and Multiprocessor Architectures. In Proceedings of the 2nd International Conference on Architectural Support for Programming Languages and Operating Systems, pages 31--39, Oct. 1987.

....5.5 for the simplest implementation to 9 for the most complex. The second stage of our project tackles this problem. 4 Persistent structures Current systems use copy on write (COW) techniques to reduce fork overhead, but both the Machand the SVR4 like COW implementations present problems. Mach [6] uses memory object chains to handle COW pages. As processes fork and modify COW pages, potentially long object chains can be created, which must be traversed at each page fault. Under this approach only objects not pages are reference counted. As a consequence, a page can be completely shadowed ....

R. Rashid, A. Tevanian, M. Young, D. Golub, R. Bar, D. Black, W. Bolosky, and J. Chew. Machine-independent virtual memory management for paged uniprocessor and multiprocessor operating system. IEEE Transactions on Computers, C-37:896-908, 1988.

....issues, the old VAX based VM system was replaced with a new VM system for 4.4BSD [12] The 4. 4BSD virtual memory system (BSD VM) is a modified version of the 1 Note that UVM is a name, not an acronym virtual memory system that was written for Carnegie Mellon University s Mach operating system [18]. The BSD VM system features a clean separation of machinedependent functions, support for mmap, and fine grain data structure locking suitable for multiprocessor systems. 1.1 Why Replace BSD VM The BSD VM system has four main drawbacks that contributed to our decision to replace it: complex ....

....systems, for example recent work on the L4 microkernel [10] has shown that a port of Linux to L4 can run with a minimal performance penalty. However, interactions with other extensions may have an adverse effect. The two virtual memory systems most closely related to UVM are the Mach VM system [18] and the SunOS VM system [4, 9, 13] Since BSD VM is based on Mach VM, most of the discussion of BSD VM in this paper applies to both VM systems (and to a lesser extent the FreeBSD VM system) As described in Section 5 UVM incorporates and extends parts of SunOS VM s anonymous memory management ....

R. Rashid et al. Machine-independent virtual memory management for paged uniprocessor and multiprocessor architectures. IEEE Transactions on Computing, 37(8), August 1988.

....the system s page fault handler to guarantee that results are the same as if data had been copied to distinct system pages. The modification affects write faults on regions for which the faulted process has write permissions and the faulted page is found in the top memory object backing the region [17]: If the page s output reference count is nonzero, the system recovers the process by invalidating all mappings of the page, copying the contents of the page to a new page, swapping pages in the memory object, and mapping the new page to the same virtual address in the process, with writing ....

R. Rashid, A. Tevanian, M. Young, D. Golub, R. Baron, D. Black,W. Bolosky and J. Chew. "Machine-Independent Virtual Memory Management for Paged Uniprocessor and Multiprocessor Architectures", in Proc. 2nd ASPLOS, ACM, Oct. 1987, pp. 31-39.

....to enforce system security policy. As has been known for quite some time [24, 15, 11] an effective security policy must be able to control all accesses to all objects exported by components (and protect itself from tampering [1] Unfortunately, early component based systems, such as Mach [18], Chorus [19] and Spring [16] only control access to component communication, not objects. In addition, despite the limited security flexibility provided by these systems, they still were shown to display less than effective performance [12] Therefore, building a high performance system that ....

....security architecture should be able to handle these mechanisms. The traditional problem in enforcing system security policies on server objects is that the system TCB knows the security policy, but the server knows the object space. Traditional micro kernel security models, such as those in Mach [18], Chorus [19] and Spring [16] control communication between protection domains, but depend on the servers to enforce the security policy on objects and control delegation. Thus, the enforcement of system security policy on object accesses is primarily the job of the servers. DTOS [15] addresses ....

R. Rashid, A. Tevanian Jr., M. Young, D. Golub, D. Baron, D. Black, W. J. Bolosky, and J. Chew. Machine-independent virtual memory management for paged uniprocessor and multiprocessor architectures. IEEE Transactions on Computers, 37(8):896--908, August 1988.

....the needs of real time systems, let al..one small memory embedded ones. In microkernels designed for general purpose computing such as Mach [1] L3 [20] and SPIN [3] researchers focused on optimizing kernel services such as thread management [5, 2] IPC [19] and virtual memory management [25]. Virtual memory is not a concern in our target applications. Thread management and IPC are important, but sources of overhead are different for embedded real time systems, necessitating different optimization techniques. Thread management is a concern in typical microkernels because either the ....

RASHID,R .,TEVANIAN,A.,YOUNG,M.,GOLUB, D., BARON,R.,BLACK,D.,BOLOSKY,W.,AND CHEW, J. Machine-independent virtual memory management for paged uniprocessor and multiprocessor architectures. IEEE Transactions on Computers 37,8 (August 1988), 896--908.

....file descriptors associated with the calling thread. The pthread exec np( also incurs latency in initializing the new protection domain. 6 Related Work The design of the Roadrunner pk memory protection subsystem and the structure of this paper were influenced most heavily by Rashid, et al. in [7]. However, the Mach VM system provides separate, demand paged, virtual address spaces as compared to the single address space in Road runner pk. The Roadrunner pk design does away with the address map abstraction. The single address space allows the use of the generic interface given in Table 1, ....

Rashid, R., et. al., "Machine-Independent Virtual Memory Management for Paged Uniprocessor and Multiprocessor Architectures", Proc. of the 2nd Symposium on Architectural Support for Programming Languages and Operating Systems (ASPLOS), ACM, 1987.

....occur when the required page is in memory, but the page table does not contain the required translation to the page. Copy on write faults are also included in this category. pendent portions of the system. The literature describes several implementations of machinedependent MMU hat layers 2 [1, 4, 5, 10]. Although hat implementations normally do support multiprocessor systems, to our knowledge there is no published material that describes the performance aspects of the hat layer and the scalability of the implementations on multiprocessor systems. Other related material includes a large body of ....

Rashid, R., A. Tevanian, M. Young, D. Golub, R. Baron, D. Black, W. Bolosky, and J. Chew. "Machine-Independent Virtual Memory Management for Paged Uniprocessor and Multiprocessor Architectures." IEEE Transactions on Computers 37 (8) (August 1988): 896--908.

....2.2 Application Managed Replacement Giving specialized applications more control over their own memory management to improve their performance has been suggested before. For instance, the Mach OS supports external pagers to allow applications to control the backing storage of their memory objects [18]. Extensions to the external pager interface have been used to implement user level page replacement polices [14] and to support discardable pages (i.e. dirty pages that do not need to be written to backing store) 20] In contrast, our approach shows that specialized applications can and should ....

....over their own memory management decisions. Most previous work in this area has focused on how the OS can provide this functionality to the applications. For instance, the Mach operating system supports external pagers to allow applications to control the backing storage of their memory objects [18]. Extensions to the external pager interface have been used to implement userlevel page replacement policies [14] and to support discardable pages (i.e. dirty pages that do not have to be written to backing store) 20] More aggressive application control of physical memory was implemented in the ....

R. Rashid, A. Tevanian, Jr., M. Young, D. Golub, R. Baron, D. Black, W. Bolosky, and J. Chew. Machine-Independent Virtual Memory Management for Paged Uniprocessor and Multiprocessor Architectures. In Proc. of the 2nd ASPLOS, Oct. 1987.

....convenience. 3 4.1. Address Space Management The implementation uses several techniques to reduce the overhead of as management. To reduce the time to find the segment for a virtual address, it maintains a hint naming the last segment found, in a manner similar to the technique used in Mach [4]. Any time the as layer translates a virtual address to a segment, this hint is used as the starting point to begin the search. Another optimization reduces the total number of segments in a given address space by allowing segment drivers to coalesce adjacent segments of similar types. This ....

Rashid, R., A. Tevanian, M. Young, D. Golub, R. Baron, D. Black, W. Bolosky, J. Chew, "Machine-Independent Virtual Memory Management for Paged Uniprocessor and Multiprocessor Architectures", Operating Systems Review, Volume 21, No. 4, October, 1987.

....the context of computer systems, the idea of grouping related tasks that share resources comes in various shapes. The concept of concurrency [11] 12] which is applied in all modern operating systems (e.g. BSD UNIX [13] abstracts the execution of applications as processes. The thread concept [14], 15] 16] allows to structure application processes along related tasks. Modern languages like Java [17] give explicit language support for threads and concurrency issues to enhance the productivity of development. The Scout operating system [18] is built around the notion of a path as a ....

R. Rashid, A. Tevanian, M. Young, D. Golub, R. Baron, D. Black, W. Bolosky, and J. Chew, "MachineIndependent Virtual Memory Management for Pages Uniprocessor and Multiprocessor Architecture", , Carnegie-Mellon University, 1987.

No context found.

R. Rashid Jr., A. Tevanian, M. Young, M. Young, D. Golub, R. Baron, D. Black, W. Bolosky, and J. Chew. Machine-independent virtual memory management for paged uniprocessor and multiprocessor architectures. In Proceedings of the 2nd Symposium on Architectural Support for Programming Languages and Operating Systems (ASPLOS), October 1987.

No context found.

R. Rashid, A. Tevanian, M. Young, D. Golub, R. Bar, D. Black, W. Bolosky, and J. Chew. Machine-independent virtual memory management for paged uniprocessor and multiprocessor operating system. IEEE Transactions on Computers, C-37:896-- 908, 1988.

No context found.

R. Rashid, A. Tevanian, M. Young, et al. Machine-independent virtual memory management for paged uniprocessor and multiprocessor architectures. In ASPLOS [ASPLOS1987], pages 31--39. 68

No context found.

R. Rashid, A. Tevanian, M. Young, D. Golub, R. Bar on, D. Black, W. Bolosky and J, Chew, `Machine-independent virtual memory management for paged uniprocessor and multiprocessor architectures', IEEE Trans. Computers, C-37, 896--908 (1988).

No context found.

R. Rashid, A Tevanian, M. Young, D. Golub, R. Baron, D. Black, W. Bolosky and J. Chew, `Machine-independent virtual memory management for paged uniprocessor and multiprocessor architectures', IEEE Trans. Computers, 37, (8), 896--908 (1988).

No context found.

Rashid, R., Tevanian, A., Young, M., Golub, D., Baron, R., Black, D., Bolosky, W. and Chew, J. Machine-Independent Virtual Memory Management for Paged Uniprocessor and Multiprocessor Architectures. IEEE Transactions on Computers 37 (8): 896-908, 1988.

No context found.

R. Rashid, A. Trevanian, M. Young, D. Golub, R. Baron, D. Black, W. Bolosky, and J. Crew. Machine-independent virtual memory management for paged uniprocessor and multiprocessor architectures. IEEE Transactions on Computers, 37(8):896--908, August 1988.

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