direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Inhalt des Dokuments

Talks, student talks, and other events of 2016

Thursday, 15. December 2016

Network Trace Analysis: Linux Toolchain versus Apache Flink
Speaker:
Yu Zhao
Type:
Masterstudent introductory talk
Time:
15 December 2016
14:00–15:00
Place:
MAR 4.033
Abstract:
The work compares the data mining process of network traces with typical Linux tools and Apache Flink jobs.

Tuesday, 13. December 2016

Network-data analytics for ISPs: A four-dimensional view on network traffic”
Speaker:
Phillip Urbanik (BENOCS)
Type:
Guest Talk
Time:
13 December 2016
16:00–17:00
Place:
MAR 4.033
Abstract:
Internet Service Providers (ISPs) rely on network data analytics to characterize traffic in their network with different objectives in mind (e.g., network planning). To gain maximum insight, ISPs cannot limit themselves to collect and process multiple data streams independently (e.g., NetFlow, BGP, ISIS, DNS, SNMP). Instead, they have to correlate these streams. For example, by combining its NetFlow and BGP data an ISP can characterize how packets will be forwarded once these enter its network.

In this presentation, I will describe how to complement state-of-the-art analytics solutions designed to help transit providers. In particular, I will explain how to dissect traffic in four different dimensions at the Autonomous System (AS) level. These dimensions are: 1) the source AS, 2) the hand-over AS (which neighboring network delivered traffic to the ISP), 3) the next-hop AS (to which neighboring network the ISP delivers traffic to), and 4) the destination AS.

The most important challenge we face to present this view to an ISP is how to determine the hand-over AS when no information about the interconnect is available (i.e., only from traffic statistics). Consequently, the main topic of this presentation will be the classification of router interfaces by leveraging observations about the egress traffic. I will then explain some of the limitations of our approach and conclude with a discussion on future work.

Tuesday, 13. December 2016

Investigating Mechanisms for Tracing Packets through SDN
Speaker:
Said Jawad Saidi
Type:
Masterstudent final talk
Time:
13 December 2016
15:00–16:30
Place:
MAR 4.033
Abstract:
With rapid growth of internet users, mobile users, increasing number of applications; complexity of businesses depending on correct operation of networks also increases. Therefore, businesses require computer networks to be increasingly reliable and bug free to prevent interruption in business operations. Reliability, troubleshooting, and catching bugs in current networks is now a fundamental element.

Network operators still use elementary, and ad hoc tools to obtain insights about behavior and operation of their networks. In Software Defined Networks, supporting network administrators with efficient, and feature-rich packet tracing mechanisms helps reducing downtime and maintenance cost.

Current packet tracing mechanisms working in Control Plane and Data Plane, still have room for improvement. Control Plane mechanisms need to be more scalable. They also must provide a better approach to model the behavior of network such as usage of ”Equal Cost Multi Path”, and Packet Encapsulation.

Packet Tracing tools in Data plane can provide insights about the actual events and behavior of network, as they unfold. However, the tools that use packet marking may cause more harm than their benefit, if the packet marking rules interfere with forwarding rules.

Control Plane packet tracing must be complemented by mechanisms at the data plane. Because data plane mechanisms can report the state of packet forwarding. The new hybrid system must operate with no assumption of consistency of state at Control Plane with the state at Data Plane.

Tuesday, 22. November 2016

Alternatives for Server Monitoring and Implementation at the Hasso Platner Institute Future SOC Lab.
Speaker:
Henning Schwaneberger (TU Berlin bachelor student)
Type:
Bachelorstudent final talk
Time:
22 November 2016
14:00–15:00
Place:
MAR 4.013
Abstract:
With the size and complexity of computing infrastructure permanently rising, keeping server monitoring manageable is increasingly becoming a critical challenge. Mindful of this problem, this bachelor thesis aims to investigate the current landscape of server monitoring solutions. We select a representative sample of state-of-the-art server monitoring tools and compare them across pinpoint criteria. We implemented the monitoring tools in a production network at HPI Future SOC Lab, a leading German computing architecture, and report on this installation.

The five monitoring tools under consideration are Nagios, Zabbix, Ganglia, Centreon, and the Elastic Stack. The assessment shows that they exhibit similarities, e.g., they are all open source in their cores, get support from their communities, and rely on the fees of an optional support for financing. However, they offer differing features and add-ons; e.g., Nagios proposes a large bundle of features while Ganglia is leaner. Overall, we believe that this work is relevant for anyone who wants to get a compact comparative overview of existing server monitoring solutions.

Monday, 21. November 2016

Building secure SDN networks from untrusted components through the Robust Combiner approach
Speaker:
Philipp Heyder
Type:
Masterstudent final talk
Time:
21 November 2016
15:00–16:30
Place:
MAR 4.033
Abstract:
The so called SYNful Knock router implant created an alarming precedent: Network equipment may contain backdoors which grant unauthorized remote operators multiple ways to temper with a network, possibly without being noticed by the network's owner. This is particularly threatening to governments and other entities which do not have the resources to manufacture their own switching and routing equipment and thus have to rely on third party vendors. We believe the robust combiner approach, when applied to network switches can successfully identify these backdoors and render them harmless. We replace a single untrusted switch with a circuit of three parallel untrusted switches, ideally manufactured by different vendors. Since these three switches mimic a single one, their forwarding logic should be equal. Furthermore since all of them experience the same input traffic, their forwarding behaviour should be equal as well. An added middlebox application compares the output of all three switches. Only a packet which was received from a majority of the switches will be forwarded. We assume that if a packet is the product of a backdoor, it will differ from its benign counterparts and therefore will not be forwarded. In the long-term we believe a backdoored switch will reveal itself, by producing a significantly higher amount of unforwardable packets than its counterparts. We focus on Software Defined Network switches, because the separation of control and data plain aids our design. Two challenges will be tackled by our work: First we will create realistic threat models to test the security benefits of our design. Secondly we will need to weigh these benefits against the strain they put on overall network performance: Placing our robust combiner in a Fat Tree topology and measuring different performance characteristics under realistic traffic patterns will guide us to further improve our design. Finally we hope to present a solution, which is secure against our thread models, reasonably light on performance and simple enough to be trustworthy.

Tuesday, 08. November 2016

SDN assisted Path Selection for MPTCP
Speaker:
Imed Bezahaf
Type:
Masterstudent introductory talk
Time:
08 November 2016
15:00–16:00
Place:
MAR 4.033
Abstract:
When applications need high throughput, using MPTCP (Multipath TCP) can be an effective approach toward this goal. It provides the ability to split a TCP stream into multiple flows (subflows) by enabling the simultaneous use of multiple network interfaces. This provides better throughput and resilience to failures. However, in practice, the current MPTCP implementation has a major limitation with regards to the throughput. The problem lies in the lack of MPTCP having control over the paths of the subflows. In an enterprise network, the path selection heavily relies on the forwarding decisions in the switches. The forwarding rules generally take the traffic between a source and a destination through the same shortest path creating contention and therefore decreased throughput. In a controlled environment, where clients can easily be upgraded to support MPTCP, we will explore the option of a SDN (Software Defined Networking) controller that has global visibility on the network and the MPTCP subflows. Using link monitoring and topology information, the controller will find paths that provide better throughput for applications by altering the paths of the subflows and reduces under-utilization by establishing subflows on unused paths.

Tuesday, 25. October 2016

On the Past, Present and Future of “Big (Internet) Data”
Speaker:
Walter Willinger (NIKSUN, Inc., Princeton NJ)
Type:
Guest Talk
Time:
25 October 2016
16:00–17:00
Place:
MAR 4.033
Abstract:
With petabytes or even zettabytes of data that are continuously collected about various aspects of the Internet, how hard can it be to obtain an accurate picture of its traffic, its physical topology (i.e., router-level Internet), its logical overlays (e.g., the Web, online social networks), or its “dark” sides and associated activities (i.e., cyber crimes)? In this talk, I will first use some well-known examples of “big (Internet) data” to illustrate what this data does and doesn’t tell us about the Internet’s traffic and its physical topology. Moving on to the problem of cyber security, I will then discuss why and how future Internet measurement studies have to change so that the much-heralded “big data” approach to Internet research can achieve ifs full potential.

Monday, 24. October 2016

Understanding Benefits of Different Vantage Points in Today’s Internet
Speaker:
M.Sc. Jan Böttger
Type:
Wissenschaftliche Aussprache zur Erlangung des akademischen Grades „Doktor der Ingenieurwissenschaften“
Time:
24 Oktober 2016
13:00–15:30
Place:
MAR 4.033
Der Promotionsausschuss setzt sich wie folgt zusammen:
Vorsitz:
Prof. Dr. Küpper
Gutachter:
Frau Prof. Feldmann, Ph.D.
Prof. Dr. Willinger (NIKSUN, USA)
Prof. Dr. Kao
Prof. Dr. Seifert

Die Dissertation und die Gutachten können von Berechtigten nach § 8 Abs. 1 der Promotionsordnung in der Fakultätsverwaltung eingesehen werden. Die wissenschaftliche Aussprache ist universitätsöffentlich.

Wednesday, 12. October 2016

An investigation of SSH NOAUTH usage in the wild
Speaker:
Oliver Machacik (TU Berlin bachelor student)
Type:
Bachelorstudent final talk
Time:
12 October 2016
14:00–15:00
Place:
MAR 4.013
Abstract:
SSH is one of the most widely used protocols on the Internet. It
offers an ubiquitous solution for remote logins and provides strong security guarantees. However, even if the protocol itself is secure in principle it is not immune to reckless configuration and incorrect usage. This thesis investigates the amount of SSH hosts, that do not perform authentication, the reasons behind it, and security implications for the affected hosts. The research is based on the analysis of datasets collected during Internet-wide scans. Our results show, that there is a notable number of SSH servers which neglect authentication and permit connection establishment with arbitrary credentials. Although additional authentication mechanisms are usually in place we show that a subset of these host can be exploited to forward TCP connections. Furthermore we explore basic parameters of reachable SSH servers on the Internet overall. This comprises popular implementations and authentication methods, geographical distribution of detected hosts and reuse of hostkeys.

Tuesday, 12. July 2016

Enabling Asynchronous I/O for the Socket Intent Framework
Speaker:
Tobias Kaiser

Type:
Masterstudent final talk
Time:
12 July 2016
15:00–16:00
Place:
MAR 4.033
Abstract:
The Socket Intent Framework allows applications to optimize network connections for their needs. The optimization choices are made by a background daemon, with whom the application communicates using a custom library. This thesis analyzes the two previously existing interfaces that this library provided to the applications, which are both based on synchronous, blocking input/output operations.

A new set of non-blocking API functions, which hide the communication with the background daemon, were defined and implemented in order to make it possible to integrate the library into applications that depend on asynchronous I/O. The disadvantage of this approach was, that it relied on the use of system call wrappers to function properly.

This is why another API based on extending the domain name resolution function getaddrinfo() has been developed. This approach looks promising for current and future uses. It currently does not support asynchronous domain name resolution, but could be extended to do that in the future.

Monday, 27. June 2016

On predictable performance for distributed systems
Speaker:
M.Sc. Lalith Suresh Puthalath
Type:
Wissenschaftliche Aussprache zur Erlangung des akademischen Grades „Doktor der Ingenieurwissenschaften“
Time:
27 Juni 2016
16:00–18:30
Place:
MAR 4.033
Der Promotionsausschuss setzt sich wie folgt zusammen:
Vorsitz:
Prof. Dr. Markl
Gutachter:
Frau Prof. Feldmann, Ph.D.
Prof. Dr. Canini (Université catholique de Louvain, Belgien)
Prof. Dr. Crowcroft (University of Cambridge, UK)
Prof. Prof. Dr. Zwaenepoel (EPFL, Schweiz)
Die Dissertation und die Gutachten können von Berechtigten nach § 8 Abs. 1 der Promotionsordnung in der Fakultätsverwaltung eingesehen werden. Die wissenschaftliche Aussprache ist universitätsöffentlich.

Monday, 20. June 2016

Improving SDN with InSPired Switches
Speaker:
Fabian Schneider (NEC Laboratories Europe)
Type:
Guest Talk
Time:
20 June 2016
16:00–17:00
Place:
MAR 4.013
Abstract:
In SDN, complex protocol interactions that require forging network packets are handled on the controller side. While this ensures flexibility, both performance and scalability are impacted, introducing serious concerns about the applicability of SDN at scale. To improve on these issues, without infringing the SDN principles of control and data planes separation, we propose an API for programming the generation of packets in SDN switches. Our InSP API allows a programmer to define in-switch packet generation operations, which include the specification of triggering conditions, packet’s content and forwarding actions. To validate our design, we implemented the InSP API in an OpenFlow software switch and in a controller, requiring only minor modifications. Finally, we demonstrate that the application of the InSP API, for the implementation of a typical ARP-handling use case, is beneficial for the scalability of both switches and controller.
Bio:
Prior to joining NEC in Jan 2012, Fabian spent 12 years researching, administrating and designing computer networks, focusing mainly on network measurement and monitoring in his scientific career. While earning his Ph.D., Fabian worked at DT Labs in Berlin where he developed broadband access network measurement solutions and studied user behavior on the Internet. During his PostDoc at Université Pierre et Marie Curie in Paris, Fabian worked on home network monitoring and network performance evaluation of virtualized systems. After joining NEC, Fabian shifted attention to network and cloud management. He is active in the ONF, participating in the Framework and Architecture working group and earlier in hybrid and northbound-api groups. Fabian holds a Ph.D. in computer science from TU Berlin, Germany and a Diplom (german M.Sc equivalent) from TU München, Germany.
Quelle: ONF

Tuesday, 07. June 2016

NAPALM - Network Automation and Programmability Abstraction Layer with Multivendor support
Speaker:
Elisa Jasinska
Type:
Masterstudent introductory talk
Time:
07 June 2016
15:00–16:00
Place:
MAR 4.033
Abstract:
The Python library dubbed Napalm - short for Network Automation and Programmability Abstraction Layer with Multivendor support - helps interact with devices from various vendors though a unified set of commands. Napalm provides methods to replace or merge the configurations on the various devices, it is open source and available on Github: [1]https://github.com/napalm-automation/napalm.
In this talk we will introduce Napalm, talk about its background as well as related projects, and dive into its capabilities in detail.

Tuesday, 31. May 2016

Masterstudent introductory talk
Speaker:
Imed Bezahaf
Type:
Masterstudent introductory talk
Time:
31 May 2016
14:00–15:00
Place:
MAR 4.033
Abstract:
One of the core properties of Software Defined Networking (SDN) is to provide network architects advanced programmability features and a dynamic automation for configuring and implementing applications like firewall, load-balancer, cloud operations and content distribution.
Nowadays, in a distributed SDN environment the capabilities of each local controller and therefore the placement of SDN applications are not taking into consideration. In our work, we incorporate the needs of SDN applications in order to select the best suited controller. Precisely, we assume that each application has its own needs, which we call “intents” (bandwidth, location, latency tolerance, throughput and so on). Using a reactive distributed intent-based approach, we consider the characteristics of each SDN controller and match them against the application's intents. We then place the application on that SDN controller, that fulfils the application's intents best.

Tuesday, 26. April 2016

Flexible WiFi deployments with commodity hardware
Speaker:
Raphael Lisicki

Type:
Masterstudent final talk
Time:
26 April 2016
14:00–15:00
Place:
MAR 4.033
Abstract:
The increasing demand for flexibility in WiFi network deployments along with more stringent requirements on performance and security stand in stark contrast to today’s ossified and expensive WiFi architecture. In particular, today’s WiFi consists of a large number of control and data plane network functions that are either bundled into a single proprietary WiFi controller or they are distributed across the networks and run on the WiFi Access Points. In this talk, we present the notion of a wireless switch and decomposition of some of the WiFi MAC functionality which enables the deployment of flexible WiFi architectures. Specifically, we present a way to decouple the client’s association, authentication, and cryptography from the WiFi device so that these function can be placed where (and when) they are most useful.

Wednesday, 20. April 2016

Realizing Software-Defined Wireless Networking – Achieving fine-grained Wi-Fi programmability with off-the-shelf hardware
Speaker:
Dipl.-Inf. Julius Schulz-Zander
Type:
Wissenschaftliche Aussprache zur Erlangung des akademischen Grades „Doktor der Ingenieurwissenschaften“
Time:
20 April 2016
17:15–19:30
Place:
MAR 4.033
Der Promotionsausschuss setzt sich wie folgt zusammen:

Vorsitz:

Prof. Dr.-Ing. Stanczak

Gutachter:

Frau Prof. Feldmann, Ph.D.

Prof. Dr.-Ing. Ott (TU München)

Frau Papagiannaki, Ph.D. (Telefonica I+D, Spanien)

Prof. Dr. Schmid (Aalborg University, Dänemark)

Die Dissertation und die Gutachten können von Berechtigten nach § 8 Abs. 1 der Promotionsordnung in der Fakultätsverwaltung eingesehen werden. Die wissenschaftliche Aussprache ist universitätsöffentlich.

Tuesday, 22. March 2016

Survey of monitoring techniques
Speaker:
Apoorv Shukla
Type:
Project group meeting (Projektgruppentreffen, PGT)
Time:
22 March 2016
14:00 - 15:00
Place:
MAR 4.033

Tuesday, 15. March 2016

Cost structure of the Internet
Speaker:
Stefan Wahl (CEO of Netsign GmbH)
Type:
Invited Talk
Time:
15 March 2016
15:00–16:00
Place:
MAR 4.033
Abstract:
This talk shall bring light in the topic on how the cost structure on todays Internet looks like.

Friday, 26. February 2016

Building secure SDN networks from untrusted components through the Robust Combiner approach
Speaker:
Philipp Heyder
Type:
Masterstudent introductory talk
Time:
26 February 2016
16:30–18:00
Place:
MAR 4.033
Abstract:
The so called SYNful Knock router implant created an alarming precedent: Network equipment may contain backdoors which grant unauthorized remote operators multiple ways to temper with a network, possibly without being noticed by the network's owner. This is particularly threatening to governments and other entities which do not have the resources to manufacture their own switching and routing equipment and thus have to rely on third party vendors. We believe the robust combiner approach, when applied to network switches can successfully identify these backdoors and render them harmless. We replace a single untrusted switch with a circuit of three parallel untrusted switches, ideally manufactured by different vendors. Since these three switches mimic a single one, their forwarding logic should be equal. Furthermore since all of them experience the same input traffic, their forwarding behaviour should be equal as well. An added middlebox application compares the output of all three switches. Only a packet which was received from a majority of the switches will be forwarded. We assume that if a packet is the product of a backdoor, it will differ from its benign counterparts and therefore will not be forwarded. In the long-term we believe a backdoored switch will reveal itself, by producing a significantly higher amount of unforwardable packets than its counterparts. We focus on Software Defined Network switches, because the separation of control and data plain aids our design. Two challenges will be tackled by our work: First we will create realistic threat models to test the security benefits of our design. Secondly we will need to weigh these benefits against the strain they put on overall network performance: Placing our robust combiner in a Fat Tree topology and measuring different performance characteristics under realistic traffic patterns will guide us to further improve our design. Finally we hope to present a solution, which is secure against our thread models, reasonably light on performance and simple enough to be trustworthy.

Tuesday, 23. February 2016

Measuring IPv6 Adoption by Crawling the BitTorrent DHT
Speaker:
Martin Ott (TU Berlin bachelor student)
Type:
Bachelorstudent final talk
Time:
23 February 2016
14:00–15:00
Place:
MAR 4.033
Abstract:
In this thesis, we develop a crawler for the BitTorrent Distributed Hashtable (DHT) in order to assess IPv6 adoption in today’s Internet. The developed crawler is capable of extracting IPv4 as well as IPv6 contact information of BitTorrent peers that participate in the global BitTorrent DHT. Using the gathered data, we present a study of IPv6 adoption as seen through the lens of the BitTorrent DHT. We compare the number of found IP addresses of each protocol and present a per-network comparison of IPv4 and IPv6. Lastly, we leverage a BitTorrent-specific unique node identifier to identify and match the IP addresses of likely dual-stacked hosts. Our results suggests that while the IPv6 DHT is still sparsely populated when compared to IPv4, our proposed method can provide a viable contribution to broaden our understanding of IPv6 adoption in the Internet.

Thursday, 04. February 2016

Dynamic Aspects of Network Virtualization - Algorithmic and Economic Opportunities
Speaker:
Dipl.-Inf. Arne Ludwig
Type:
Wissenschaftliche Aussprache zur Erlangung akademischen Grades „Doktor der Ingenieurwissenschaften“
Time:
04 February 2016
16:00–19:00
Place:
MAR 4.033
Der Promotionsausschuss setzt sich wie folgt zusammen:

Vorsitz:

Prof. Dr.-Ing. Nestmann

Gutachter:

Frau Prof. Feldmann, Ph.D.

Frau Prof. Richa, Ph.D. (Arizona State University, USA)

Prof. Dr.-Ing. Kellerer (TU München)

Prof. Dr. Schmid (Universität Aalborg, Dänemark)

Die Dissertation und die Gutachten können von Berechtigten nach § 8 Abs. 1 der Promotionsordnung in der Fakultätsverwaltung eingesehen werden. Die wissenschaftliche Aussprache ist universitätsöffentlich.

Tuesday, 12. January 2016

Temporal and Spatial Classification of Active IPv6 Addresses
Speaker:
Arthur Berger (Akamai / MIT) see bio below.
Type:
Invited Talk
Time:
12 January 2016
11:00-12:00
Place:
MAR 4.033
Abstract:

There is striking volume of World-Wide Web activity on IPv6 today. In early 2015, one large Content Distribution Network handles 50 billion IPv6 requests per day from hundreds of millions of IPv6 client addresses; billions of unique client addresses are observed per month.
Address counts, however, obscure the number of hosts with IPv6 connectivity to the global Internet. There are numerous address assignment and subnetting options in use; privacy addresses and dynamic subnet pools significantly inflate the number of active IPv6 addresses.

As the IPv6 address space is vast, it is infeasible to comprehensively probe every possible unicast IPv6 address. Thus, to survey the characteristics of IPv6 addressing, we perform a year-long passive measurement study, analyzing the IPv6 addresses gleaned from activity logs for all clients accessing a global CDN.

The goal of our work is to develop flexible classification and measurement methods for IPv6, motivated by the fact that its addresses are not merely more numerous; they are different in kind. We introduce the notion of classifying addresses and prefixes in two ways: (1) temporally, according to their instances of activity to discern which addresses can be considered stable; (2) spatially, according to the density or sparsity of aggregates in which active addresses reside. We present measurement and classification results numerically and visually that: provide details on IPv6 address use and structure in global operation across the past year; establish the efficacy of our classification methods; and demonstrate that such classification can clarify dimensions of the Internet that otherwise appear quite blurred by current IPv6 addressing practices.

Bio:

Arthur Berger is a Principal Research Scientist at Akamai Technologies and a Research Affiliate at the MIT Computer Science and Artificial Intelligence Laboratory. Dr. Berger received his PhD in applied mathematics from Harvard University in 1983 and then worked at Bell Labs, AT&T Labs, until 1999 when he joined Akamai. From 1990 to 1998, Dr. Berger represented the U.S. at the International Telecommunications Union and was a key contributor to standards on congestion controls and traffic engineering. Dr. Berger has done research in the areas of design and performance of high-speed data equipment and networks, congestion controls, quality of service, content delivery over the Internet, overlay routing, and Internet measurements.

Wednesday, 06. January 2016

Towards Improved Internet Infrastructure Geolocation
Speaker:
Dipl.-Inf. Carlo Fürst
Type:
Wissenschaftliche Aussprache zur Erlangung akademischen Grades „Doktor der Ingenieurwissenschaften“
Time:
06 January 2016
16:00–19:00
Place:
MAR 4.033
Der Promotionsausschuss setzt sich wie folgt zusammen:

Prof. Dr. Markl

Gutachter:

Frau Prof. Feldmann, Ph.D.

Prof. Dr. Michiardi, Ph.D. (EURECOM, Frankreich)

Dr. Costa (Imperial College London, UK)

Prof. Dr. Schmid (Universität Aalborg, Dänemark)

Die Dissertation und die Gutachten können von Berechtigten nach § 8 Abs. 1 der Promotionsordnung in der Fakultätsverwaltung eingesehen werden. Die wissenschaftliche Aussprache ist universitätsöffentlich.

Wednesday, 06. January 2016

Thinking Outside the (Network) Box
Speaker:
Paolo Costa (Microsoft Research UK)
Type:
Talk
Time:
06 January 2016
11:00–12:00
Place:
MAR 4.002
Abstract:
Data centers are the infrastructure providing access to online services such as Azure, Bing, and Office 365 for hundreds of millions of users around the world. They comprise millions of servers interconnected by a fast network fabric. The network is therefore a critical component of data centers and it is often cited as one of the main bottlenecks, affecting performance and costs.

Existing network deployments are heavily influenced by Internet-based technology. While this approach has served us well, it starts showing its limits. In this talk, I will review some of these shortcomings and discuss how the research undertaken in our group aims to address the challenges of future data centers through a deep rethinking of the way data center networks are built and operated.
Bio:
Paolo Costa is a researcher in the Systems and Networking group at Microsoft Research in Cambridge. In the past, he had been a faculty at Imperial College London and a Postdoctoral Researcher at Vrije Universiteit Amsterdam. He holds a M. Sc. and Ph.D. degree in Computer Science from the Politecnico di Milano.

His research interests lie at the intersection of distributed systems and networking with particular emphasis on large-scale networked systems, ranging from sensor and mobile networks to overlay networks and data centers. While at MSR, he has been working on the CamCube and Predictable Datacenters (PDC) projects, aiming to providing a better integration and synergy between applications and networks in data centers to improve performance and reduce complexity. His current research investigates how to improve data center networks through a deep rethinking of the way these networks are built and operated, leveraging the availability of new hardware technologies such as systems-on-chip (SoCs), distributed switching fabrics, and optical networks.

Zusatzinformationen / Extras

Quick Access:

Schnellnavigation zur Seite über Nummerneingabe

Auxiliary Functions