Computer Science Department
School of Computer Science, Carnegie Mellon University


The Manna Plug-In Architecture for
Content-based Search of VM Clouds

Wolfgang Richter, Glenn Ammons*, Jan Harkes, Adam Goode,
Nilton Bila**, Eyal de Lara**, Mahadev Satyanarayanan

August 2010


Keywords: Data-intensive computing, virtual machines, cloud computing, interactive search, forensic analysis, computer vision, pattern recognition, distributed systems, Diamond, OpenDiamond, Mirage

As cloud computing becomes more popular, collections of virtual machine (VM) images are growing in size. Management of VM collections requires the ability to inspect and search data stored within VM images. We present a plug-in-based architecture, called Manna, for efficiently searching state within VM images through both index and non-index based search. The architecture offers a flexible framework for creating a wide range of new applications that are valuable to both end users and administrators of VM images. We showcase this flexibility through three applications built using Manna's API: one for searching images, one for searching source code, and one for performing virus scanning. Efficient search for such diverse applications is achieved using two independent mechanisms: plug-ins that are data-type-specific, but independent of data source, and use of VM-specific metadata to shrink the search space of non-indexed data. Trace-driven measurements on our prototype confirm that Manna searches incur low performance overhead.

21 pages

*IBM Research
**University of Toronto

Return to: SCS Technical Report Collection
School of Computer Science

This page maintained by