主流云计算的安全架构与技术分析.pptVIP

主流云计算的安全架构与技术分析Cloud Technology - Google Google File System S. Ghemawat, H. Gobioff, and S.T. Leung Proc. 9th ACM symposium on Operating systems principles (SOSP’03) Google Disk Farm Early days… …today GFS GFS shares many of the same goals目的 as previous distributed file systems分布式文件系统 performance, scalability可扩展性, reliability可靠性, and availability有效性 Different with earlier file system Assumptions假定 Building from many inexpensive commodity components廉价的商品元件 that often fail Constantly Monitor监控 itself, Detect, tolerate, and recover promptly立即的 from component failures Storing a modest适度的 number of large files Million files (typically 100 MB or larger) Multi-GB files are the common case常见情况 should be managed efficiently Small files must be supported必须支持小文件 Assumptions假定 The workloads工作量 consist of two kinds of reads Large streaming reads大流读取 Individual operations单独操作 typically read hundreds of KBs, more commonly一般地 1 MB or more Small random reads小的随机读取 typically reads a few KBs at some arbitrary offset The workloads also have many large, sequential连续的 writes that append data to files Typical operation sizes are similar those for read Once written, files are not often modified again不经常修改 Small writes at arbitrary positions in a file are supported支持任意位置的写操作 not efficient Assumptions假定 The system must efficiently implement执行 for multiple clients that concurrently append to the same file多客户端同时添加到同一个文件 Our files are often used as producer-consumer queues生产者消费者队列or for many-way merging多路合并 Hundreds of producers, running one per machine, will concurrently append to a file Atomicity原子性 with minimal synchronization overhead最小同步开销 is essential The file may be read later or a consumer may be reading through the file simultaneously消费者可以同时读取文件 Assumptions假定 High sustained bandwidth is more important than low latency持续的高带宽比低延时更重要 Our target applications place a premium on processing data in bulk at a high rate我们的目标应用诱发高速处理数据 Few have stringent resp