C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet. AlphaSort: A RISC Machine Sort. In Proc. of the Sigmod Conference, pages 233--242, 1994.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... group by l return 7 replacement selection because it has good adaptive behavior in the presence of changing memory conditions, allows straightforward implementation of record aggregation and only requires memory proportional to the output of the aggregation or projection, not the input relation [Nyberg94]. A solution using hashing would have similar benefits given the appropriate choice of hash functions to match the skew within a particular set of data. 3.3. Joins Selective joins will benefit significantly from a reduction in data transfer by operating directly at the drives, and from the ....

Nyberg, C., Barclay, T., Cvetanovic, Z., Gray, J. and Lomet, D. "AlphaSort: A RISC Machine Sort" SIGMOD, May 1994. 21

....load balancing and overlapping of I O and computation has been intensively studied [21, 7, 3, 14, 13, 12] But we have not seen results that guarantee overlapping of I O and computation during parallel disks merging of arbitrary runs. There are many good practical implementations of sorting (e.g. [19, 1, 30, 20]) that address parallel disks, overlapping of I O and computation, and low internal overhead. However, we are not aware of fast implementations that give theoretical performance guarantees on achieving asymptotically optimal I O. Most practical implementations use a form of striping that ....

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet. AlphaSort: A RISC machine sort. In SIGMOD, pages 233--242, 1994.

....the reader to Graefe s survey [12] for an overview of database query processing techniques and highlight the most relevant work on parallel query processing. Early work concentrated on parallelizing individual, traditional content sensitive operators like hybrid hash join [25] and sort (e.g. [10, 20, 1]) The abstractions which inspired Flux, Exchange [11] and RiverDQ [23] were proposed to compose such operators into a dataflow. Shatdal and Naughton [27] describe how to leverage shared virtual memory across a shared nothing cluster to implement hybrid hash join. DeWitt et al. present practical ....

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. B. Lomet. AlphaSort: A RISC Machine Sort. In SIGMOD, 1994.

....direction. We refer the reader to [13] for an overview of database query processing techniques and highlight the most relevant work on parallel query processing. Early work concentrated on parallelizing individual, traditional content sensitive operators like hybrid hash join [25] and sort (e.g. [11, 20, 1]) The abstractions which inspired Flux, Exchange [12] and RiverDQ [23] were proposed to compose such operators into a dataflow. In [10] and [9] the authors present practical techniques for handling data skew for a hash join and external sort, respectively. These techniques rely on sampling a ....

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. B. Lomet. AlphaSort: A RISC Machine Sort. In SIGMOD, 1994.

....has diminished, we believe the benchmark remains relevant for three reasons. First, Datamation stresses the importance of start up time. Start up time has been noted as one of three factors limiting performance of parallel systems [7] and has been a problem for both clusters [2] as well as SMPs [8]. Second, Datamation is an example of an interactive parallel application. For parallelism to become commonplace, interactive applications must become a reality. Third, interactive parallel # This technical report describes work completed in the spring of 2001. 0.1 1 10 100 1000 1986 1988 ....

Chris Nyberg, Tom Barclay, Zarka Cvetanovic, Jim Gray, and Dave Lomet. AlphaSort: A RISC Machine Sort. In Proceedings of 1994.

....are introduced [11, 31, 34] their behaviors and properties will likely become even more divergent than they are today. Although this set of devices is disparate, one commonality pervades them all: the time to access them is high, especially as compared to CPU cache and memory latencies [23]. Due to the cost of fetching blocks from storage media, caching of blocks in main memory often reduces execution time of individual applications and increases overall system performance often by orders of magnitude. However, while storage technology has dramatically changed over the past few ....

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet. AlphaSort: A RISC Machine Sort. In 1994.

....steps are involved: Key creation: A key is created by scanning through the database and picking part of the elds or entire elds. The consonants are usually picked up. Typically the key may be made of 3 letters from each eld. One of the elds is taken as the primary eld. Fig. 2) Sorting[Nyb94]: Any one of the good sorting methods like quick sort or merge sort can be used to sort the data set, based on the key created in the previous step. 11 Figure 2: Key creation in Sorted Neighborhood method Figure 3: Merging of records using sliding window Merge: A window of size w is made to ....

Nyberg C. , Barclay T., Cvetanovic Z. , Gray J. and Lomet D. AlphaSort: A RISC Machine Sort. In Proceedings of the ACM-SIGMOD Conference, pages 233-242, 1994.

....are introduced [11, 31, 34] their behaviors and properties will likely become even more divergent than they are today. Although this set of devices is disparate, one commonality pervades them all: the time to access them is high, especially as compared to CPU cache and memory latencies [23]. Due to the cost of fetching blocks from storage media, caching of blocks in main memory often reduces execution time of individual applications and increases overall system performance often by orders of magnitude. However, while storage technology has dramatically changed over the past few ....

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet. AlphaSort: A RISC Machine Sort. In

....has diminished, we believe the benchmark remains relevant for three reasons. First, Datamation stresses the importance of start up time. Start up time has been noted as one of three factors limiting performance of parallel systems [7] and has been a problem for both clusters [2] as well as SMPs [8]. Second, Datamation is an example of an interactive parallel application. For parallelism to become commonplace, interactive applications must become a reality. Third, interactive parallel jobs are particularly sensitive to performance fluctuations that occur in large scale systems [3] Thus, if ....

Chris Nyberg, Tom Barclay, Zarka Cvetanovic, Jim Gray, and Dave Lomet. AlphaSort: A RISC Machine Sort. In Proceedings of

....a pointer to the record on disk. Memory usage considers the operating system (OS) memory overhead, around 38 17 in practical cases. The exact overhead depends on the total memory, measured with 64 256Mbytes. Around 20 of the time to run the sort algorithm is due to CPU time, when using quicksort [3]. The approach proposed here is based on two technological advances. First, there is a trend to standardize the communication between the main processor and peripherals in Personal Computers and or UNIX Workstations. A good example is the PCI bus standard. In our approach, one or more peripherals ....

C. Nyberg et alii. AlphaSort: A RISC Machine Sort. In: Proceedings of

....extracted from the USGS GNIS [USGS95] and Land Use 5 Our bulk load routine uses the standard technique of extracting key,TID pairs from the base table, sorting the pairs into index leaf pages and then building the rest of the tree bottom up. Our external sorting routine follows the recent trend [DEWI91, GRAE92,NYBE94] toward quicksort based run generation. 11 Index Base Table (Heap) Type Cardinality Size Distributions (tuples) bytes) clustered unclustered B tree 10 4 10 6 sorted random 10 5 10 7 sorted random 10 6 10 8 sorted random R tree 1. 4 10 5 1. 3 10 7 Hilbert (H 22 ) ....

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray and D. Lomet, "AlphaSort: A RISC Machine Sort," Proc. 1994 ACMSIGMOD Conf. on Management of Data, Minneapolis, MN, May 1994, 233-242.

.... 50 or more of execution time is often wasted due to SRAM cache misses [1, 2, 10, 18] For main memory databases, it is even clearer that SRAM cache performance is crucial [19] Hence several recent studies have revisited core database algorithms in an effort to make them more cache friendly [5, 17, 19, 20, 21]. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy ....

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet. AlphaSort: A RISC Machine Sort. In Proceedings of the 1994 SIGMOD Conference, pages 233--242, May 1994.

.... create EFS 1 thread Figure 2. Read Throughput. Figure 3. Write Create Throughput. whereas the single threaded cases are not. 7.2. Database Sort Benchmark Results Using XFS, Silicon Graphics recently achieved record breaking performance on the Datamation sort [Anon85] and Indy MinuteSort [Nyberg94] benchmarks. The Datamation sort benchmark measures how fast the system can sort 100 megabytes of 100 byte records. The MinuteSort benchmark measures how much data the system can sort in one minute. This includes start up, reading the data in from disk, sorting it in memory, and writing the sorted ....

Nyberg, C., Barclay, T., Cvetanovic, Z., Gray, J., Lomet, D., "AlphaSort: A RISC Machine Sort," Proceedings of the 1994 SIGMOD International Conference on Management of Data, Minneapolis, 1994.

....at an average of N K P 2 bytes per message, where K is the record size in bytes. Each processor sends P 1 messages, so the total network trac is approximately N K bytes. The bucket sort utilizes two bu ers on each node to hold data before and after sorting. It has been shown by Nyberg et al. [14] that sorting via pointers is considerably faster than sorting large records; therefore, our implementation uses an additional bu er of pointers on which the sort is performed. With a uniform distribution each processor s bucket is assured to be N K P size. This leads to a total memory ....

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet, \AlphaSort: A RISC Machine Sort," in Proceedings of 1994 ACM SIGMOD Conference, May 1994.

No context found.

Nyberg, C., T. Barclay, Z. Cvetanovic, J. Gray, D. Lomet, "AlphaSort: A RISC Machine Sort", Proceedings of the 1994.

No context found.

Nyberg, C.T., Barclay, T.D., Cvetanovic, Z.Z., Gray, J.N., Lomet, D.L., "AlphaSort - A RISC-machine Sort", Submitted to SIGMOD 94.

No context found.

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet. AlphaSort: A RISC Machine Sort. In 1994.

....products On a standard sort benchmark, Nsort running on an Origin2000 system is an order of magnitude faster than its competition. In a separate test, Nsort sorted a terabyte of data in 2.5 hours. The next section describes these results. MinuteSort MinuteSort is a standard sorting benchmark [3]. The benchmark measures the number of 100 byte records that can be sorted in one minute of elapsed time. The input records have 10 byte random keys. The minute limit includes the time to: read the input le sort the data create and write the output le There are two categories for the MinuteSort ....

.... time to: read the input le sort the data create and write the output le There are two categories for the MinuteSort benchmark: Indy (a custom, benchmark special sort program) and Daytona (a commercial, general purpose sort program) The rst winner of the Indy MinuteSort benchmark was AlphaSort [3], a sort program designed to show that RISC processors could be used for high performance sorting. It used striped input and output les to achieve high bandwidth disk i o one input le and output le on each disk. AlphaSort also demonstrated that judicious use of processor caches is crucial to ....

Nyberg, C., T. Barclay, Z. Cvetanovic, J. Gray, D. Lomet, "AlphaSort: A RISC Machine Sort", Proceedings of the

No context found.

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet. AlphaSort: A RISC Machine Sort. In Proc. of the Sigmod Conference, pages 233--242, 1994.

No context found.

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet. AlphaSort: A RISC Machine Sort. In Proc. of the Sigmod Conference, pages 233--242, 1994.

No context found.

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet. AlphaSort: A RISC machine sort. In SIGMOD, pages 233--242, 1994.

No context found.

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet. AlphaSort: A RISC Machine Sort. In Proceedings of the 1994.

No context found.

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet. Alphasort: A RISC machine sort. In SIGMOD, pages 233--242, 1994.

No context found.

C. Nyberg, T. Barclay, Z Cvetanovic, J. Gray, and D. Lomet. AlphaSort: A RISC Machine Sort. In Proc. of the ACM SIGMOD, pages 233-- 242, Minneapolis, May 1994.

No context found.

C. Nyberg, T. Barclay, Z. Cvetanovic, J. Gray, and D. Lomet. AlphaSort: A RISC Machine Sort. In Proc. of the ACM SIGMOD Int'l. Conf. on Management of Data, pages 233--242, Minneapolis, MN, USA, May 1994.

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