Tuesday, May 28

TUE 11:15 - 12:30 P1: Management of Big Data - Big Data for Management
Chair: Rolf Stadler (KTH)
Room: Refter
Panelists: Vasanth Bala (IBM), Marcus Brunner (Swisscom), Alexander Clemm (Cisco Systems), Sven van der Meer (Ericsson)
Abstract: Big Data is the next disrupting technology wave that has reached critical mass through advances in low-cost networked computing and storage, inexpensive intelligent sensors, and analysis techniques. Many disciplines, from commerce to science and public health, will become data-driven, data-centric, and data-intensive. This panel will assess and debate the implications, opportunities and challenges of the Big Data paradigm for network and service management and for operations. One aspect is that new technologies are needed to manage the infrastructure and operations of Big Data processing and analytics. Second, Big Data technologies can potentially harness the large amount of operational data that is constantly produced in networks/clouds, perform analytics, possibly in-network, to improve real-time fault, performance, and security management.

Wednesday, May 29

WED 17:00 - 18:15 P2: Diagnosing and Troubleshooting Problems in the Cloud
Chair: J.P. Martin-Flatin (EPFL)
Room: Rector Vermeylen
Panelists: David Breitgand (IBM), Raju Rajan (Hewlett-Packard), Helmut Reiser (Leibniz Supercomputing Center), Sebastian Wenner (T-Systems)

Abstract: Public and hybrid clouds have become increasingly successful in recent years. It used to be customary for SaaS services to be deployed in hosted environments or internally. But these days, many SaaS services use public clouds instead. For instance, some providers build their SaaS on top of an Infrastructure-as-a-Service (IaaS) such as Amazon Web Services, or a Platform-as-a-Service (PaaS) such as Heroku or Google App Engine, and use optional add-ons such as database services. Businesses now often use hybrid clouds that integrate SaaS services deployed in public clouds with legacy applications deployed in private clouds. And corporations usually need to integrate applications deployed and supported by fairly independent business units (federated clouds).

Troubleshooting problems that affect end-users is complex in such environments, because software and hardware are not entirely controlled by the same NOC/IT Department. Multiple service providers or business units are involved, monitoring information is siloed, management domain boundaries come in the way of root cause analysis, and cooperation sometimes gives way to finger-pointing.

During this panel, we will characterize the underlying issues, suggest possible solutions, and present research challenges to the community.

Thursday, May 30

THU 11:00 - 12:45 P3: Software Defined Networks - Dumb, Smart or Both?
Chair: Bruno Klauser (Cisco)
Room: Refter
Panelists: Michael H. Behringer (Cisco Systems), Axel Clauberg (Deutsche Telekom), Paul Glynn (Davra Networks), George Pavlou (University College London)

Abstract: Software Defined Networks is one of the hottest recent terms in the networking buzzword universe. The race for killer applications and use cases seems open, while the real meaning of SDN it is still an emerging story. The underlying problem statements, motivations and hopes however have been seen before – or maybe not?

One school of thought holds that SDNs are essentially a resurgence of the "dumb" network, in which smarts are provided by controllers and the network itself is reduced to fairly dumb pipes. Isn’t this just recreating the same solutions to the same problems as in traditional networks but a slightly different form factor? Does a new management and control plane justify the effort?

Another school of thought holds that SDNs are fundamentally about making the network more programmable and hence ultimately host more intelligence than today’s networks, not less. Doesn’t this contradict the objective in the first place? What applications are programming the network if not the management and control plane of a central controller?

So, which one is it going to be? Could it be a hybrid or a third one? Do SDNs foreshadow the re- emergence of a glorified Internet-age version of the PSTN, concentrating the smarts in a few very powerful controllers? Or are they an expression of devices becoming increasingly smart and independent? Are both perspectives as contradictory as they seem? Or do they complement each other as two perspectives of the same thing?

THU 14:15 - 15:30 P4: The Race Against the Machine: Automating the Human Manager
Chair: Nikos Anerousis (IBM)
Room: Refter
Panelists: Joseph Betser (The Aerospace Corporation), Raouf Boutaba (University of Waterloo), Mark Burgess (CFEngine), Larisa Shwartz (IBM)

Abstract: Why do we still rely on humans to perform the most critical tasks in server and network administration? In an age when almost everything can be automated, humans still rule across many areas of infrastructure management, sometimes performing manual and time-consuming tasks. Despite countless predictions for the complete domination of autonomic computing and the replacement of humans by machines, it seems that automation is not fully ready yet - but it's getting close. Now more than ever, there are numerous initiatives to address the automation needs of IT infrastructure: from popular OpenSource packages like CFengine, Puppet and Chef, to complex enterprise suites from IBM, HP and others, there is plenty of automation to suit every need. Yet, despite range of available choices, human administrators are still in control.

This panel will investigate the evolving role of a human manager in an increasingly automated environment, by assembling experts representing infrastructure automation, IT support organizations, human-computer interaction and academia. We will discuss the many challenges behind automation - particularly the areas where automation falls short, and therefore where humans are still critical; but also where humans need to eventually be replaced by machines - particularly in situations with significant risks of human error, or where scale is absolutely critical - as is the case in emerging Cloud environments.