Things You Can Do in Budapest During Summer Time
description
Transcript of Things You Can Do in Budapest During Summer Time
Things You Can Do in Budapest During Summer Time
Aleksandar Kuzmanovic
Rice University & Ericsson Traffic LabOctober 2001
Rice University & Ericsson Traffic Lab
Budapest
• Buda-Pest– History– Architecture– Decadency…
• Ericsson Traffic Lab– Unix– Young people– Security
Rice University & Ericsson Traffic Lab
People
• Ericsson Traffic Lab, Budapest– Andras Veres
• EMULab, The University of Melbourne– Attila Pasztor, Darryl Veitch
• Budapest University, Department of Physics– Gabor Simon, Gabor Vattay
Rice University & Ericsson Traffic Lab
My Goals
• Modeling and reality?
Rice University & Ericsson Traffic Lab
Active Probing
Internal network inaccessible
• Poor understanding of origins of complex network dynamics
• Statistics is like the bikini…
Goal: Estimate the sample path of cross-traffic
Rice University & Ericsson Traffic Lab
Background
• Precise measurement infrastructureO
ne w
ay d
ela
y (
sec)
->
Packet departure time (sec) ->
Rice University & Ericsson Traffic Lab
Probing Uncertainty Principle
• Should not allow queue to empty between probe packets
• Small T for accurate measurements– but probe traffic would disturb
cross-traffic (and overflow bottleneck buffer!)
• Larger T leads to measurement uncertainties– queue could empty between probes
Rice University & Ericsson Traffic Lab
Theory
• Lindley’s equation
• CT information imbedded in delay• Ideal case: minimally backlogging condition
[QK99]
)( 11 iiii txww
Rice University & Ericsson Traffic Lab
Design Space
• No time synchronization between end points $ 2.000.000.000
• Delay difference
• Non-intrusiveness
• Stability
Time (sec) ->
On
e w
ay d
ela
y (
sec)
->
Rice University & Ericsson Traffic Lab
Probing schemeS e nde r R e c e ive r
U D P
T C P
T im e
One
way
del
ay
Rice University & Ericsson Traffic Lab
Aside…
• Experiments on LAN• Time-sharing in Linux
Time (sec) ->
Inte
r-arr
ival ti
me (
sec)
->
Rice University & Ericsson Traffic Lab
Traffic Lab – Experiment Setup
• Tcpdump, iperf (HTTP, TCP), sender, receiver…
e thr1 (1 9 5 .2 2 8 .1 3 9 .6 5 )
m o d e m p o o l
256 K b p s
Rice University & Ericsson Traffic Lab
Routes
• Tech. University Budapest6 nec2.ttt.bme.hu (152.66.247.2) 14.980 ms 14.212 ms 14.252 ms
• Columbia University, NY16 bongo.comet.columbia.edu (128.59. 64.210) 150.725 ms * 127.967 ms
• University of Melbourne23 potoroo.ee.mu.OZ.AU (128.250.76.186) 384.271 ms 383.408 ms 382.977 ms
• Houston, we have a problem… ->Time (sec) ->
On
e w
ay d
ela
y (
sec)
->
Rice University & Ericsson Traffic Lab
Wide Area vs. Metropolitan Area Network
Rice University & Ericsson Traffic Lab
Cross Traffic EstimationSample Path
• Differentiate delay, know C
• Cross traffic: from: Budapest to: NY iperf - 1 TCP
• Probing traffic from: Budapest to: Melbourne
Rice University & Ericsson Traffic Lab
Cross Traffic EstimationMoving Average
Rice University & Ericsson Traffic Lab
Signs
• Delay difference
• => two queues
• => two casesQ 1 Q 2
Rice University & Ericsson Traffic Lab
Wide Area Network revisited
• Differentiation btw. primary (256K) & secondary (transatlantic) bottlenecks
• Secondary traffic underestimated for Csec./256K
Rice University & Ericsson Traffic Lab
Primary & Secondary Bottlenecks
• LAN traffic controllable
• Separation between primary (256K)& secondary (LAN) cross traffic
Rice University & Ericsson Traffic Lab
TCP Cross Traffic
• Iperf (number of TCP connections >=3)
• => delay decreases yet we see clusters of packet losses (queue size ~ delay~pck. loss?)
Rice University & Ericsson Traffic Lab
Role of packet sizes
• => change packet size
Rice University & Ericsson Traffic Lab
Bit/sec. vs. Packet/sec.
• Up: 540 Bytes• Down: 60 Bytes• 540/60 = 9
– 256Kbps*9= 2.3Mbps
• UDP: 1500/30=50– 256K*50=
12.8Mbps - 1.5M*50= 75Mbps
Rice University & Ericsson Traffic Lab
Secondary Bottleneck
• Difference of CT estimate and tcpdump CT followed by clusters of packet losses
P ro b e s : E ric s s o n -A u s tra lia
C ro s s tra ffic : E ric s s o n - N Y (tc p d u m p )C ro s s tra ffic e s tim a te
Rice University & Ericsson Traffic Lab
Role of TCP
• Secondary CT (necessary for losses*) picks up the periodicity of probe traffic!
– Through losses– Through delay
• When phases match– Long burst of small packets causes losses on secondary bottleneck
P ro b e s : E ric s s o n -A u s tra lia
C ro s s tra ffic : E ric s s o n - N YS e c o n d a ry c ro s s tra ffic
Rice University & Ericsson Traffic Lab
Ongoing work
• Short and long lived TCP connections• Heterogeneous TCP traffic…
P ro b e s : E ric s s o n -A u s tra lia
C ro s s tra ffic : E ric s s o n - T e c h . U n iv .
P ro b e s : E ric s s o n -A u s tra lia
C ro s s tra ffic 1 : E ric s s o n - N YC ro s s tra ffic 2 : E ric s s o n - B u d a p e s t
Rice University & Ericsson Traffic Lab
Conclusions
• Scalable, edge-based tool for on-line network analysis, modeling, and measurement
• Scheme for estimating sample path of adaptive cross-traffic
• Differentiation between primary and secondary bottlenecks: two queue model
• Role of TCP (delay variation) and variable packet sizes– Low bit rate denial of service attack?– QoS in access networks (xDSL…)
Rice University & Ericsson Traffic Lab
The End
Rice University & Ericsson Traffic Lab
One more picture