You are on page 1of 1

Minimizing Flow Completion Times using Adaptive Routing over

Inter-Datacenter Wide Area Networks


Introduction and Definitions Microsoft Global WAN (datacenters in 38 regions)
I Throughput-oriented flows deliver large volumes of data.
I Less sensitive to propagation and initial routing latency.
I Inter-datacenter networks managed by one organization allow for
flexible application of custom routing techniques.
I Focus on single path routing and aim to minimize completion times
and bandwidth usage.
I Adaptive flow routing according to network and flow properties.
I Fast heuristic schemes use a cost (distance) metric and select the
minimum cost (shortest) path.

Path Selection Heuristics


I Three edge cost metrics: load, load+demand, utilization
I load: total remaining bytes to be sent on a link (given flows on it) https://azure.microsoft.com/en-us/blog/
I demand: total bytes of the new flow (its size) how-microsoft-builds-its-fast-and-reliable-global-network/
I utilization: current fraction of capacity in use on a link
I Two path cost functions of MINMAX() and MINSUM()
Cogent WAN (197 nodes and 243 edges)
I Extensively used for traffic engineering: MINMAX(utilization)

Simulation Setup
I Cogent WAN (uniform capacity of 1.0 for all links).
I Four flow demand distributions of light-tailed, heavy-tailed,
Cache-Follower and Hadoop (last two reported by Facebook).
I Poisson distribution with rate λ = 1.0 for flow arrivals.
I Flow demands with an average of 20 units and a maximum of 500
units (heavy-tailed had a minimum demand of 2 units).
I Three scheduling policies of FCFS, SRPT and Fair Sharing using
Max-Min Fairness (MMF). Topology Info: http://www.topology-zoo.org/files/Cogentco.gml

Simulation Results
MFCT TFCT Total Bandwidth Used
Light-tailed Heavy-tailed Light-tailed Heavy-tailed Light-tailed Heavy-tailed
F S M F S M F S M F S M F S M F S M
MINSUM(load)
MINMAX(load)
MINSUM(load+demand)
MINMAX(load+demand)
MINSUM(utilization)
MINMAX(utilization)
MinHop

MFCT TFCT Total Bandwidth Used


Cache-Follower Hadoop Cache-Follower Hadoop Cache-Follower Hadoop
F S M F S M F S M F S M F S M F S M
MINSUM(load)
MINMAX(load)
MINSUM(load+demand)
MINMAX(load+demand)
MINSUM(utilization)
MINMAX(utilization)
MinHop
< 10% from min < 20% from min < 30% from min < 40% from min < 50% from min ≥ 50% from min

Fig. 1. Performance of various cost metrics for path selection over Cogent WAN [6], with uniform capacity of 1 and λ = 1.0 (F , S and M represent the
Figure: Performance of various
FCFS, SRPT and MMFcostscheduling
metrics for path selection
policies, oversimulation
respectively), Cogent WAN (F , S and
was repeated manyM represent
times the FCFS,
and average SRPT and MMF scheduling policies, respectively)
was computed

Discussion Future Directions Online


Schemes based on utilization are at least 40% above the mini- is also an effective metric for selection of multicast forwarding
I Schemes based mumonfor utilization are atofleast
the majority 40% above
scenarios. Also, the
MINMAX(load)I Software treesDefined Networking
that reduce completion cantimes
be used for realization
via load of [8].
balancing [7],
minimum for the
and majority of scenarios.
MINMAX(load+demand) are more than 50% above MINSUM(load+demand)
It is also interesting tohaving access
note that to a global view of
MINMAX(utilization),
I MINMAX(load) and MINMAX(load+demand)
the minimum in mean completion times for multiple
are 50% or network
scenar- status
which and flow used
is frequently demands.
in traffic engineering research, is far
ios. Overall,
more above minimum it cantimes
in mean be seen that schemes
for multiple based on I
scenarios. “load” from the
Distributed best solution for
implementation viathe majorityand
end-point of evaluated
switch scenarios.
I Schemes based as link cost offer
on “load” offermuch
muchbetter
bettertail
tail completion
times (less times (less
cooperationCentralized frameworks,
(flow demand such in
encoded as packet
SDN [9], are good candi-
header).
than 10% away thanfrom away from
10%minimum for minimum
majority offorcases).
majority of cases).IAlso,
In case dates
exactfor flow
realization
size is of this scheme
unknown, since theycan
an estimate offer
beaccess to
MINSUM(load+demand)
I MINSUM(load+demand) offers the bestoffers
meanthecompletion
best mean completionused.global
Further view of network
research status and
is needed flow flow
on how demands.
demandTo properly
times all
times considering considering
scenarios. all scenarios. updateaccuracy
estimation load variables associated
can affect with
quality links, onepaths.
of selected needs knowl-
Total Bandwidth
I MINSUM(load+demand) offers Usage: MINSUM(load+demand)
the minimum extra of- edge and
I Extending of flow demands.
evaluating In case exact flow size isforunknown,
MINSUM(load+demand) [Poster Abstract]
fers compared
bandwidth usage the minimum extra which
to MinHop bandwidth usage
is below compared multipath
20%. to an estimate
routing can
(what be demand
used. Further
to useresearch is needed on how
per sub-flow).
MinHop which is below 20% at all times. Schemes flow demand estimation accuracy can affect quality of selected
based on MINMAX() consume at least 40% extra band- paths. In addition, we plan to extend and evaluate our proposed
width. MINSUM(load) and
Mohammad Noormohammadpour andMINSUM(utilization)
Cauligi S. Raghavendra use adaptive approach for multipath traffic engineering.
at ofleast more bandwidth
10% Engineering, at all times compared to of Southern California
Ming Hsieh Department Electrical Viterbi School of Engineering, University R EFERENCES

You might also like