[Ns-developers] GSoC 2012 NS-3 LTE scheduler

Dizhi Zhou q5frc at unb.ca
Thu Mar 29 06:00:07 PDT 2012


Hi Nicola,

Sorry for late response. I didn't receive this email and just found it in the maillist in this morning.

The reason for me to choose LTE is that I will consider LTE MAC layer
as lower layer protocol for my multipath transport layer solution.
I did a 12 months LTE project with Huawei in 2010,a leading global
information and communications technology solutions provider. This
project aims to propose IETF standard and patent about LTE core network
(I hold two patents as first author from this project). So I'm also
familiar with the basic knowledge of LTE, both core network and access
network (E-UTRAN). In the past two weeks, I read lots of papers and
standard about LTE MAC scheduler and already have some idea for this
project. The approach part of my proposal can be found at NS-3 maillist:

http://mailman.isi.edu/pipermail/ns-developers/2012-March/010181.html

Your suggestions and feedbacks will be highly appreciated.

Currently, I'm working on the TCP scheduler for multipath transmission.
Yes, scheduler for transport layer and MAC layer is different. However,
their principle is very similar. The LTE MAC scheduler aims to decides
which UEs are to be scheduled(varied priority metrics) and which
resources(TD or FD or mixed) to be allocated to the scheduled UEs. On
the other hand, TCP multipath scheduler also tries to decide which packets
are to be scheduled and which resources (path, bandwidth, QoS) to be
allocated to scheduled packets. The research experience of scheduler
in transport layer can help me a lot to easily and quickly understand
varied LTE MAC schedulers.

Yes, the LTE standard only defines the API for LTE MAC scheduler. So the
MAC scheduler can be seen as a plugin model for MAC layer. I list my
principle to choose potential MAC scheduler in the proposal which contains
two parts:
1, Basic scheduler which can be easily extend to advanced algorithm
2, Scheduler for time domain OR frequency domain

Looking forward to your reply!

Best regards
Dizhi

///////////////////original mail/////////////////////////////
On 03/26/2012 02:01 AM, Dizhi Zhou wrote:
>/  Dear mentors,
/>/
/>/  I'm a second year Ph.D student in UNB, Canada and my research focuses on
/>/  multipath video delivery in wireless network. I have three years NS-2
/>/  development experience.
/>/  In this week, I went through the related documents of several projects
/>/  I'm interesting, including SCTP, PackMime-HTTP, Tmix and LTE scheduler.
/>/  In the end, I decide to select LTE scheduler to write my proposal.
/
I am glad of your interest for the LTE Scheduler project. Anyway, out of
curiosity, why did you choose this project instead of the SCTP project?
  From what you told us about you, I would say that you probably have a
good background on SCTP, whereas your background on LTE is not clear. I
suggest that, in order to carry out successfully the LTE scheduler
project, a good knowledge of LTE is required.


>/  In recent two months, I'm doing the research project about application
/>/  aware packet scheduler for multipath TCP and SCTP. So, I read lots of
/>/  scheduler algorithms in the transport
/>/  layer. After this projects, I will consider the cross-layer scheduler
/>/  algorithm that considers the mac layer parameters. Therefore, LTE
/>/  scheduler project in NS-3 is very close to
/>/  what I'm doing now.
/
Can you please comment a bit more in detail what do you think are the
similarities between transport layer schedulers and LTE schedulers? Do
you expect that there will be some differences?


>/  Before discussing the detail, I have a question here, in NS-3 idea list,
/>/  it says that we need to implement an existing and well-defined
/>/  FemtoForum MAC Scheduler algorithm.
/>/  However, what is the definition of well-defined here? Some algorithms
/>/  published in high level journal? Or algorithms recommended in standard?
/
The 3GPP standards do not specify any scheduling algorithm, because it
is left open to manufacturers to design and implement customized
schedulers for differentiating their products. "Well-defined scheduler
algorithms" in the ns-3 GSoC ideas page refers mainly to algorithms
previously published in a scientific publication. Whether it comes from
a conference paper, a journal or a thesis does not matter much, what
really matters is that the algorithm itself is technically sound,
suitable for LTE, and already defined. In other words, we don't like
that you are trying to invent your own algorithm as part of the GSoC,
because if you were to do so the chances of succeeding would be very
low, due to the limited timespan of the GSoC.

Regards,

Nicola

-- 
Dizhi Zhou
Ph.D. Candidate
Faculty of Computer Science
University of New Brunswick
540 Windsor Street
Fredericton,New Brunswick,Canada
E3B 5A3

E. q5frc at unb.ca
Homepage: www.cs.unb.ca/~q5frc/




More information about the Ns-developers mailing list