【正文】 s partitioning scheme unless the index is explicitly partitioned differently. When a table and its indexes are aligned, then SQL Server can move partitions in and out of partitioned tables more effectively, because all related data and indexes are divided with the same algorithm. When the tables and indexes are defined with not only the same partition function but also the same partition scheme, they are considered to be storageal。 however, you will get better performance when you have multiple disks and, preferably, multiple CPUs. When the scheme is used with a table, you will define the column that is used as an argument for the partition function.For range partitions, the dataset is divided by a logical and datadriven boundary. In fact, the data partitions may not be truly balanced. Data usage dictates a range partition when the table is used in a pattern that defines specific boundaries of analysis (also known as ranges). The partition key for a range function can consist of only one column, and the partitioning function will include the entire domain, even when that data may not exist within the table (due to data integrity/constraints). In other words, boundaries are defined for each partition but the first partition and the last partition will, potentially, include rows for the extreme left (values less than the lowest boundary condition) and for the extreme right (values greater than the highest boundary condition). Therefore, to restrict the domain of values to a specific dataset, partitions must be bined with CHECK constraints. Using check constraints to enforce your business rules and data integrity constraints allows you to restrict the dataset to a finite range rather than infinite range. Range partitions are ideal when maintenance and administration requires archiving large ranges of data on a regular basis and when queries access a large amount of data that is within a subset of the ranges.Index PartitioningIn addition to partitioning a table39。2005, tables and indexes can be created with the same syntax as previous releases—as a single tabular structure that is placed on the DEFAULT filegroup or a userdefined filegroup. Additionally, in SQL Server 2005, table and indexes can be created on a partitioning scheme. The partitioning scheme maps the object to one or more filegroups. To determine which data goes to the appropriate physical location(s), the partitioning scheme uses a partitioning function. The partitioning function defines the algorithm to be used to direct the rows and the scheme associates the partitions with their appropriate physical location (., a filegroup). In other words, the table is still a logical concept but its physical placement on disk can be radically different from earlier releases。s structure with regard to creation and placement. In previous releases, a table was always both a physical and a logical concept, yet with SQL Server2005, you must be familiar with a few new concepts, terms, and syntax. To understand these new concepts, let39。GB of old data requires rowbyrow manipulation of the table, as well as its associated indexes. The process of deleting data creates a significant amount of log activity, does not allow log truncation for the duration of the delete (note that the delete is a single automit transaction。2005 the solution is more elegant: you can simply switch in the newly filled partition(s) as an extra partition of the existing partition scheme and switch out any old partition(s). From end to end, the process takes only a short period of time and can be made more efficient by using parallel bulk loading and parallel index creation. More importantly, the partition is manipulated outside of the scope of the table so there is no effect on the queried table until the partition is added. The result is that, typically, adding a partition takes only a few seconds. The performance improvement is also significant when data needs to be removed. If one database needs a slidingwindow set of data in which new data is migrated in (for example, the current month), and the oldest data is removed (maybe the parallel month from the previous year), the performance of this data migration can be improved by orders of magnitude through the use of partitioning. While this may seem extreme, consider the difference without partitioning。 however, these can directly affect all other users, and sacrifices concurrency for speed. If this data were isolated into a newly created (empty) and unindexed [heap] table, the load could occur first and then the indexes could be created after loading the data. Often you will realize gains in performance of ten times or better by using this scheme. In fact, by loading into an unindexed table you can take advantage of multiple CPUs by loading multiple data files in parallel or by loading multiple chunks from the same file (defined by starting and ending row position). Since both operations can benefit from parallelism, this can yield even further gains in performance. In any release of SQL Server, partitioning allows you this more granular control, and does not require that you have all data in one location。 最后，向在百忙中抽出時間對本文進行評審并提出寶貴意見的各位老師表示衷心地感謝！附錄1:相關(guān)英文文獻Partitioned Tables and Indexes in SQL Server 2005 Kimberly L. New Security Partitioned Tables and Indexes in SQL Server 2005 [D]. 2005,312..Abstract: Tablebased partitioning features in SQL Server 2005 provide flexibility and performance to simplify the creation and maintenance of partitioned tables. Trace the progression of capabilities from logically and manually partitioning tables to the latest partitioning features, and find out why, when, and how to design, implement, and maintain partitioned tables using SQL Server 2005.Partitioned Tables in SQL Server 2005While the improvements in SQL Server and SQL Server 2000 significantly enhanced performance when using partitioned v