This paper presents an architecture for performing logging of network requests and responses, by inserting task-identifiable metadata into the request. The catch is that all the network elements/protocols that lie in the paths of this would need to modified in order to handle and propagate such metadata. Logging of data is handled by individual devices thus ensuring that sensitive information is not shared unwittingly, while the task id in the metadata allows for cooperation across domains. A (more than) proof of concept implementation is also presented.
Given that there is a hit in performance, no matter how small, I wonder if X-Trace would only be used mostly for debugging, rather than for logging of all network operations. Even if devices were X-Trace enabled, they would only propagate metadata when it is present.
As the paper mention, a user study on how much such a tool improves fault finding would provide much motivation for deploying such an architecture.
With respect to security, how easy would it be for a malicious node to hide its activity by propagating modified X-Trace data such that everything seems normal? This could make it difficult to pin-point where the true problem lies.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment