V. Singhal, S. V. Kakkad, and P. R. Wilson: Texas: An Efficient, Portable Persistent Store. In A. Albano and R. Morrison, editors, Persistent Object Systems, Workshop In Computing, pages 11-33. Springer-Verlag, 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....[46, 67] has served to demonstrate that a memory mapped architecture with direct application access to data is critical for a significant class of applications. Other memory resident or memory mapped storage architectures such as Recoverable Virtual Memory [65] Cricket [66] and Texas Store [70] have demonstrated significant performance gains when user application code has direct access to some or all of the database storage structures. In these systems, the risk of data corruption due to application errors is significantly higher than for standard database management systems where ....

.... than access in the same process space, even with highly tuned implementations [8] Due to this cost, an increasing number of systems give application code direct access to system buffers, including extensible systems [73] object databases [46, 3] and memory mapped or in memory architectures [41, 66, 70]. In [75] Sullivan and Stonebraker investigate the use of hardware memory protection to improve software fault tolerance in a DBMS environment by guarding data in the buffer cache. For example, calls were added to POSTGRES [73] to unprotect the page containing a tuple before it is updated and to ....

[Article contains additional citation context not shown here]

V. Singhal, S. Kakkad, and P. R. Wilson. Texas: An efficient, portable persistent store. In Proc. Fifth Int'l. Workshop on Persistent Object Systems, September 1992.

....used as a basis for indexing objects in an object manager, it was designed for file system updates, which are short transactions, and is less suitable for long living transactions. Log structured file systems have also been used as the basis for several object managers: the Texas persistent store [19], as a part of the Grasshopper operating system [4] and the Lumberjack object store [5] The storage managers in Texas and Grasshopper are page based, i.e. when an object has been modified, the whole page it resides on has to be written back. Similar to these approaches was Seltzer s ....

V. Singhal, S. Kakkad, and P. Wilson. Texas: An efficient, portable persistent store. In Proceedings of the Fifth International Workshop on Persistent Object Systems, 1992.

....based on the L4 kemel and Pebble [ 10] 1.2 DBMS Componentisation Work on evolving the architecture the DBMS architecture has followed a very similar path to that of the OS but has to some extent lagged behind. For example some research has been carded out on adapting lightweight DBMS (LWDB) [4, 27, 16]. Often, these systems omit features found in traditional DBMS, and include more specialised features which maximise performance. Such systems are very limited to particular applications and most were custom built and remained monolithic. Research has also considered extensible systems to tailor ....

Singhal V., Kakkad S., Wilson P.,' Texas: An Efficient, Portable Persistent Store, Persistent Object Systems': In Proc. Fifth International Workshop on Persistent Object Systems, September 1992

.... are explicitly evaluated (typically, before each update of an object s persistent state) This can be either application controlled or system controlled (based on hardware supported reference identification) In the latter case, the techniques of early or late pointer swizzling are typically used [SKW92, VD92]. Having possible integration with the Relationship Service in mind, we selected the evaluation of dependencies technique for our implementation, since it emulates the graph traversal mechanism used by the Traversal Criteria object from the Relationship Service. 3.2 Design architecture issues ....

V. Singhal, S. V. Kakkad, and P. R. Wilson: Texas: An Efficient, Portable Persistent Store. In Proceedings of the Persistent Object Systems, San Miniato, A. Albano and R. Morrison (Editors), Springer, pages 11-33, 1992.

....from legal values. This data could also be shared by many objects. Since we have a compile time approach, there is no good way of handling pointers. There are different approaches to solve the problem of pointers, such as using smart pointer structures, pointer swizzling and object pickling [20]. But they are of limited utility here, given the presence of non object data in a C program, and we seek a generalized solution. Two decisions are made to simplify the pointer support. First, the malloc library is modified so that when queried, it can provide information about memory blocks. ....

Vivek Singhal, Sheetal Kakkad, and Paul R. Wilson. Texas: An efficient, portable persistent store. in persistent object systems. In Persistent Object Systems: Proceedings of the Fifth International Workshop on Persistent Object Systems, September 1992.

....particularly low overhead along the critical path of object access. Memory resident objects are quickly located using the resident object table, and non resident objects are faulted in with little additional processing. This can be contrasted with page mapping architectures of other object stores [49, 77] which have a fairly high penalty for accessing a non resident object. These systems are optimized for localized processing of a large number of small objects, where the cost of faulting a page of objects can be amortized over many access to the objects in the page. This pattern of access differs ....

Singhal, V., Kakkad, S. V., and Wilson, P. R. Texas, an efficient, portable persistent store. In Proc. of the 5th Inter. Workshop on Persistent Object Systems, pages 11--33, San Miniato, Italy, Sept. 1992.

....several reasons, the most important being the buffer force strategy used. Vagabond is based on the same philosophy as log structured file systems (LFS) which was introduced by Rosenblum and Ousterhout [15] LFS has been used as the basis for two other object managers: the Texas persistent store [16], and as a part of the Grasshopper operating system [5] Both object stores are page based, i.e. when an object has been modified, the whole page it resides on has to be written back. To our knowledge, there has been no publications on log only ODBs operating on object granularity as in the ....

V. Singhal, S. Kakkad, and P. Wilson. Texas: An efficient, portable persistent store. In Proceedings of the Fifth International Workshop on Persistent Object Systems, 1992.

....bunch. 2.5 Distribution An object may refer, using normal pointers to some other object, irrespective whether it is located in the same cluster or in another one, even when the pointed to object is located in a cluster that happens to be currently located on another node. References are swizzled [25]. An application navigating through an object structure may implicitly access a remote cluster. For instance, Figure 2 shows two clusters 4 owned by two different nodes. Starting from persistent root my root in the left most cluster, one can navigate trough the objects like e.g. my ....

....(objects created in the same transaction are allocated in the same cluster) that are prefetched and cached is determined dynamically. Caching is not cooperative. Thor has a client server architecture, provides implicit transactions and offers persistence by reachability. 4.2. 4 Texas Texas [25] is a page driven object store based on pointer swizzling at page faults. It uses the standard virtual memory hardware to detect page faults, then the requested page is fetched from disk, is swizzled and mapped into the virtual memory space of the application. It has a single user single machine ....

Vivek Singhal, Sheetal V. Kakkad, and Paul R. Wilson. Texas: An efficient, portable persistent store. In Proceedings of the 5th International Workshop on Persistent Object Systems (POS5), San Miniato, Italy, September 1992. http://cs.utexas.edu/pub/garbage.

....a file that correspond to object boundaries. HAC could also be used within a server cache. In persistent object systems, objects are clustered in fixedsize pages on disk, and pages are much larger than most objects [Sta84] Most systems manage the client cache using page caching [LLOW91, WD94, SKW92] when a miss occurs, the client evicts a page from its cache and fetches the missing object s page from the server. Page caching achieves low miss penalties because it is simple to fetch and replace fixed size units; it can also achieve low miss rates provided clustering of objects into pages is ....

V. Singhal, S. V. Kakkad, and P. R. Wilson. Texas: An efficient, portable persistent store. In 5th Int. Workshop on Persistent Object Systems (POS), pages 11--33, San Miniato, Italy, Sept. 1992.

....of persistence. We suspect that this is because the weak control of pointers in C meant that they were never able to become sufficiently simple and robust. Persistent extensions to Pascal [49, 48] only became available as Pascal was on the wane. A general purpose approach was the Texas store [191], which uses memory management techniques to manage swizzling and page faulting, and could therefore bring durable and atomic commitment to legacy C and C code. We believe that it did not investigate any of the other persistence requirements, such as evolution. A number of orthogonally ....

....is why we have kept them out of the OPJ specification until we are persuaded that they are really necessary. 12.2.2 Normal Execution Performance Ideally no cost would be paid for persistence once the working set of objects needed by an application have been faulted into main memory. Some systems [191, 207, 166, 71, 133, 101] approach this goal by utilising the memory management hardware to detect both non resident objects, and mutated objects for checkpoint purposes. It is currently the case that the cost of servicing the faults, when they do occur, can be very high with existing commercial operating systems, such ....

V. Singhal, S.V. Kakkad, and P.R. Wilson. Texas: An Efficient, Portable Persistent Store. In Albano and Morrison [5], pages 11--33.

....National University fSteve.Blackburn,Robin.Stantong anu.edu.au Abstract The paper is about design of object stores for distributed memory computers. Problems of distribution within such stores are described through the design of Distributed Texas, a system derived from the Texas store [6] and implemented on a scalable multicomputer (a Fujitsu AP1000) Problems discussed include creating a single image of an object space across multicomputer architectures, tradeoffs between page based and object based granularity, the distribution of store level services and the management of ....

....point in the performance spectrum distributed memory machines appear to be unavoidable and multicomputers provide a refined testbed for the problems that distributed memories generate. 1. 2 Structure of the Paper With the objectives outlined above in mind, the experiment was designed around Texas [6] as an example of recently developed object store technology and the Fujitsu AP1000 [5] a machine with a typical multicomputer architecture. Although coherency, concurrency and address mapping problems in scalable object stores can readily be abstracted away from particular implementations, the ....

[Article contains additional citation context not shown here]

SINGHAL, V., KAKKAD, S. V., AND WILSON, P. R. Texas: An efficient, portable persistent store. In Fifth International Workshop on Persistent Object Systems (San Miniato, Italy, Sept. 1992), MorganKaufman.

....met en oeuvre un m ecanisme de m emoire partag ee r epartie (DSM, Distributed Shared Memory) 28, 57, 80] L acc es a une donn ee n ecessite de trouver sa r ef erence par navigation en partant d une racine persistante bien connue (par ex. un serveur de noms) Dans un syst eme utilisant la PBR [34, 48, 84, 112], un objet atteignable depuis la racine persistante est lui meme persistant, il doit ainsi etre stock e. Un objet atteignable par transitivit e a partir de la racine persistante est aussi persistant. Les objets non atteignables ne sont plus n ecessaires, ils peuvent donc etre ramass es (et la ....

....disjoint, Larchant implements a distributed shared memory (DSM) mechanism [28, 57, 80] Data is replicated in multiple sites for performance and availability. Accessing data entails finding its reference by navigation from a well known persistent root (e.g. a name server) In systems with PBR [34, 48, 84, 112] a datum reachable from the persistent root is persistent, thus it should persist on permanent storage. A datum transitively reachable from the persistent root is persistent also. Unreachable data is not needed and can be reclaimed (and memory compacted) Such data is said to be garbage. ....

K. Singhal, S. Kakkad, and P. Wilson. Texas: An efficient, portable persistent store. In Proc. of the Fifth International Workshop on Persistent Object Systems Design, Implementation and Use, pages 13--28, San Miniato Pisa (Italy), September 1992.

....DBMS is an application specific, high performance DBMS that omits one or more features of a heavyweight DBMS and specializes the implementations of its features to maximize performance. Examples of LWDBs include main memory DBMSs (e.g. Smallbase [Hey95] persistent stores (e.g. Texas [Sin92]) and primitive code libraries (e.g. Booch Components [Boo87] Code Farms C Data Object Library [Cod95] Each of these examples strips features from a general purpose DBMS (e.g. Smallbase removes the databases larger than main memory feature, Texas removes client server architectures, and ....

....or log anchor at commit xact( log sync [xact : XACT, log : DS, log anchor : DS] XACT The synchronous log manager. Synchronously flush log and log anchor at commit xact( Figure 5.15 Log manager components. 13.P2 has previously been combined with the Texas uniprocess persistent store [Sin92, Jim98]. 92 decision to use operation logging does not exclude the possibility that we may implement page based logging in a future log manager. Operation logging has two fundamental problems: partial actions and action consistency [Gra93b] The partial action problem is that individual operations are ....

[Article contains additional citation context not shown here]

Vivek Singhal, Sheetal Kakkad, and Paul R. Wilson. Texas: An Efficient, Portable Persistent Store. In Persistent Object Systems: Proceedings of the 179 Fifth International Workshop on Persistent Object Systems, San Miniato, Italy, September 1992, pages 11-33. Springer-Verlag Workshops in Computing.

....storage, it is desirable, and in practice necessary, that parts of the OIDX itself can be migrated as well. This is dicult when current and historical versions reside on the same leaf pages. The composite index is used in the POST C temporal object store [16] based on the Texas object store [14]. In POST C , objects are indexed with physical OIDs, and a variant of the composite index structure is used to index historical versions. Because of the use of physical OIDs, when an object is updated in POST C , a new object is created to hold the previous version. After the previous version ....

V. Singhal, S. Kakkad, and P. Wilson. Texas: An ecient, portable persistent store. In Proceedings of the Fifth International Workshop on Persistent Object Systems, 1992.

.... systems based on log structured techniques was given in [8, 9] Logstructured file systems, whose philosophy the log only approach of Vagabond is based on, was introduced by Rosenblum and Ousterhout [11] LFS has been used as the basis for two other object managers: the Texas persistent store [12], and as a part of the Grasshopper operating system[3] Both object stores are page based, i.e. operations in the database are done on page granularity. To our knowledge, there has been no publications on other object LFS based log only OODBs. 3 The Need for a New Architecture When designing ....

V. Singhal, S. Kakkad, and P. Wilson. Texas: An efficient, portable persistent store. In Proceedings of the Fifth International Workshop on Persistent Object Systems, 1992.

....a general benchmark for DBMSs, this paper uses it to compare storage managers only. This is achieved by running the benchmark on versions of LabBase implemented on top of different storage managers, as described above. This paper compares ObjectStore (version 3.0) 32, 42] and Texas (version 0. 3) [51, 59]. Compared to relational systems, these storage managers have been used in few production applications, so this analysis is interesting in its own right. Since they are a relatively novel technology, we compare these storage managers not only in terms of performance, but also in terms of client ....

....here, we used several versions of the LabBase data server, which varied in storage management. The five versions for which we report results are: 1. OStore a version relying on ObjectStore (v3.0) 32, 42] for storage management. 2. Texas a version relying on the Texas storage manager (v0.3) [51, 59] for storage management. 3. Texas TC a version almost identical to Texas, and using the same storage manager, but with additional object clustering implemented in client code. 4. Ostore mm and Texas mm versions without any persistent storage management, and running entirely in main memory. ....

[Article contains additional citation context not shown here]

Vivek Singhal, Sheetal V. Kakkad, and Paul R. Wilson. Texas: an efficient, portable persistent store. In Proceedings of the Fifth International Workshop on Persistent Object Systems (POS-V), San Minato, Italy, September 1992. Available at ftp://cs.utexas.edu/ pub/garbage/texaspstore.ps.

No context found.

Vivek Singhal, Sheetal Kakkad, and Paul Wilson. Texas: an efficient, portable persistent store. In Persistent Object Systems: Proceedings of the Fifth International Workshop on Persistent Object Systems (San Miniato, Italy), September 1992, pages 11-33.

....has hardly been explored. Work to date has concentrated primarily on decreasing the cost of writes, or of avoiding fragmentation of the disk, but an LFS s ability to store any data anywhere on the disk opens up many possibilities for improving spatial locality, reducing read costs as well [SKW92]. Conventional memory allocators (like C s malloc( and free( also implicitly perform an important kind of clustering, simply by choosing where in memory to put objects when they are initially allocated. A good understanding of the principles of clustering could lead to the design of better ....

Vivek Singhal, Sheetal V. Kakkad, and Paul R. Wilson. Texas: an efficient, portable persistent store. In Antonio Albano and Ron Morrison, editors, Fifth International Workshop on Persistent Object Systems, pages 11--33, San Miniato, Italy, September 1992. Springer Verlag.

No context found.

V. Singhal, S. V. Kakkad, and P. R. Wilson: Texas: An Efficient, Portable Persistent Store. In A. Albano and R. Morrison, editors, Persistent Object Systems, Workshop In Computing, pages 11-33. Springer-Verlag, 1992.

No context found.

V. Singhal, S. V. Kakkad, and P. R. Wilson. Texas: An Efficient, Portable Persistent Store. In A. Albano and R. Morrison, editors, Persistent Object Systems, Workshop In Computing, pages 11-33. Springer-Verlag, 1992.

No context found.

Singhal, V., S. Kakkad, and P. Wilson, Texas: An Efficient, Portable Persistent Store, in Proc. 5th International Workshop on Persistent Object Stores. 1992. p. 11-13.

No context found.

V. Singhal, S.V. Kakkad, and P.R. Wilson. Texas: An Efficient, Portable Persistent Store. In Albano and Morrison [5], pages 11--33.

No context found.

V. Singhal, S.V. Kakkad, and P.R. Wilson. Texas: An Efficient, Portable Persistent Store. In Proc. of the 5th Intl. Workshop on Persistent Object Systems, pp. 11--33, 1992.

No context found.

Vivek Singhal, Sheetal V. Kakkad, and Paul R. Wilson. Texas: an efficient, portable persistent store. In IWPOS92 [IWPOS921992].

No context found.

V. Singhal, S. V. Kakkad, and P. R. Wilson. Texas: An E#cient, Portable Persistent Store. In Proceeding of the Fifth Int'l Workshop on Persistent Object Systems, San Miniato, Italy, pages 11#33, September 1992.

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