第五课(二)|学习笔记

简介: 快速学习第五课(二)

开发者学堂课程【高校精品课-西安交通大学-数据库理论与技术:第五课】学习笔记,与课程紧密联系,让用户快速学习知识。

课程地址:https://developer.aliyun.com/learning/course/12/detail/15861


第五课(二)

 

内容介绍

一、Data Guides

二、Existence of Multiple DataGuides

三、Annotations

四、Strong Data Guides

五、Mapping Long Records to Pages

六、Allocation and Free-Space Management

七、文件组织

八、文件中的记录组织

四、Strong Data Guides

We define a class of DataGuides that supports annotations as described in the previous subsection.

Intuitively , we are interested in DataGuides where each set of label paths that share the same ( singleton ) target set in the DataGuide is exactly the set of label paths that share the same target set in the source. 

Definition 7.:Consider OEM objects s and d where d is a DataGuide for a sources.Given a label path l of slet Ts(l) be the target set of l in s, and let Td(l) be the ( singleton ) target set of l in d. Let Ls(l) = { m | Ts(m) =Ts(l)}. That is, Ls(l) is the set of all label paths in s that share the same target set as l.Similarly,let Ld(l)={ m | Td(m)=Td(l) }. That is, Ld(l) is the set of all label paths in d that share the same target set as l. If, for all label paths l of s, Ls(l) = Ld(l), then d is a strong DataGuide for s. 

For example , Figure 3(c) is not a strong DataGuide for Figure 3(a). The source target set Ts(B.C)is {6,7},and the DataGuide target set Td(B.C) is {20}.

In the source, Ls(B.C) is {B.C}, since no other source label paths have the same target set. 

In the DataGuide(c),however,Ld(B.C)is{B.C,A.C}. SinceLs(B.C)# Ld(B.C), the DataGuide is not strong.The reader may verify that Figure 3(b)is in fact a strong DateGuide 

Next, we prove that a strong DataGuide is sufficient for storage of annotations.

Theorem 1:Suppose d is a strong DataGuide for a source s. lf an annotation p of some label path l is stored on the object o reachable via l in d,then p describes the target set in s of each label path that reaches o.

We also prove that a strong DataGuide induces a straightforward one-to-one correspondence between source target sets and DataGuide objects.This property is useful for incremental maintenance and query processing.

Theorem 2:Suppose d is a strong DataGuide for a source s.Given any target sett of s,t is by definition the target set of some label path l.Compute Td(l),the target set of l in d,which has a single element o.Let F describe this procedure, which takes a source target set as input and yields a DataGuide object as output. In a strong DataGuide,F induces a one-to-one correspondence between source target sets and DataGuide objects.

 

五、Mapping Long Records to Page

1.Long Pages

(1)Allows long pages (up to 64K bytes).

(2)Internally, long records are managed just as short ones.

(3)Simple solution, but inadequate for very large records.

(4)Used in DB2 and Oracle.

2.Separate Pages for Long Fields

(1)Instead of keeping entire long field in a record.

(2)maintains only a pointer to the long field stored in another page.

(3)The main record is then treated as a short tuple.

(4)Good results if long attributes are accessed infrequently.

(5)Unfortunately, performance suffers in situations when ong attributes are accessed frequently.

 图片254.png

 

六、Allocation and Free-Space Management

1.Free-Space Information Pages

Advantages:

(1)FSIPs need not always be updated during allocation and deallocation.

(2)Deallocations are relatively fast and simple

(3)Provides good storage utilization and reclamation(回收) of free and unused space.

Disadvantages:

Allocations may require several disk accesses.

部分习题如下:

图片255.png

图片256.png

相关文章
|
运维 监控 算法
从定时任务-到任务调度系统xxl-job
定时任务的今生前世以及xxl-job调度系统
3367 0
从定时任务-到任务调度系统xxl-job
|
弹性计算 自然语言处理 测试技术
30分钟私有部署Deepseek-R1和V3,轻松拥有企业专属超大模型
阿里云计算巢推出了基于 ECS镜像、Vllm和Ray集群的超大模型一键部署方案。该方案通过预置标准化环境的ECS镜像,结合自动化云资源编排模板(如Ros模板),将Deepseek-R1和Deepseek-V3的私有化部署流程简化为一键操作。用户无需深入理解底层技术细节,即可在 30分钟内 快速完成模型环境搭建与服务启动,真正实现“开箱即用”的企业级超大模型服务。
|
API 数据库 决策智能
基于百炼平台qwen-max的api 打造一套 检索增强 图谱增强 智能工具调用决策的智能体
本文介绍了一种基于阿里云百炼平台的`qwen-max` API构建的智能体方案,该方案集成了检索增强、图谱增强及智能工具调用决策三大模块,旨在通过结合外部数据源、知识图谱和自动化决策提高智能回答的准确性和丰富度。通过具体代码示例展示了如何实现这些功能,最终形成一个能灵活应对多种查询需求的智能系统。
1469 11
|
运维 Kubernetes 监控
K8S异常诊断之俺的内存呢
K8S异常诊断之俺的内存呢
|
机器学习/深度学习 算法
深度学习笔记(四):神经网络之链式法则详解
这篇文章详细解释了链式法则在神经网络优化中的作用,说明了如何通过引入中间变量简化复杂函数的微分计算,并通过实例展示了链式法则在反向传播算法中的应用。
1128 0
深度学习笔记(四):神经网络之链式法则详解
|
机器学习/深度学习 人工智能 测试技术
NeurIPS D&B 2024 突破短视频局限!MMBench-Video解读MLLM视频理解能力
GPT-4o 四月发布会掀起了视频理解的热潮,而开源领军者Qwen2也对视频毫不手软,在各个视频评测基准上狠狠秀了一把肌肉。
|
安全 测试技术 网络安全
|
Linux
Linux主机名重启失效解决
Linux主机名重启失效解决
941 1
|
存储 数据管理 数据处理
阿里云oss收费标准
阿里云oss收费标准,阿里云对象存储OSS收费有两种计费模式,即包年包月和按量付费,包年包月是指购买存储包、流量包来抵扣OSS产生的存储费核流量费,OSS标准(LRS)存储包100GB优惠价33元、500GB存储包半年162元、OSS存储包40GB一年9元,OSS流量包100G 49元/月
13299 1
[Halcon&小结] 判断Halcon区域region是否为空
[Halcon&小结] 判断Halcon区域region是否为空
1413 0