One of my earlier talks on agentless, introspection-based monitoring for VMs. It is interesting how long we have been pushing on this, and how recent containerization business made these ideas resonate more. I just copied from the SDI/ISTC Seminar Series for nostalgia:
DATE: Thursday, January 31, 2013
TIME: Noon – 1:00 pm
PLACE: Gates 8102
SPEAKER: Canturk Isci, IBM TJ Watson
TITLE: Agentless, Near-realtime VM Introspection in Origami
Enterprise data centers continue to embrace virtualization and cloud computing technologies due to their dramatic benefits for simplifying and streamlining system provisioning and management, as well as to improve overall resource use efficiency of the underlying hardware infrastructure via virtual machine (VM) consolidation and dynamic, distributed resource management. As virtualization and cloud automation has made it cheap and simple to create, deploy and recycle VMs on a large scale, modern data centers have become much more dynamic environments than they were a few years ago. This is creating new challenges for routine maintenance operations like security and compliance scans, that continue rely on a decades-old model based on in-VM monitoring agents and rule-driven analytics. When machines become transient and fungible resources, this model breaks down, resulting in increasing operational cost and risk.
In this talk we present a different approach to maintaining a large-scale dynamic data center. We show that out-of-VM monitoring agents can approach the data fidelity and real-time view as in-VM monitoring agents. Using this foundation we show how systems can be viewed as documents, enabling data mining and ad-hoc analytics techniques to be exploited for data center maintenance operations. These technologies are under development as part of the Origami system, an ambitious project to treat systems as data, and transform the way dynamic Cloud data centers are managed.