Pasi Sarolahti

Naming in content-oriented architectures

There have been several recent proposals for content-oriented network architectures whose underlying mechanisms are surprisingly similar in spirit, but which differ in many details. In this paper we step back from the mechanistic details and focus only on the area where the these approaches have a fundamental difference: naming. In particular, some designs adopt a hierarchical, human-readable names, whereas others use self-certifying names. When discussing a network architecture, three of the most important requirements are security, scalability, and flexibility. In this paper we examine the two different naming approaches in terms of these three basic goals.

An experimental study of home gateway characteristics

Many residential and small business users connect to the Internet via home gateways, such as DSL and cable modems. The characteristics of these devices heavily influence the quality and performance of the Internet service that these users receive. Anecdotal evidence suggests that an extremely diverse set of behaviors exists in the deployed base, forcing application developers to design for the lowest common denominator. This paper experimentally analyzes some characteristics of a substantial number of different home gateways: binding timeouts, queuing delays, throughput, protocol support and others.

Deadline-based resource management for information-centric networks

Unlike in traditional IP-based end-to-end network sessions, in information-centric networks the data source may change during a communication session. Therefore the response time to subsequent data requests may vary significantly depending on whether data comes from nearby cache, or a distant source. This is a complication for designing resource management, reliability and other algorithms, that traditionally use RTT measurements for determining when data is considered lost and should be retransmitted (along with related congestion control adjustments). This paper discusses a different approach for designing resource management in information-centric networks: data packets are assigned with a lifetime, that is used as a basis for scheduling and resource management in the network, and for congestion control and retransmission logic at the end hosts. We demonstrate an initial evaluation of this approach based on ns-3 simulations on CCN framework.

Reducing server and network load with shared buffering

Today, content replication methods are common ways of reducing the network and servers load. Present content replication solutions have different problems, including the need for pre-planning and management, and they are ineffective in case of sudden traffic spikes. In spite of these problems, content replication methods are more popular today than ever, simply because of an increasing need for load reduction. In this paper, we propose a shared buffering model that, unlike current proxy-based content replication methods, is native to the network and can be used to alleviate the stress of sudden traffic spikes on servers and the network. We outline the characteristics of a new transport protocol that uses the shared buffers to offload the server work to the network or reduce the pressure on the overloaded links.

Network expansion in OpenStack cloud federations

Cloud federation is receiving increasing attention due to the benefits of resilience and locality it brings to cloud providers and users. Our analysis of three diverse use cases shows that existing solutions are not addressing the federation needs of such use case applications. In this paper, we present an alternative approach to network federation, providing a model based on cloud-to-cloud agreements. In our scenarios, companies hosting their own OpenStack clouds need to run machines transparently in another cloud, provided by a company they have an agreement with. Our solution provides multiple benefits to cloud providers and users detailed in this paper. Our implementation outperforms the VPNaaS solution in OpenStack in terms of throughput.

Marooned magic numbers - An adaptive congestion control architecture

TCP and other Internet transport protocols rely on series of hard-coded initial values for their connection state, particularly for the congestion control parameters. For example, recently the initial value of congestion window has been under much debate, as there is a desire to make TCP more efficient for common use cases, while not endangering its performance on scenarios with limited network bandwidth. Our take on this discussion is that there is no clear single set of initial values that would work for all cases. Earlier research has proposed sharing connection and congestion control state among multiple connections over time, but that approach is limited to sharing connections to a particular host, which is not sufficient, because services are often distributed across multiple hosts, and opening multiple connections to the same host is a rather rare use case. We aim to solve this problem by proposing the Pathlet Transport Architecture that models the network paths as a series of pathlets, and uses those as the basis of initializing and maintaining the various transport parameters, particularly those related to congestion control. We analyze our initial instantiation of the PTA architecture using ns-3 simulations for TCP congestion control parameters, and show how it improves the communication performance in various different network scenarios, where single common set of magic values would fail.

WePR: A Tool for Automated Web Performance Measurement

In this paper, we present WePR, a web performance measurement tool for evaluating the download and the rendering time of a website at large scale. The first component of the tool, WebPerf, measures the technical metrics of a website such as DNS lookup time, TCP connection establishment time, page download time from different vantage points. The second component renders the website based on the metrics recorded by the WebPerf and measures the rendering time of the website in a browser window. The user experience on a website is then inferred from the measured rendering time. Along with describing WePRs architectural design and the metrics it measures, we present initial measurement results recorded March to August 2015. The results show that transmission latency, DNS lookup time, the number of HTTP elements affect the rendering time of the website and therefore the end user experience.

Locations vs. identities in internet content: Applying information-centric principles in today's networks

Recently, several proposals for informationcentric networks have surfaced and gained traction. Many of these are focused on developing a clean-slate communication architecture to replace the current model based on TCP/IP. In such architectures also the naming of data needs to be rethought. This article discusses the different approaches to naming, based on the expectations that users have about the names, and the properties that are important to network. We show that HTTP could be used as an information-centric transport protocol, and many of the benefits of information-centric networking could be achieved in the current network architecture with small protocol modifications. We also show an architecture based on using information-centric HTTP proxies and discuss some of the expected performance implications of such approach.