Wireless Resources Virtualization in
Cellular Networks
NVS is a substrate within the
basestation for virtualizing the wireless resources of a
cellular network, such as a WiMAX or an LTE network. Conference [ACM MobiCom 2010]    Journal [TON 2012] CellSlice
is an overlay scheduler designed for mobile core gateways to
achieve the functionality of NVS with minimal modifications to
the basestations. Conference [Comsnets
2013] NetShare
is designed as a novel two-level scheduler split between the mobile gateway and
the cellular basestations to effectively manage and allocate the
wireless resources of the radio access network composed of
multiple basestations among multiple different entities.
It leverages NVS as the link layer scheduler in
the basestations. Conference [ICNP 2013]
Virtualizing wireless resources in cellular
networks fosters several interesting deployment scenarios that
are of interest to both academia and industry:
Active RAN sharing: One radical change that is
receiving considerable attention recently is active RAN
(Radio Access Network) sharing among Mobile Network
Operators (MNOs). RAN sharing enables significant reduction
in equipment in low traffic areas and results in at-least
100% increased rollout speed with a given cost.
Mobile Virtual Network Operators (MVNO): In the
recent past, several MVNOs have emerged as strong players in
the cellular market providing enhanced services. Such MVNOs
often do not own spectrum and rely on sharing the wireless
resources of a MNO in that region. Virtualization helps
partition the wireless resources in an MNO’s network
effectively, thereby encouraging stricter and fine-grained
SLAs between MVNOs and MNOs.
Corporate Bundle Plans: Currently, network
operators offer data plans to enterprises and corporations
that allow sharing of bandwidth dynamically across their
users. However, no bandwidth guarantees are provided.
Virtualization may help realize better guarantees on
resource allocation, and hence fosters more sophisticated
data plans.
GENI and Beyond:
Virtualization can help researchers evaluate their new ideas
on commercial base stations and deployments.
Controlled evaluation: Virtualization enables MNOs
to isolate partial wireless resources to deploy and test
novel ideas without affecting the operational networks.
Currently, MNOs often use dedicated (small scale)
deployments to test new ideas.
Services with Leased
Networks(SLNs): With the increased use of wireless and
mobile networks for Internet services, we envision
application service providers reserving bandwidth with
MNOs and paying on the behalf of their users to enhance
users’ quality of experience. Virtualization helps in
ensuring that such reservations are met.
NVS and CellSlice enable effective wireless
resource virtualization by meeting the following crucial
requirements:
Isolation
of wireless network resources across groups of flows (or
slices)
Efficient utilization
of the wireless resources.
Customization
of flow management within individual slices.
To
achieve the above, both NVS and CellSlice contain two novel components:
Slice
Scheduling: NVS includes a simple and optimal slice scheduling
algorithm that is light-weight to implement. It enables
co-existence of both bandwidth and resource based slice
provisioning: with bandwidth based reservation, slices can
reserve certain amount of bandwidth in Mbps on a cellular base
station, while resource based reservation allows slices to
reserve a percentage of the resource slots in the OFDMA
frames.
Flow
Scheduling Framework: NVS enables individual slices to employ custom flow
scheduling policies through a flexible framework.
We
instantiate the NVS and CellSlice prototypes on a WiMAX testbed
consisting of:
A Picochip’s IEEE 802.16e (WiMAX) compliant
Base station.
A Profile-C ASN-gateway installed on a Linux
Desktop.
Six Netbooks with Intel WiMAX mini PCI-X
cards.
A Video server that streams videos to the
clients.
The following demos show the
efficacy of NVS
on a WiMAX testbed.
In this
demo, we configure
the virtualized
WiMAX base station
for the
co-existence of
two slices with
resource-based
reservation. SLICE
1 and 2 reserve
50% of the total
resources each.
While SLICE 2
employs
traditional
scheduling like
round robin and
prop fairness
across its flows,
SLICE 1 performs
dynamic admission
control and
advertising to
improve QoE for
its users.
In the demo video, the left three netbooks
are in SLICE 1 and
the right three
are in SLICE 2.
The demo shows
that all clients
in SLICE 2 suffer
similar
performance
degradation,
whereas one client
in SLICE 1 is
shown an
advertisement in
the interest of
maximizing the
Quality of
Experience of the
other two users.
The demo clearly
shows two
customized
resource
management
policies running
within slices on a
single base
station
simultaneously.
This scenario
represents two
MVNOs on the same
MNO using
different flow
management schemes
for its respective
users. It also
represents two
experimenters
sharing the same
testbed for
conducting
scheduler
research.