Franklin, M.J., Carey, M.J. and Livny, M. "Global Memory Management in Client-Server DBMS Architectures," In Proc. of the InternationalConference on Very Large Data Bases, pp. 596--609, August 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....client server environment are usually page based, as they try to exploit spatial locality assuming that clustering of tuples to pages is effective. Page level caching assumes that only pages from the base relations are stored in the cache. Lots of work has been done in page level caching, [FCL92], FCL93] FC94] Fra96] FCL97] etc. In the work of [DM90] the caching mechanism with a granularity of a single object is implemented and compared with a page level caching. CSL98] investigates three different levels of granularity of caching, namely, attribute caching which caches ....

Michael J. Franklin, Michael J. Carey and Miron Livny, Global Memory Management in ClientServer DBMS Architectures, Proceedings of VLDB Conference, 1992

.... spatial distance based on the distance between the user and objects is included for cache replacement algorithm in MRM (Most Required Movement) 3] Compared to a traditional LRU replacement scheme, MRM shows the better performance in the situation where the strong spatial coherence is existent [4] in applications like a virtual walkthrough. Since this method ignores individual user s interest in objects, it does not support efficient cache replacement and prefetching adapting various behavioral pattem of users, as the number of users increases. QUICK proposes a task dependent importance ....

M. Franklin, M. Carey, and M. Livny, "Global Memory Management in Client-Server DBMS Architectures," in Proceedings of the International Conference on Very Large Database, pages 596-609, 1992.

....Work The design of Dynamo is motivated by several systems such as NASD [6] which employs third party transfer of data from network attached storage devices directly to client machines. Initial work on a more scalable approach to client server architectures was done by Franklin, Carey and Livny [5]. The serverless network file system (xFS) was developed at the University of California at Berkeley [1] as part of the NOW (Network of Workstations) project to address the problem of a highly scalable file system in a distributed environment using cooperative caching. There are two research ....

M. Franklin, M. Carey, and M. Livny. Global memory management in client-server DBMS architectures. Proc. of the 18th VLDB Conference, 1992.

....directly 28 J. YANG, ET. to client toochines. In the remoinder of this section, we will cornpore Dynomo with reloted file system orchitectures, ond cooperotive coching strotegies. Initiol work on o more scoloble opprooch to client server orchitectures wos done by Fronklin, Corey ond Livny [11]. Here, the server observes whot the clients ore doing ond uses the foct thor o poge might be in some client s moin memory so thor the server is oble to serve other clients requests directly from o client s coche insteod of its own limited buffer or disk resources. The serverless network file ....

. M. Franklin, M. Carey, and M. Livny. Global memory management in client-server DBMS architectures. P,oc, of the 18th VLDB Confc,ence, 1992.

....42] Cooperating nodes use approximate knowledge of the global memory state to make caching and ejection decisions that benefit a page faulting client and the whole cluster. Perhaps the closest work to ours in spirit is a global memory management protocol developed for database management systems [14]. Here, the database server keeps a directory of pages in the aggregate cache. This directory allows the server to forward a page request from one client to another that has the data, request that a client demote rather than discard the last in memory copy of a page, and preferentially discard ....

M. J. Franklin, M. J. Carey, and M. Livny. Global memory management in client-server DBMS architectures. In Proc. of the 18th Very Large Database Conf., pages 596--609. Aug. 1992.

....collective memory on all client machines is usually much larger than that on server machines, the cache hit rate can be improved, and disk I O avoided. However, the cooperative cache management in Dynamo is different from existing cooperative caching systems such as [And96] Cao95] Cao96] Fee95] [Fra92] [voe97] Voe98] In most of these systems, a client wanting to replace data in its own cache first tries to place the data in the cache of other nodes if space is existing at these nodes. If not, the data will be discarded or written back to disk. In determining the candidate remote node to which ....

....dynamically migrating functionality to other machines. In the remainder of this section, we will compare Dynamo with related file system architectures, and cooperative caching strategies. Initial work on a more scalable approach to client server architectures was done by Franklin, Carey and Livny [Fra92]. Here, a server was introduced that extensively used the available client cache for keeping data pages in memory so that the server was able to service other clients requests directly from a client cache instead of its own limited buffer or disk resources. The serverless network file system ....

M. Franklin, M. Carey, and M. Livny. Global memory management in client-server DBMS architectures. Proc. of the 18th VLDB Conference, 1992.

.... spatial distance based on the distance between the user and objects is included for cache replacement algorithm in MRM (Most Required Movement) 3] Compared to a traditional LRU replacement scheme, MRM shows the better performance in the situation where the strong spatial coherence is existent [4] in applications like a virtual walkthrough . Since this method ignores individual user s interest in objects, it does not support efficient cache replacement and prefetching adapting various behavioral pattern of users, as the number of users increases. QUICK proposes a task dependent importance ....

M. Franklin, M. Carey, and M. Livny, Global Memory Management in Client-Server DBMS Architectures, in Proceedings of the International Conference on Very Large Database, pages 596-609, 1992.

....or incorrect, a client can always retrieve a block from the server. Using hints, block migration is done in a manner similar to that of GMS [13] Unlike MBFS, none of the above systems considers a client s CPU or memory load when deciding the movement or replacement of pages. Franklin et al. [12] use remote memory to cache distributed database records and move data around using an algorithm similar in nature to that of Nchance forwarding. Client load was not considered by the data migration mechanism. The Trapeze network interface [2] provides an additional order of magnitude improvement ....

M. Franklin, M. Carey, and M. Livny. Global Memory Management in Client-Server DBMS Architectures. In 18th International Conference on Very Large Data Bases, 1992.

....reduce the demand on I O bandwidth. If the average length of the video chains is l, this scheme requires a video bandwidth which is nearly l times less than that of a conventional system. We note that a technique similar to Chaining in concept was proposed by Franklin, Carey and Livny [7] for conven tional client server computing. In this scheme, a client workstation is allowed to retrieve data cached in other workstations which have the latest version of the data items. The similar studies on the operating systems design were discussed in [3, 6] Their works, however, are ....

M. Franklin, M. Carey, and M. Livny. Global memory management in client-server DBMS architectures. In Proc. of VLDB Conf., pp. 596--609, Aug. 1992.

....of clients increases. They utilize inter object spatial relations and more appropriate communication protocols (such as multicast) to reduce the amount of data for transmission[12] Some other studies consider the data management problem in a large scale world from a client s point of view[5] 6][9]. This is also the focus of this paper. Determining the set of visible objects at the current viewpoint in a large 3D virtual world is usually carried out in a preprocessing step due to the computation complexity involved. In [14] a k D tree is used to partition the world into cells connected ....

....more relevant data for faster accesses. Many replacement policies for caching have been proposed but no universally best policy is found. The well known LeastRecently Used (LRU) policy is a good one in general for data with temporal locality. However, it does not perform as well for spatial data[9]. In [5] an effective cache replacement policy, called Most Required Models (MRM) was introduced, based on multiple resolutions of an object that can be transmitted progressively. A good prefetching strategy can help system performance on the smoothness of user navigation. Most previous systems ....

M. Franklin, M. Carey, and M. Livny, "Global Memory Management in Client-Server DBMS Architectures, " in Proceedings of the International Conference on Very Large Database, pp596-609, 1992.

....over the basic cooperative caching mechanism and provides better performance when there is fine grained sharing. 1 Introduction This paper describes how to integrate two techniques that have been shown to improve performance in distributed client server storage systems: cooperative caching [8, 11, 12, 20] and efficient support for fine grained sharing [1, 4, 6, 9, 10, 15] Cooperative caching combines client caches so that misses at one client can be serviced from caches of other clients. It takes advantage of fast local area networks to reduce access latency, replacing slow disk accesses by ....

....of what we have accomplished. 2 Related Work Our proposal builds on previous work on cooperative caching and support for fine grained sharing. Cooperative caching has been studied in several contexts: distributed virtual memory [11] file systems [8, 7, 20] and a transactional database [12]. All studies show that cooperative caching can reduce access latency and improve system scalability in workloads with coarse grained sharing. These studies are complementary to ours; we explain how to extend the benefits of cooperative caching to workloads with fine grained sharing. We can use ....

[Article contains additional citation context not shown here]

M. Franklin, M. Carey, and M. Livny. Global Memory Management in Client-Server DBMS Architectures. In VLDB, 1992.

....they only used the memory of idle workstations. The xFS file system [Dahlin et al. 94] caches file data at clients. In this system the server coordinates data transfers between clients. A similar idea applied to an object oriented database server was implemented at University of Wisconsin Madison [Franklin et al. 92] Their server coordinates the use of peer clients active buffering regions to service database buffer misses. They used specialized page management at clients to manage coherency, and at the database server to improve hit rates. Clients send data to and receive data from the server, which may ....

Michael J. Franklin, Michael J. Carey, and Miron Livny. Global Memory Management in ClientServer DBMS Architectures. In Proceedings of the 18th International Conference on Very Large Data Bases, pages 596--609, Vancouver, BC, Canada, August 1992.

....on a parallel hardware make use of aggregate 17 memory. Efficiently managing and exploiting the aggregate memory is left to the application. It is not clear how database applications use such a distributed shared memory systems for buffer management. 1.3. 3 Utilizing Idle Memory Franklin et al. FCL92, FCL93] study ways to augment the memory of the server in a client server database system by utilizing the memory and disk resources on client workstations. The key idea behind their work is that, the server uses the data cached in one client s memory to serve the request of another client. This ....

....up the directory entry, and request the page to be sent to the requesting server, if it is present in memory at some other server. The technique in option 2 is known as forwarding. Forwarding has been shown to make good use of data already present in global memory in studies by Franklin et al. FCL92] and Dahlin et al. DWAP94] We therefore use option 2, forwarding, instead of always reading the page from disk. With forwarding, the owner reads the page from disk only if the page is absent in global memory. When the owner reads the page from disk and sends it to the primary server, it ....

[Article contains additional citation context not shown here]

M. Franklin, M. Carey, and M. Livny. Global Memory Management in Client-Server DBMS Architectures. In Proceedings of the 18th International Conference on Very Large Data Bases, Vancouver, British Columbia, Canada, August 1992.

....reduce the amount of data requested over the network for faster response time. This can be achieved by caching and prefetching mechanisms. A caching mechanism allows a client to utilize its memory and local storage to cache currently visible objects that are likely to be visible in the near future [11]. A prefetching mechanism allows a client to predict objects that will be visible in the future and obtain the objects in advance to improve response time. A good caching mechanism should retain objects with high affinity while a good prefetching mechanism should predict those objects which will ....

....cache has an advantage of persistence. When disconnected from the server, a client can still operate on the database objects in its local storage. In a conventional client server database environment, data objects are usually transferred from the database server to a client on a per page basis [2, 11]. This is primarily because the server s storage is also page based. The overhead for transmitting one item or a page is similar. In a virtual walkthrough environment, virtual objects are represented using object models and are usually very complex and large in size, occupying possibly multiple ....

[Article contains additional citation context not shown here]

M. Franklin, M. Carey, and M. Livny. Global Memory Management in Client-Server DBMS Architectures. In Proceedings of International Conference on Very Large Databases, pages 596--609, 1992.

....discussed in [24] show that the page server architecture, in the form described above, yields performance superior to both file server and object server architectures, provided a good data clustering scheme is in use. In the RAD UNIFY CSD, there is no interaction between the clients. In [27], retrieval of information from other clients caches is presented as a way to augment the local cache. By adding client caches, CSDs follow the trend in building global memory hierarchy systems [51] This makes the volume of data available in memory buffers (other than in the server s cache) ....

....at any client is the server s disk accessed. A number of algorithms have been developed that allow this method to be used to reduce the server load without affecting data consistency in the database, as well as maximizing the amount of data that is available for retrieval from global memory. [27] indicates that this configuration is best suitable for environments where there is low to medium data sharing. 20 The notion of the Enhanced CSD and the use of client s disks described earlier can be extended in the page server environment. A proposal along these lines appears in [28] There ....

[Article contains additional citation context not shown here]

M. Franklin, M. Carey, and M. Livny. Global Memory Management in Client-Server DBMS Architectures. In Proceedings of the 18th International Conference on Very Large Data Bases, Vancouver, Canada, August 1992. 45

.... scalability, enhanced fault tolerance and better dynamic load balancing [15] 16] A significant amount of research effort has investigated various issues for such data sharing environments, or related client server paradigms, including the impact of data skew [4] global buffer management [1] 5] [7], locking, and concurrency and coherency protocols [6] 11] 14] As pointed out in [4] a skewed input, while helping the buffer hit ratio, will increase the cross invalidation effect as well as the level of lock contention, and thus has various impacts on system performance. As a result, several ....

M. J. Franklin, M. J. Carey, and M. Livny. Global Memory Management in Client Server DBMS Architectures. Proceedings of the 18th International Conference on Very Large Data Bases, pages 596-- 609, August 1992.

....explores global caching in a scalable serverless storage architecture. The xFS coherence protocol improves on earlier file level coherence protocols by supporting disk block level sharing. In contrast, the fragment reconstruction protocol supports sharing of fine grained objects. Franklin et al. [7] studies global caching in a client server database. Clients interact with each other via servers. Servers redirect fetch requests between clients. The study evaluates several global memory management algorithms. This work is similar to ours in that it considers a transactional database. The ....

M. Franklin, M. Carey, and M. Livny. Global memory management in client-server dbms architectures. In Proceedings of 18th VLDB Conf., 1992.

.... 9] have explicitly addressed the issue of Web cache filter effects, wherein a higher level cache in a multi cache system only handles requests that miss in the lower level cache(s) Similar cache design problems have been addressed previously in the context of CPU cache hierarchies [14] databases [10], and client server systems [16] Among the papers that focus on Web cache filter effects, most focus on the frequency domain aspect of the Web cache filter effect. Doyle et al. 9] refer to this as the trickle down effect, and conduct a detailed simulation study to quantify its impact. Che ....

M. Franklin, M. Carey, and M. Livny, "Global Memory Management for Client-Server DBMS Architectures ", Proceedings of the 19th International Conference on Very Large Databases (VLDB), August 1992.

....the idea is to apply the expensive Pro#table policy only if this is necessary and otherwise rely on the cheap Reference Counting policy. Furthermore, weintend to study how the policies presented in this work can be combined with techniques for global memory management such as those proposed in #Franklin et al. 1992#; here, the goal is to make good caching decisions for all #or a large group of# clients rather than for every client individually. Finally,wewould like to study cache investment in the presence of proxy caching or more general caching hierarchies. Acknowledgments We would liketothank Bj#orn ....

Franklin, M., Carey, M., and Livny, M. 1992. Global memory management in clientserver DBMS architectures. In Proc. of the VLDB Conf. #Vancouver, Canada, 1992#, pp. 596#609.

No context found.

Franklin, M.J., Carey, M.J. and Livny, M. "Global Memory Management in Client-Server DBMS Architectures," In Proc. of the InternationalConference on Very Large Data Bases, pp. 596--609, August 1992.

No context found.

M. Franklin, M. Carey, and M. Livny, Global memory management in client-server DBMS architectures, in Proceedings of the 18th VLDB Conference, August 1992, pp. 596--609.

No context found.

M. J. Franklin, M. J. Carey, and M. Livny. Global memory management in client-server DBMS architectures. In International Conference On Very Large Data Bases (VLDB '92), pages 596--609, San Mateo, Ca., USA, August 1992. Morgan Kaufmann Publishers, Inc.

No context found.

M. Franklin, M. Carey, and Livny M. Global memory management in client-server DBMS architectures. In Proceedings of the Eighteenth International Conferenceon Very Large Databases, Vancouver, British Columbia, pages 596#609, August 1992.

No context found.

M. Franklin, M. Carey, and Livny M. Global memory management in client-server DBMS architectures. In Proceedings of the Eighteenth International Conference on Very Large Databases, Vancouver, British Columbia, pages 596#609, August 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