P. Scheuermann, J. Shim and R. Vingralek, Watchman: A data warehouse intelligent cache manager, Proc. 22nd VLDB Conf., Mumbai (Bombay), India, 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....WORST, least recently used(LRU) CLOCK, etc. The optimal policy is often approximated by LRU, LRU is further generalized into LRU k which chooses the replacement victim according to the time of the k th previous access of a cached item( CSL98] The cache replacement algorithm in WATCHMAN([SSV96]) identifies the replacement victim by considering its average reference rate, its size and execution cost of the associated query. Instead of maximizing the cache hit ratio, WATCHMAN aims at increasing the profit of the cache by minimizing the execution time of queries that miss the cache. ....

Peter Scheuermann, Junho Shim and Radek Vingralek, WATCHMAN: A Data Warehouse Intelligent Cache Manager, VLDB Conference, 1996

....and store and materialize classes in descending order of this ratio till all available space for materialized data is used up. We are however developing a more sophisticated scheme for admitting and replacing classes of materialized data, such as described for data warehouse cache management [18] where 20 40 60 80 100 Precision Percentage of queries in pattern P q=0.5 q=0.4 q=0.3 q=0.2 q=0.1 q=0 0 20 40 60 80 100 Recall Percentage of queries in pattern P q=0.5 q=0.4 q=0.3 q=0.2 q=0.1 q=0 (a) b) Precision Recall Figure 4. Effectiveness of CM Algorithm the ....

P. Scheuermann, J. Shim, and R. Vingralek. Watchman: A data warehouse intelligent cache manager. In Proceedingsof the 22nd VLDB Conference, Mumbai(Bombay), India, 1996.

....proposed in [KB96] Every cached predicate is assigned a bene t metric, which takes into account its result size, frequency of reference, and execution cost etc. The one with the smallest bene t value will then be chosen as the replacement victim. The strategy they use is very similar to that of [SSV96]. DRSN98] works in the OLAP domain, cache replacement is done at the chunk level and uses a bene t based CLOCK policy. The bene t metric mainly involves execution cost which varies from chunk to chunk in OLAP applications. Unlike these works, DFJ96] adopts several other cache replacement ....

Peter Scheuermann, Junho Shim and Radek Vingralek, WATCHMAN: A Data Warehouse Intelligent Cache Manager, Proceedings of the VLDB Conference, 1996, pp. 51-62

....is extremely useful since lots of requests are being served by transcoding it to other requested versions (very high transcoding utility) t To accommodate such a scenario, we assign a profit metric to each cached object. This profit metric is an augmented version of the metric used in WATCHMAN [9]. Each cached object, Oi, has a profit value, Pi, given by: Pi = Xici q 7iybiy si si j: All possible versions into which Oi can be transcoded. Ai: Average rate of direct reference of the object. ffij: Average rate of reference of the object, when it is referenced for transcoding to version ....

....14 on the present load conditions on the proxy and it will make sure that they stay true to the current conditions. 4.3. ALGORITHM Since the objects with lesser references have less reliable estimates of Ai and ffij, the cache replacement algorithm gives them a higher priority for eviction. As [9] suggests, we consider all objects with just one reference (Direct and Transcoding) and evict the ones with least profit scores. Then we consider the objects with two references and so on. The parameters used are (i) Size of object to be cached, ii) C: Set of objects to be replaced, and (iii) ....

Scheuermann, P., J. Shim, and R. Vingralek: 1996, 'WATCHMAN: A Data Warehouse Intelligent Cache Manager'. In: The VLDB Journal. pp. 51-62.

....but is extremely useful since lots of requests are being served by transcoding it to other requested versions (very high transcoding utility) To accommodate such a scenario, we assign a pro t metric to each cached object. This pro t metric is an augmented version of the metric used in WATCHMAN [9]. Each cached object, O i , has a pro t value, P i , given by: P i = i c i j ij b ij j: All possible versions into which O i can be transcoded. i : Average rate of direct reference of the object. ij : Average rate of reference of the object, when it is referenced for ....

....on the present load conditions on the proxy and it will make sure that they stay true to the current conditions. 4.3. Algorithm Since the objects with lesser references have less reliable estimates of i and ij , the cache replacement algorithm gives them a higher priority for eviction. As [9] suggests, we consider all objects with just one reference (Direct and Transcoding) and evict the ones with least pro t scores. Then we consider the objects with two references and so on. The parameters used are (i) S: Size of object to be cached, ii) C: Set of objects to be replaced, and (iii) ....

Scheuermann, P., J. Shim, and R. Vingralek: 1996, `WATCHMAN : A Data Warehouse Intelligent Cache Manager'. In: The VLDB Journal. pp. 51-62.

....but is extremely useful since lots of requests are being served by transcoding it to other requested versions (very high transcoding utility) To accommodate such a scenario, we assign a profit metric to each cached object. This profit metric is an augmented version of the metric used in WATCHMAN [8]. Each cached object, has a profit value, given by: 103240 5 076 8:9 ; 0 9 0 9 0 = All possible versions into which can be transcoded. Average rate of direct reference of the object. Average rate of reference of the object, when it is referenced for ....

....depend on the present load conditions on the proxy and it will make sure that they stay true to the current conditions. 4.3. Algorithm Since the objects with lesser references have less reliable estimates of , the cache replacement algorithm gives them a higher priority for eviction. As [8] suggests, we consider all objects with just one reference (Direct and Transcoding) and evict the ones with least profit scores. Then we consider the objects with two references and so on. The parameters used are (i) Size of object to be cached, ii) Set of objects to be replaced, and ....

P. Scheuermann, J. Shim, and R. Vingralek. WATCHMAN : A data warehouse intelligent cache manager. In The VLDB Journal, pages 51--62, 1996.

....caching system for OLAP queries on a dedicated infrastructure. In this work we follow a di#erent approach focusing on the client side. Continuing the previous example, assume that users from Hong Kong pose queries to the NYSE warehouse and some results are cached at their local computer [4, 5, 15, 21] hoping that subsequent queries can reuse this data. However, the size of each client s cache is relatively small compared to the size of the warehouse, while the network cost of transferring large amounts of data from overseas is high. On the other hand, it is possible that some other user in ....

....are selected once when the warehouse is set up. A dynamic approach is inspired by semantic data caching [3, 13] instead of caching a list of physical pages or tuple identifiers, the results of previous queries together with their semantic description are stored. For the special case of OLAP, [21] developed a semantic cache manager called Watchman. The system stores in the cache the results of the query together with the query string. Subsequent queries can be answered by the cached data if there is an exact match on their query strings. The authors present admission and replacement ....

[Article contains additional citation context not shown here]

P. Scheuermann, J. Shim, and R. Vingralek. Watchman : A data warehouse intelligent cache manager. In VLDB, pages 51--62, 1996.

....but is extremely useful since lots of requests are being served by transcoding it to other requested versions (very high transcoding utility) To accommodate such a scenario, we assign a profit metric to each cached object. This profit metric is an augmented version of the metric used in WATCHMAN [8]. Each cached object, 10 , has a profit value, 230 , given by: 0 5476 896 : 6 ; A 6 6 6 where, All possible versions into which can be transcoded. Average rate of direct reference of the object. Average rate of reference of the object, when it is ....

....depend on the present load conditions on the proxy and it will make sure that they stay true to the current conditions. 4.3. Algorithm Since the objects with lesser references have less reliable estimates of , the cache replacement algorithm gives them a higher priority for eviction. As [8] suggests, we consider all objects with just one reference (Direct and Transcoding) and evict the ones with least profit scores. Then we consider the objects with two references and so on. The parameters used in the algorithm are: Size of object to be cached. This is the minimum amount of ....

P. Scheuermann, J. Shim, and R. Vingralek. WATCHMAN : A data warehouse intelligent cache manager. In The VLDB Journal, pages 51--62, 1996.

....The same file after it is read into cache, is also referenced by different users. Some earlier works on file caching in distributed systems and the staging of files from tertiary storage have been presented in [8, 10, 14, 15, 16, 18] Recent studies on caching have focused more on webcaching [1, 3, 5, 12, 17]. Cao and Irani [3] present a relative comparison of various cache replacement policies that have been proposed for web caching. Their work discusses some of the merits and concerns 3 Network Site D HRM HRM Hierarchical Storage System Site C Clients Workgroup Server Clients Acessing Data ....

....function defined in the preceding subsection. Restated differently, we have that whenever a file in the cache needs to be evicted at time t, the eviction candidate is the one that has the minimal utility function # #t# given by p # #t# # c # #t# (5) Similar conclusions have been reached in [10, 12] but under different assumptions. The problems studied for which they derived their results were different. The utility function as expressed by equation (5) is not practical to apply since we do not know the probabilities and even more, these probabilities are not stationary. Under the ....

P. Scheuermann, J. Shim, and R. Vingralek. Watchman: A data warehouse intelligent cache manager. In Proc. 22nd VLDB Conference, pages 51 -- 62, Bombay, India, Sept. 1996.

....these algorithms is proposed, to select both views and indices on them. 1] employs a method which identifies the relevant views of a lattice for a given workload. 22] uses a simple and fast algorithm for selecting views in lattices with special properties. Dynamic alternatives are exploited in [18, 3, 11]. These systems reside between the data warehouse and the clients and implement a disk cache that stores aggregated query results in a finer granularity than views. Most of these papers assume that the OLAP queries are sent to the system one at a time. Nevertheless, this is not always true. In ....

Scheuermann P., Shim J., Vingralek R., "WATCHMAN: A Data Warehouse Intelligent Cache Manager", Proc. VLDB, 1996.

....consequences. The view selection problem aims at balancing the trade off between performance improvement and maintenance overhead because of materialization. View selection has been studiedextensively in relational databases and data warehouses [Rou82, Han87, Sel88, BM90, GM95, RCK 95, SSV96, Gup97, CD97, KR99] In data warehouses, View Selection is performed off line during the down time of the warehouse. Web servers, on the other hand, must remain online all the time and thus, updates are applied in the back end database while the web server continues to serve user requests. ....

....Response Time Figure 15: QoS over time 5 Related Work To the best of our knowledge this is the first paper that attempts to solve the Online View Selection problem. There is a lot of research on the off line version of the view selection problem in Data Warehousing literature [Sel88, GM95, SSV96, Gup97, GM99, KR99] Materialization has also been explored in [YFIV00] where multiple maintenance policies were supported, but the selection problem was not addressed. LR00] presented a cost model for the view selection problem, but did not provide an algorithm. CIW 00] described a system ....

Peter Scheuermann, Junho Shim, and Radek Vingralek. "WATCHMAN : A Data Warehouse Intelligent Cache Manager". In Proc. of the 22nd VLDB Conference, Bombay, India, September 1996.

....results. Due to the interdependencies among OLAP queries, keeping semantic information about the cache contents leads to better utilization of the cache than page caching. Some early work on semantic caching for general SQL queries appears in [DFJ 96, KB96] For the special case of OLAP, [SSV96] developed a cache manager called watchman. Their system stores in the cache the results of the query, together with the query string. Subsequent queries can be answered by the cached data if there is an exact match on their query strings. The authors present admission and replacement algorithms ....

....be evicted. Traditional page replacement algorithms such as LRU and LFU are not suitable in the OLAP environment since they don t consider the cost differences to calculate results at different levels of aggregation, and the varying sizes of the results. A more suitable alternative is presented in [SSV96]. Let f(v) be the frequency of queries that are answered exactly by view v.Letsize(v) be the size of v and cost(v) be the cost to calculate v.Thegoodness of v is defined as: cos ) v size v t v f v goodness = The view with the lowest goodness is the first to be evicted. The ....

Scheuermann P., Shim J., Vingralek R., "WATCHMAN: A Data Warehouse Intelligent Cache Manager", Proc. VLDB, 1996.

No context found.

P. Scheuermann, J. Shim, and R. Vingralek. WATCHMAN: A data warehouse intelligent cache manager. In Proceedings of the International Conference on Very Large Databases, 1996.

No context found.

P. Scheuermann, J. Shim and R. Vingralek, Watchman: A data warehouse intelligent cache manager, Proc. 22nd VLDB Conf., Mumbai (Bombay), India, 1996.

No context found.

SCHEUERMANN, P., SHIM,J.,AND VINGRALEK, R. 1996. WATCHMAN: A data warehouse intelligent cache manager. In Proceedings of the 22th VLDB Conference (Bombay, India, Sept.), pp. 51--62.

No context found.

Peter Scheuermann, Junho Shim, and Radek Vingralek. Watchman : A data warehouse intelligent cache manager. In Proceedings of 22th International Conference on Very Large Data Bases (VLDB), pages 51--62, 1996.

No context found.

P. Scheuermann, J. Shim, R. Vingralek, WATCHMAN : A Data Warehouse Intelligent Cache Manager, VLDB Conf., Bombay, 1996.

No context found.

P. Scheuermann, J. Shim, and R. Vingralek. WATCHMAN: A Data Warehouse Intelligent Cache Manager. In VLDB Journal, pages 51--62, 1996.

No context found.

P. Scheurmann, J. Shim, and R. Vingralek. WATCHMAN: A Data Warehouse Intelligent Cache Manager. Proc. of the 22nd VLDB Conf., pp. 51-62, September 1996.

No context found.

P. Scheuermann, J. Shim, and R. Vingralek, "WATCHMAN: A Data Warehouse Intelligent Cache Manager," Proc. 22nd VLDB Conf., pp. 51-62, Sept. 1996.

No context found.

P. Scheuermann, J. Shim, and R. Vingralek, "WATCHMAN: A Data Warehouse Intelligent Cache Manager," Proc. 22nd VLDB Conf., pp. 51-62, Sept. 1996.

No context found.

P. Scheuermann, J. Shim, and R. Vingralek. WATCHMAN: A data warehouse intelligent cache manager. In Proceedings of the 22nd VLDB Conference, pages 51--62, Mumbai, India, September 1996.

No context found.

P. Scheuermann, J. Shim and R. Vingralek, Watchman: A data warehouse intelligent cache manager, Proc. 22nd VLDB Conf., Mumbai (Bombay), India, 1996.

No context found.

Peter Scheuermann, Junho Shim, Radek Vingralek, "WATCHMAN, A Data Warehouse Intelligent Cache Manager", 51-62, Proc. VLDB'96, 1996.

No context found.

Peter Scheuermann, Junho Shim, Radek Vingralek, "WATCHMAN, A Data Warehouse Intelligent Cache Manager", 51-62, Proc. VLDB'96, 1996.

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