Mic Bowman

PlanetLab: an overlay testbed for broad-coverage services

PlanetLab is a global overlay network for developing and accessing broad-coverage network services. Our goal is to grow to 1000 geographically distributed nodes, connected by a disverse collection of links. PlanetLab allows multiple service to run concurrently and continuously, each in its own slice of PlanetLab. This paper discribes our initial implementation of PlanetLab, including the mechanisms used to impelment virtualization, and the collection of core services used to manage PlanetLab.

An Analysis of Performance Interference Effects in Virtual Environments

Virtualization is an essential technology in modern datacenters. Despite advantages such as security isolation, fault isolation, and environment isolation, current virtualization techniques do not provide effective performance isolation between virtual machines (VMs). Specifically, hidden contention for physical resources impacts performance differently in different workload configurations, causing significant variance in observed system throughput. To this end, characterizing workloads that generate performance interference is important in order to maximize overall utility. In this paper, we study the effects of performance interference by looking at system-level workload characteristics. In a physical host, we allocate two VMs, each of which runs a sample application chosen from a wide range of benchmark and real-world workloads. For each combination, we collect performance metrics and runtime characteristics using an instrumented Ken hypervisor. Through subsequent analysis of collected data, we identify clusters of applications that generate certain types of performance interference. Furthermore, we develop mathematical models to predict the performance of a new application from its workload characteristics. Our evaluation shows our techniques were able to predict performance with average error of approximately 5%

MILSA: A Mobility and Multihoming Supporting Identifier Locator Split Architecture for Naming in the Next Generation Internet

Naming and addressing are important issues for next generation Internet (NGI). In this paper, we discuss a new mobility and multihoming supporting identifier locator split architecture (MILSA). There are three main contributions of our solution. First, we separate trust relationships (realms) from connectivity (zones). A hierarchical identifier system for the realms and a Realm Zone Bridging Server (RZBS) infrastructure that performs the bridging function is introduced. Second, we separate the signaling and data plane functions to improve the performance and support mobility. Third, to provide transparency to the upper layer applications, identifier locator split happens in network layer. A Hierarchical URI-like Identifier (HUI) is used by the upper layers and is mapped to a locators set by HUI Mapping Sublayer (HMS) through interaction with RZBS infrastructure. Further scenarios description and analysis show the benefits of this scheme for routing scalability, mobility and multihoming.

Distributed scene graph to enable thousands of interacting users in a virtual environment

Virtual environments are currently limited to no more than a hundred interacting users by the simulator-centric server architectures used for many of these applications. There are some potential new usages such as virtual concerts and sporting events involving hundreds or thousands of users and we seek to enable these exciting new applications. We propose a distributed scene graph (DSG) architecture which enables massive scaling of scene complexity and participants with the addition of hardware. A prototype implementation of the DSG components to manage client communications demonstrates an order of magnitude increase in the number of concurrent users. We present the design of this component within the DSG architecture, prototype implementation based on an open source virtual environment server and our experimental setup, workloads and results.

Scaling virtual worlds: simulation requirements and challenges

Virtual worlds use simulation to create a fully-immersive 3D space in which users interact and collaborate in real time. It is still a great challenge to scale virtual worlds to provide rich user experiences, high level of realism, and innovative usages. There are three unique simulation requirements in scaling virtual worlds: (1) large-scale, real time and perpetual simulations with distributed interaction, (2) simultaneous visualization for many endpoints with unique perspectives, and (3) multiple simulation engines with different operation characteristics. In this paper, we review the challenges in meeting these requirements, present the scalability barriers we observed in current virtual worlds, and discuss potential virtual world architecture and solutions to address the challenges and overcome the barriers.

Survey of state melding in virtual worlds

The fundamental goal of virtual worlds is to provide users with the illusion that they are all seeing and interacting with each other in a consistent world. State melding is the core of creating this illusion of a shared reality. It includes two major parts: consistency maintenance and state update dissemination. Well-designed state melding technologies are also critical for developing a virtual world that can scale to a large number of concurrent users and provide satisfying user experiences. In this article, we present a taxonomy of consistency models and categorization of state update dissemination technologies for virtual worlds. To connect theories and practices, we then apply the taxonomy to case study several state-of-the-art virtual worlds. We also discuss challenges and promising solutions of state melding in large-scale virtual worlds. This survey aims to provide a thorough understanding of existing approaches and their strength and limitations and to assist in developing solutions to improve scalability and performance of virtual worlds.

Globus and PlanetLab resource management solutions compared

PlanetLab and Globus Toolkit are gaining widespread adoption in their respective communities. Although designed to solve different problems - PlanetLab is deploying a worldwide infrastructure testbed for experimenting with network services, while Globus is offering general, standards-based, software for running distributed applications over aggregated, shared resources - both build infrastructures that enable federated, extensible, and secure resource sharing across trust domains. Thus, it is instructive to compare their resource management solutions. To this end, we review the approaches taken in the two systems, attempt to trace back to starting assumptions the differences in these approaches, and explore scenarios where the two platforms can cooperate to the benefit of both user communities. We believe that this is a key first step to identifying pieces that could be shared by the two communities, pieces that are complementary, and how Globus and PlanetLab might ultimately evolve together.

Scale Virtual Worlds through Dynamic Load Balancing

Dynamic load balancing holds the potential to scale virtual worlds flexibly by dynamic allocation of hardware to match load. In this paper, we study the benefits and overheads of space based load partitioning, in particular, distributed binary space partitioning (BSP). Our evaluation is based on Open Simulator, a virtual world system compatible with Second Life® viewers. Our work reveals that although simple and effective, distributed BSP has several limitations and suffers from high overhead. We then analyze the fundamental reasons of these limitations. To overcome the limitations, we argue that it is necessary to break away from the simulator-centric architecture used in today’s virtual worlds, and present potential new directions.

Enabling behavior reuse in development of virtual environment applications

Virtual environments (VEs) provide simulated 3D spaces in which users can interact, collaborate, and visualize in real time. Accordingly, virtual environments have the potential to transform education, creating classrooms that ignore geographic boundaries and immerse students in experiences that would be difficult or impossible to arrange in the real world. A major impediment to the widespread adoption of educational VEs is the high cost of developing VE applications. We believe application development must become tractable for non-expert users in the same way that Web development is no longer the exclusive purview of professional programmers. In this position paper, we describe our experiences in enabling behavior reuse across VE applications. Our approach replaces, whenever possible, application-specific behaviors with general purpose, reusable simulation modules. These modules bootstrap one another until a rich ecosystem develops; thus, VE application development is reduced to compositing content and behaviors instead of developing them from scratch.

CADIS: aspect-oriented architecture for collaborative modeling and simulation

The development of large and complex simulated models often requires teams to collaborate. One approach is to break a large model into independently developed partial models that, when combined, capture the overall behavior. However, maintaining consistent world state across independently developed simulations is a challenge. In this paper, we introduce the Collaborative Aspect-Oriented Distributed Interactive Simulation (CADIS) architecture and development platform. CADIS embodies a new paradigm for integrating independently developed time-discrete partial models and simulations, focusing on transparently maintaining synchronized shared state. Data is pulled and instantiated in the beginning of each time step, and pushed at the end of each time step. An urban simulation is used to demonstrate CADIS capabilities and performance. We show how simple optimizations can bring the performance of the framework to acceptable levels, making CADIS a viable modeling and simulation methodology supporting separation of concerns.