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

Commodity computer systems contain more and processor cores exhibit increasingly diverse architectural tradeoffs, including memory hierarchies, interconnects, instruction sets variants, IO configurations. Previous high-performance computing have scaled in specific cases, but the dynamic nature of modern client server workloads, coupled with impossibility statically optimizing an OS for all workloads hardware variants pose serious challenges operating system structures.

This paper argues that a new class of geographically distributed network services is emerging, and the most effective way to design, evaluate, deploy these by using an overlay-based testbed. Unlike conventional testbeds, however, we advocate approach supports both researchers want develop services, clients use them. dual use, in turn, suggests four design principles are not widely supported existing testbeds: should be able run continuously access slice overlay's resources, control over...

Support for multimedia applications by general purpose computing platforms has been the subject of considerable research. Much this work is based on an evolutionary strategy in which small changes to existing systems are made. The approach adopted start ab initio with no backward compatibility constraints. This leads a novel structure operating system. aims decouple from one another and provide multiplexing all resources, not just CPU, at low level. motivation structure, design its...

Recent device hardware trends enable a new approach to the design of network server operating systems. In traditional system, kernel mediates access by applications, enforce process isolation as well and disk security. We have designed implemented Arrakis, that splits role in two. Applications direct virtualized I/O devices, allowing most operations skip entirely, while is re-engineered provide protection without mediation every operation. describe software changes needed take advantage this...

In this paper, we present techniques for provisioning CPU and network resources in shared hosting platforms running potentially antagonistic third-party applications. The primary contribution of our work is to demonstrate the feasibility benefits overbooking platforms, maximize platform yield: revenue generated by available resources. We do first deriving an accurate estimate application resource needs profiling applications on dedicated nodes, then using these profiles guide placement...

In this paper, we propose a new approach to preventing and constraining denial-of-service (DoS) attacks. Instead of being able send anything anyone at any time, in our architecture, nodes must first obtain "permission send" from the destination; receiver provides tokens, or capabilities, those senders whose traffic it agrees accept. The then include these tokens packets. This enables verification points distributed around network check that has been certified as legitimate by both endpoints...

Overlay networks are used today in a variety of distributed systems ranging from file-sharing and storage to communication infrastructures. However, designing, building adapting these overlays the intended application target environment is difficult time consuming process.To ease development deployment such overlay we have implemented P2, system that uses declarative logic language express highly compact reusable form. P2 can Narada-style mesh network 16 rules, Chord structured only 47...

The networking and distributed systems communities have recently explored a variety of new network architectures, both for application-level overlay networks, as prototypes next-generation Internet architecture. In this context, we investigated declarative networking: the use recursive query engine powerful vehicle accelerating innovation in architectures [23, 24, 33]. Declarative represents significant application area database research on processing. paper, address fundamental issues...

Declarative Networking is a programming methodology that enables developers to concisely specify network protocols and services, which are directly compiled dataflow framework executes the specifications. This paper provides an introduction basic issues in declarative networking, including language design, optimization, execution. We present intuition behind of networks, roots Datalog, extensions for networked environments, semantics long-running queries over state. focus on sublanguage we...

Recent device hardware trends enable a new approach to the design of network server operating systems. In traditional system, kernel mediates access by applications enforce process isolation as well and disk security. We have designed implemented Arrakis, that splits role in two. Applications direct virtualized I/O devices, allowing most operations skip entirely, while is re-engineered provide protection without mediation every operation. describe software changes needed take advantage this...

It is widely accepted that the current Internet architecture insufficient for future: problems such as address space scarcity, mobility and non-universal connectivity are already with us, stand to be exacerbated by explosion of wireless, ad-hoc sensor networks. Furthermore, it far from clear ubiquitous use standard transport name resolution protocols will remain practicable or even desirable.In this paper we propose Plutarch, a new inter-networking architecture. subsumes existing...

PlanetLab is a geographically distributed platform for deploying, evaluating, and accessing planetary-scale network services. shared community effort by large international group of researchers, each whom gets access to one or more isolated slices PlanetLab's global resources. Because we deployed started supporting users before fully understood what its architecture would be, being able evolve the system became requirement. This paper examines set design principles that guided this...

Systems software like databases and language runtimes typically manage memory themselves to exploit application knowledge unavailable the OS. Traditionally deployed on dedicated machines, they are designed be statically configured with sufficient for peak load. In virtualization scenarios (cloud computing, server consolidation), however, static provisioning of RAM applications dramatically reduces efficiency cost-saving benefits virtualization. Unfortunately, existing "ballooning" techniques...

In this paper, we present techniques for provisioning CPU and network resources in shared hosting platforms running potentially antagonistic third-party applications. The primary contribution of our work is to demonstrate the feasibility benefits overbooking platforms, maximize platform yeld: revenue generated by available resources. We do first deriving an accurate estimate application resource needs profiling applications on dedicated nodes, then using these profiles guide placement...

Overlay networks are used today in a variety of distributed systems ranging from file-sharing and storage to communication infrastructures. However, designing, building adapting these overlays the intended application target environment is difficult time consuming process.To ease development deployment such overlay we have implemented P2, system that uses declarative logic language express highly compact reusable form. P2 can Narada-style mesh network 16 rules, Chord structured only 47...

This paper motivates and describes an example network Information Plane, called Sophia, currently deployed on PlanetLab. Sophia is a distributed system that collects, stores, propagates, aggregates, reacts to observations about the network's current conditions. Sophia's approach novel: it can be viewed as multi-user expression evaluator in which sensors actuators form ground terms, statements take complete expressiveness of logic language like Prolog. argues this has several advantages...