To:  new or prospective students entering our PhD or MS program who are interested in computer networking.

Please refer to the latest PhD and MS program requirements as this provides the details  that you must meet to get your degree. 

This page briefly summarizes my approach to working with PHD students.  I begin by sharing my expectations and then I go into more detail as to typical paths for a successful PHD student.

EXPECTATIONS: 
Dissertation research in the systems areas is usually quite challenging as the amount of background material that must be read and understood is extremely large.  By roughly your last year in the PHD process,  you should be more knowledgeable in the very narrow problem you end up addressing.  When you reach this level, completing your dissertation usually goes smoothly.  If you appear to continuously 'stuck' and not able to make progress without significant assistance from me,   and this continues as you approach the later stages described below, I will inform you that you are not making sufficient progress - we would brainstorm to come  up with a plan that would be designed to help.  After another 6 months or so,  if you still are 'stuck' - I reserve the right to tell you that a PHD is not the right path for you.   If you have successfully accomplished the items mentioned below, by the time you defend your dissertation to your committee,  you need to be able to clearly articulate the motivations, the specific problem statement,  your methodology,  and your results and analysis that supports your 'thesis statement'.  As long as you have good publications and appear to know what you are talking about when we have committee reviews, that is usually sufficient for the committee members.  My expectations are a bit higher:  The nature of systems research unfortunately leads to quite a bit of very poor research that ends up being published.   By the time you defend your work to the committee,  it should be crystal clear that you are one of the leading researchers in the world on your topic.  In the past, I've had to send a PHD candidate's work to colleagues (outside of Clemson) to get their opinion of your work.  By the time you are ready to defend, if you have followed my guidance, there is a 95% chance the committee will sign the forms saying PASSED.   If you have not followed my guidance, I most likely would have told you much earlier that you should no longer pursue a PhD.  If you did not follow my guidance and by some chance you do make it to your defense,  you likely will not pass.  Or, more likely,  you will have to do additional work to address the committee's comments and try to defend once more.  That might cost you a couple of months but no more than 6 months.   If more than 6 months is needed,  I would likely tell you that you need to move on (without the degree).  

ILLUSTRATIVE WORK FLOW:

Any graduate student working in my Networking Lab must have a background in systems modeling and analysis  AND have a deep understanding of operating systems, specifically Linux.   Students in my lab are trained to follow two paths concurrently: 

• Systems performance and modeling: 
  
• Basic level (level 1) :  probability and statistics,   experimental statistics
  
• Level 2 :   random processes (stochastic systems), system performance analysis using at least 2 of the 4 approaches:
• Measurement-based evaluation performed within our lab's controlled ethernet and wireless networks.
• Measurement-based evaluation enhanced with some form of emulation.
• Linux's tc (traffic control) allows a Linux system to operate as a fully functional router with additional support from tc allowing an admin to control or manage how each flow is treated from a resource allocation (scheduling) perspective.
• miniNet, CORE, EMANE are other examples.
• Level 3:  Performance evaluation using simulation.
• Our lab uses ns2, ns3, VSIM, or matlab.  The latter example typically requires our students to develop a simulation model using matlab.
• Simulation models must be validated by comparing simple models with baseline data based on a measurement activity using a prototype or a production level system.
• Before moving on to Level 4,  you must have at least two papers published (or at least accepted). One should be a conference paper, the other should be a journal publication.  
  
• The nature of these two papers would be: Performance Analysis of a system that exists.  Here is one example:  Problem statement:  We conjecture that 5G operators will realize the conflict between ensuring low latency bounds with economics.  Operators plan to support large numbers of containers that are used by different stakeholders.   As the number of organizations (i.e., container instances) the low latency bound will have to increase. One direction for a study would require the following
• A detailed literature survey of past and ongoing research in the area of 5G MECs.  There are well known problems at this point with no widely accepted solutions.  Finding these open problems is the goal of the literature review.
• Identify a very specific problem that can be expressed (at a high level) in a single sentence.  Call this the 'thesis statement' of the paper. 
  
• Example:
•  Vehicle/machines that communicate with peers to cooperatively perform tasks pose significant problems if some applications in the system require low latency and high reliability. A portion of the academic community that study this area assume infrastructure is required to reduce the processing and subsequently lowers power consumed by vehicles.  Interesting problems include innovative techniques for ensuring latency bounds,  highly mobile machines leads to  problems related to 'hand-offs' of a container's state being migrated from one MEC to another. 
  
• Top conferences:   SIGCOM, SIGMetrics, IEEE flagship conferences:  Infocom,  ICC,  ICNP,  or an appropriate Usenix Association conference.
  
• Top journals:   any of the IEEE's flagship journals and transactions.
  
• Level 4:   Advanced modeling
  
• Your third step in your journey towards getting your degree will require you to do at least the following two items:
• First, you will need to take advanced courses in random processes and optimization.
  
• Second, you will need to expand the scope of the system and after further exploration, identify a very specific problem that is critical to the larger system.
• Third, you will need to come up with an innovative solution to the problem and conduct a thorough performance study validating your ideas.
  
• The e expand the scope of the system that is of interest, while at the same time narrow down the focus to find a specific problem. 
  
• Level 4 must result in another pair of papers (conference paper and then an extended paper that published in a top tier journal. 
  
• Become an expert in a particular academic area in the broad direction of systems.  You will get some of the foundations through careful selection of your classes, through the first 3 levels described above, and from your committee members.  My role as your committee chair will be to guide you through these levels.  However, once you are at Level 4,  you should be able to make progress independently.
• With a new PHD student,   I typically will have you assist other researchers in the lab along with one-on-one meetings with me.  The first couple years usually involves quite a bit of reading of related papers that have been published. 
  
• Given the complexity of any system that might be of interest, you will have to keep up with what is published in the top conference venues -  this is the most effective way for you to ensure you are reading the correct set of papers.   This
  

Knowing in advance what I am looking for will help guide your graduate course selection.

• Attention students.....
  
• The Networking Lab in the School of Computing is looking for PHD students with a background and/or interest in systems, modeling, system analysis and optimization, and distributed computing.  But most importantly, we are looking for students with the curiosity, passion and drive to help us, working with industry leaders and supported by multiple federal government agencies, shape how Internet access will evolve over the next 10 years. 
  
• Keep you eyes open for updated versions of  core CS classes such as CPSC3600, CPSC4240/6240, CPSC8510, CPSC8520.  Aspects of the research performed by our lab research team will be introduced into these courses the next several semesters.
  
• Undergrads, consider taking our lab's Creative Inquiry, called CyberTiger  or asking about Research Experience for Undergrads (REU) programs.

LAB RESEARCH:

• Emerging Application Systems (APPSYS): We are developing an information-centric system that supports emerging Application Systems  (APPSYS).  An APPSYS  involves a group of mobile, cooperative machines that are tasked to perform work.  Early examples of APPSYS's include
•  A group of cooperating autonomous vehicles that collectively are able to offer benefits including higher traffic flow rates,  reduced power consumption while reducing the systems probability of accidents potentially by an order of magnitude (note:  , and a much lower rate of attempt to improve safety by working as a group;  or a drone swarm tasked with a mission (such as Coordinated Search and Tracking). 
  
• An APPSYS requires nodes to dynamically form working groups and leverage whatever computation and communications capabilities that are collectively possible in a fair manner.   The concept  differs from 'Edge Computing' systems in that each node might be capable of serving as an edge node.    One way to view an APPSYS is a group of nodes that form an adhoc cloud that collectively provide services that can support distributed applications that need to operate 'out in the wild'.
  
•  Government funding agencies have invested billions of dollars in adhoc wireless networks.  The technology has seen only small scale, niche use-case adoption in society, public safety, and the military. The APPSYS concept borrows some of the Internets architectural model, that of separate autonomous systems cooperate and internetwork.  The result is  a unified network that does not much more than transport data to the destination network.  In some cases an APPSYS is similar to the Internet's autonomous system.  In other cases, an APPSYS is more open and can support any node that respects the protocols and policies.  In other words,  an APPSYS might appear in a geographic area that has one or more APPSYS in operation.  The problem with wireless is the baggage such as access rights to spectrum and decades old protocols that are still the basis for accessing spectrum.   Our direction involves an architecture that is inherently information centric.  One open research issue that we are currently exploring is to  develop an appropriate architecture for group communications that blends application level publish and subscribe with traditional wireless group communications such as broadcast and wireless multicast. 
  


•  Future broadband access:  Over-the-air television content still exists, as does the broadcast of cable content over coaxial cable.  These decades old methods are quickly being replaced by streamed content from a content provider using a broadband Internet connection.  While cable-based Internet access is available in top-tier markets, the current COVID crisis tells us that a large percentage of the population have inadequate Internet connectivity.  We are explore the following:
  


• The contention between cable,  wifi, and 3GPP -  historically these represent three separate worlds.  To go the next step,  we must redefine heterogeneous broadband access.
  
• One game changer due to hit in 2021 will be Internet access through constellations of LEO Cubesats.
  
• Dense constellations of LEO Cubesats that can provide broadband access to any point on the globe. If things go as planned, this will be a nuclear blast to the current economic models driving cellular and Internet access as it instantly could provide competition anywhere.
  




• Cybersecurity:  The DHS has identified 17 industry sectors that represent our Nation's Critical Infrastructure Sectors.  The majority of these sectors, such as the transportation system, the financial system, energy systems and agriculture systems are at various stages in a dangerous trend:  Becoming dependent and consequently vulnerable  on commodity, commercial wireless systems.   Specific examples:
  
• Government agencies such as the DOJ rely on a commercial 3GPP operator to manage the nation's first responder's  FirstNet public safety broadband network.
• Deployed military personnel claim that commercial 3GPP systems are essential to their work. 
  
• We are proactively developing an architectural model that requires a minimal set of functionality required by any Autonomous Wireless System, inclusive of 3GPP and public WiFi network operators,  along with minimal National wireless support infrastructure operated by the FCC (or equivalent) that oversees the rules. 
  

 

  Page last updated: 1/24/2021