英语翻译The need for dynamic view management was forcefully made

英语翻译
The need for dynamic view management was forcefully made by Kotidis et al.
in their DynaMat system[14].As they observe,“This static selection of views[...]
contradicts the dynamic nature of decision support analysis.”There are a num-
ber of ways to approach the dynamic view selection problem,which we review in
Section 2.2.In this paper we explore an alternative approach to Dynamic View
Selection.We consider an OLAP system with two phases of operation:Startup
and Online.In the Startup Phase an initial set of views must be selected based
on some estimated query probabilities.This is the classical(static)view selection
problem.In the Online Phase an“in use”OLAP system is considered,for which
a set of views M has already been selected and materialized.Since over time
the relative importance of each type of aggregate query may change due to the
changing demands of its users,the system may elect to select a new set of views,
M,which better serves the incoming queries.However,view materialization is
computationally expensive and the time window in which new views can be ma-
terialized may prohibit selection of an entirely new view set.Thus the problem
becomes selecting a new view set M by discarding some views from M,and
adding new ones to be materialized.We refer to the problem of incrementally
updating a view set as the Online View Selection Problem(see Section 2).We be-
lieve that the online view selection problem is an important addition to the static
variant,as OLAP systems in practice are restarted from scratch only infrequently
and must be able to tune their performance to changing conditions on-the-fly.
Our approach to online view selection is to adapt methods that have proven to
be ective for the static variant.In this paper we develop online adaptations of
the greedy heuristic,BPUS,introduced by Harinarayan et al.[1]and three ran-
domized techniques(iterative improvement,simulated annealing and two-phase
optimization)initially proposed for static view selection by Kalnis et al.[3](see
Sections 3 and 4).Our challenge is two-fold.For the static phase,the randomized
methods must be adapted so as to take into account maintenance cost in addi-
tion to the space constraint and for the online phase all of the methods must be
adapted to take into account the existing pool of previously materialized views.

需要动态视图管理是有力地发了言kotidis等人.
在他们dynamat制度[ 14 ] ,因为他们的观察, "这静态的选择意见[...]
违背了动态性的决策支持分析" ,也确有序号-
误码率的方式对待动态视图选择问题,这是我们检讨
第2.2.in本文探索一种替代办法,以动态的观点
selection.we考虑OLAP系统与两阶段的操作：启动
和online.in启动阶段,一套初步的意见,必须选择基于
对一些估计查询probabilities.this是古典（静态）视图选择
problem.in线上的一个阶段" ,在使用" OLAP系统被认为是,为这
一套看法米已被选为materialized.since随着时间的推移
相对重要性每类聚集查询可能会改变,由于该
转变的需求,其用户时,系统可将选择一个新的一套看法,
男,更好地服务于来袭queries.however ,观点是物化
运算昂贵和时间窗,其中一些新的看法,可以马
terialized可能禁止选择一种全新的观点set.thus问题
成为选择了一条新的看法一套米乘抛弃一些看法,从米,
加入新的要予以materialized.we提及的问题,逐步
更新期定为网上视图选择问题（见第2条） ,我们得到-
lieve这家线上视图选择问题,是一项重要的,除了静态
变,因为OLAP系统,在实践中是重新开始,从无到有,只有偶尔
而且必须能够调整自己的表现不断变化的情况,对有关的禁飞区.
我们的做法,以在线视图选择,是适应方法已经证明
电子商务?选择性为静态variant.in本文我们发展网上改编
贪婪启发式, bpus介绍所里那瑞安等[ 1 ]和三个然
domized技术（迭代改善,模拟退火法,并分两阶段
优化）最初提出的静态视图选择由kalnis等[ 3 ] （见
第3和第4段） ,我们的挑战是两fold.for静态阶段,随机
方法,必须适应等,以顾及维修费用添加剂-
原因,空间的限制,并为网上阶段,所有的方法必须
改编,以顾及现有一批以前物化视图.