LSM-Tree（6）

1. Introduction（5）

We have considered a B-tree for the Acct-ID||Timestamp index on the History file because it is the most common disk-based access method used in commercial systems, and in fact no classical disk indexing structure consistently gives superior I/O cost/performance.
我们假设B-tree存储历史文件中Acct-ID||Timestamp的索引，因为这是商业中通用方案，实际上没有磁盘架构可以给出一致的高性能表现。
We will discuss the considerations that lead us to this conclusion in Section 5.
我们在第五节讨论这个结论的注意内容。

The LSM-tree access method presented in this paper enables us to perform the frequent index inserts for the Account-ID||Timestamp index with much less disk arm use, therefore at an order of magnitude lower cost.
LSM-tree操作实现允许频繁插入Account-ID||Timestamp索引，但使用很少的磁盘操作，在一个较低级别的磁盘使用。
The LSM-tree uses an algorithm that defers and batches index changes, migrating the changes out to disk in a particularly efficient way reminiscent of merge sort.
LSM推迟和批量操作索引，迁移磁盘内容以合并排序实现。
As we shall see in Section 5, the function of deferring index entry placement to an ulti- mate disk position is of fundamental importance, and in the general LSM-tree case there is a cascaded series of such deferred placements.
第五节会看到，函数推迟索引内容放入磁盘功能非常重要，一般的LSM有级联的推迟结构。
The LSM-tree structure also supports other operations of indexing such as deletes, updates, and even long latency find operations with the same deferred efficiency.
LSM也支持七天索引操作，例如删除、更新、甚至长延迟的查找推迟操作。
Only finds that require immediate response remain relatively costly.
只有立即返回的查找相对消耗更多性能。
A major area of effective use for the LSM-tree is in applications such as Example 1.2 where retrieval is much less frequent than insert (most people don't ask for recent account activity nearly as often as they write a check or make a deposit).
LSM的一个主要使用场景是像示例1.2的应用，插入远多于插入的操作（大多数人不会像写支票或存钱那样经常询问最近的账户活动情况（有道翻译））。
In such a situation, reducing the cost of index inserts is of paramount importance;
这种情况下，减少索引插入操作成本至关重要。
at the same time, find access is frequent enough that an index of some kind must be maintained, because a sequential search through all the records is out of the question.
同时，find访问非常频繁，因此必须维护某种类型的索引，因为不可能对所有记录进行顺序搜索。（有道翻译）

Here is the plan of the paper.
一下是本文的结构。
In Section 2, we introduce the two-component LSM-tree algorithm.
第二章，介绍LSM的两个组件。
In Section 3, we analyze the performance of the LSM-tree, and motivate the multi- component LSM-tree.
第三章，分析LSM-tree性能和多组LSM-tree。
In Section 4 we sketch the concepts of concurrency and recovery for the LSM-tree.
第四章，分析并发和恢复功能。
In Section 5 we consider competing access methods and their performance for ap- plications of interest.
第五章，分析分析并发操作方法和感兴趣的应用性能。
Section 6 contains conclusions, where we evaluate some implications of the LSM-tree, and provide a number of suggestions for extensions.
第六章，结论，评估一个LSM-tree影响，提供一些扩展建议。