Abstract
Metadata operations constitute a significant fraction of the network traffic between an SMB2 client and server running a typical file server workload. The simplified command set defined by the SMB2 protocol requires that a client implementation must make judicious use of metadata caching and compounding in order to effectively utilize network bandwidth. In the first part of the talk we will discuss the metadata caching strategies employed by the Windows SMB (should this be “SMB2”?) client implementation to reduce metadata traffic to the server. Specifically we will look into the 3 different metadata caches – the file-information cache, the directory cache and the file-not-found cache and see how these caches help reduce network round trips. We will also look at various configuration settings to control the sizes and lifetimes of these caches. The second part of the talk will look at some of the consistency issues that arise due to opportunistic metadata caching. We will also look at potential protocol improvements and show how we can mitigate these problems by improving the consistency and lifetimes of these metadata caches.
Learning Objectives
Gain insight into the data and metadata caching policies of the Windows SMB2 client
Learn about client settings that can be used to tune the metadata caching behavior
Understand the benefits these caches provide in different scenarios."
Investigate how the introduction of new protocol capabilities could further refine data and metadata caching