云计算的安全架构与技术_第6节.ppt

Chains of Privileges and Metadata * Dirty Data Management in Cloud Database Introduction Data quality problem is caused by dirty data refers to inconsistent, inaccurate, erroneous, redundant, and outdate data Massive data sets contain dirty data in higher probability the difficulty of maintaining massive data errors in large storage devices will lead to more errors in data the probability of inconsistent data becomes large Platform for massive data management is necessary to manage dirty data in cloud databases * Introduction Building cloud databases is a feasible way to manage massive data sets In cloud database, three types of compute nodes master is to manage slave nodes router is to store index in cloud database slave stores data and process query locally * Cloud database * Cloud database Example of cloud database one master node two router nodes To store node index to locate slave nodes which possibly contain query results six slave nodes To store representatives and data index * Cloud database When a query is injected it is sent to a router the router searches the index to find the set of slaves containing the query results. After index searching, the query is sent to a set of slave nodes Slave nodes return local results these results are merged as final query results * Storage Model for Dirty Data Based on the inconsistency in dirty data cluster-based strategy is used to store the data the tuples in table T is partitioned into {C1:{P1, P2, P3}, C2:P4, C3:{P5, P6, P7}, C4: P8g} In a cloud database dirty data are distributed on multiple slave nodes partitioned based on the differences among tuples * Indexing Structures for Dirty Data The indexing structures have three tiers Representatives tuples in same cluster are identified by a representative tuple Data index data indices are constructed to find representative for a query Node index efficiently locating all relative compute nodes possibly contain query