国庆长假，老外在热烈的讨论这个话题，

原文地址： http://natishalom.typepad.com/nati_shaloms_blog/2007/10/why-most-scalab.html

我摘录部分观点如下：

Most of these sites are using LAMP as the core runtime stack

Some develop their own file system (Google, GFS)

Some are using caching to solve the database bottleneck (memcached and the like)

--

introducing similar solutions for addressing the scalability challenges: 在解决伸缩性方面的尝试：

On the Data Tier we see the following:

1. Adding a caching layer to take advantage of memory resources

availability and reduce I/O overhead
2. Moving from a database-centric approach to partitioning, aka shards

注：shards是google贡献给hibernate的一个项目，他是hibernate的一个扩展，可以

透明的实现数据的分布存储。至于如何分布存储由你自己定义策略，例如典型的

RoundRobinShardSelectionStrategy

 

On the Business Logic Tier:

3. Adding parallelization semantics to the application tier (e.g.,

MapReduce)
注：MapReduce是Google开发的C++编程工具，用于大规模数据集（大于1TB）的并行运算。MapReduce通过把对数据集的大规模操作分发给网络上的每个节点实现可靠性；每个节点会周期性的把完成的工作和状态的更新报告回来。MapReduce会生成大量的临时文件，为了提高效率，它利用Google文件系统来管理和访问这些文件。

4. Moving to scale-out application models to achieve linear scalability
注：
scaling：the ability of an application to address growth in throughput,

usage, and capacity
有两种策略：scale out 和 scale up
scale up vs scale out 的区别在于：
 1 Reliance on hardware versus reliance on software
 2 Equation with "first-class" hardware versus "commodity" hardware
 3 Massive versus incremental capacity increases
 4 Centralized versus partitioned application architectures
简单的说就是一个靠提升单个硬件的配置，另一个靠用低端配置实现集群

5. Moving away from the classic two-phase commit and XA for transaction

processing  (See: Lessons from Pat Helland: Life Beyond Distributed

Transactions)
注： 不使用传统的分布式事务处理，例如EJB的事务，使用新的基于space的中间件平台 Space Based Architecture (SBA) and GigaSpaces XAP