Challenges in the Design and Evaluation of Content-Centric Networks

thoughtlessskytopNetworking and Communications

Oct 29, 2013 (3 years and 7 months ago)

58 views

Challenges in the Design and
Evaluation
of Content
-
Centric Networks

Jim
Kurose

Department of Computer Science

University of Massachusetts

Amherst MA USA




Visiting Scientist, Technicolor Paris Lab

Professeur

Invite, LINCS


Sigcomm ICN Workshop, Aug 2013


Overview


architecture, system, prototype


networks

of caches


challenges


approximation algorithms


n
etwork calculus for cache networks


w
orkload: logical mobility among networks

Architecture, System, Prototype

architecture


system


prototype

(realization)


high
-
level design/structuring principles, service/function
modularity

i
nstantiated set of interoperating protocols, mechanisms,
platforms conforming to design principles

Implemented (sub)set of protocols, platforms in particular
existing technologies

guides, informs, inspires, constrains

“here’s what it does (function), you
tell me how”

*

*
a
ck
: D. Clark, J.
Wroclawksi

Architecture, System, Prototype

architecture


system


prototype


Minimalist


principles,
modularities

Maximalist


protocols,
mechanisms go here

Architecture, System, Prototype

architecture


system


prototype


Internet

t
elephony

ICN

end
-
end circuit,
stateless endpoints,
stateful core, QoS,
single service

SS7, ESS, MSC,
VLR, HLD

many .. over the
years

datagram, stateful
endpoints, best
-
effort
stateless core,
multiple services, IP

TCP, UDP, DNS,
BGP, IS
-
IS, OSPF

many .. over the
years

content,
naming, stateful
core, caching

Ongoing

(routing, congestion
control, caching,
name resolution,
search,…)

Ramping up …

Architecture, System, Prototype

architecture


system


prototype


Internet

t
elephony

ICN

end
-
end circuit,
stateless endpoints,
stateful core, QoS,
single service

SS7, ESS, MSC,
VLR, HLD

many .. over the
years

datagram, stateful
endpoints, best
-
effort
stateless core,
multiple services, IP

TCP, UDP, DNS,
BGP, IS
-
IS, OSPF

many .. over the
years

content, stateful
core, caching

Ongoing

(routing, congestion
control, caching,
name resolution,
search,…)

Ramping up …

Architecture, System, Prototype

architecture


system


prototype


Internet

ICN

datagram, stateful
endpoints, best
-
effort
stateless core,
multiple services, IP

TCP, UDP, DNS,
BGP, IS
-
IS, OSPF

content, stateful
core, caching

Ongoing

(routing, congestion
control, caching,
name resolution,
search,…)

Architecture research/evaluation more difficult, more
“fundamental” (?), more impactful (?)

Research/evaluation mostly here: mechanism, protocols

Architecture, System, Prototype

architecture


system


Internet

t
elephony

ICN

end
-
end circuit,
stateless endpoints,
stateful core, QoS,
single service

SS7, ESS, MSC,
VLR, HLD

datagram, stateful
endpoints, best
-
effort
stateless core,
multiple services

TCP, UDP, DNS,
BGP, IS
-
IS, OSPF

content, stateful
core, caching

Ongoing

(routing, congestion
control, caching,
name resolution,
search,…)

Kleinrock

64

Cerf, Kahn 74


Clark 1988


McCanne

1998

?

?

Blocking

networks


Queueing networks

Delay calculus

Effective bandwidths

TCP, NUM optimization


Evaluation

?


ICN: architecture
vs

protocol/mechanism


elucidation, value
-
proposition of design principles,
service/function
modularities


e
valuation tool: networks of caches


Challenges

Interlude ….

Overview


architecture, system, prototype


networks

of caches


challenges


approximation algorithms


n
etwork calculus for cache networks


w
orkload: logical mobility among networks

Cache networks

miss

content

miss

miss

miss

miss


consumer
requests content


request
routed

(e.g., shortest
path) to known
content

custodian


en
-
route to custodian,
cache
s

inspect

request


hit:

return
local copy



miss
:
forward request
towards custodian


during content download,
store

in caches

along path


content request
s

from different users
interact:
cache replacement

request

content

custodian

content

custodian

miss

miss

Challenge: networks of caches


n
etwork effect: interaction
among

content
request/reply flows from different users:


content replacement
: requested content by one user
replaces content previously requested by others


x

Circuit
-
switching:

blocking networks

(
Erlang
, 1917, Kelly 1986)

Packet
-
switching:

queueing networks

(
Kleinrock
, 1963)

Content
-
caching:

c
ache networks

Modeling a network of caches

content


node
i
:
exogenous

(external)
arrivals for content
j
:
l
(
i,j
)


node
i
:
internal

arrivals (miss
stream) for
content
j

from
downstream neighbors
h
:

m(
j,h
)


r(
i
,j
):
aggregate rate of arrival
requests at
i

for content
j





ZDD: zero download delay
assumption

l
(
i,
j
)

i

x

w

m(
j
,w
)

m(
j
,x
)

r(
i,j
) =
l
(
i,
j
) + m
(
j
,h
)

S

a
ll downstream

neighbors,
h


y

m(
j
,i
)

j

Modeling a network of caches


SCA:
standalone cache
i

approximation algorithm: given
r(
i,j
),

compute miss rate for all content j


Independence
Reference Model (IRM) of incoming
requests:



SCA approximation algorithm for LRU: [Dan 1985]



r(i,1)

c
ache
i

Pr
(
X
t

=
f
j

| X
1
,..,
X
t
-
1
) =
Pr
(
X
t
=
f
i
)

r(
i,n
)

m
(i,1)

m(
i,n
)

But we need {
(
r
i,j
,
m
i,j
)
} for a
network

of caches

F
ixed
-
point iteration

Set r(
i,j
)=
λ
(
i,j
)

Compute miss
rate m(
i,j
)

Compute arrival
rate r(
i,j
)

Return r(
i,j
), m(
i,j
)

Using Routing Matrix


Using



SCA alg潲ithm

R数敡t until

捯nv敲g敮捥

Note: tree
-
network (feed
-
forward) require single iteration

15

Using
SCA

algorithm


Quality of
a
-
NET



a
-
NET: approximation error


line topology with
9 nodes


e
rrors decrease in
networks with
high node degree


0 1 2 3 4 5 6 7 8 9

Cache ID

1.14

1.12

1.10

1.08

1.06

1.04

1.02

1.0

0.98

Sim

/ Approx Misses

s
ources of approximation errors in
a
-
NET?


SCA algorithm inaccuracies


c
ascading errors


approximated output rates of one iteration is input to
next iteration


v
iolating IRM assumed by SCA algorithm


m
iss process for file

j

negatively correlated


a
-
NET: error factor analysis


Quality of
a
-
NET


Cascade Err
. removed


Non
-
IRM removed


Quality of
SCA

a
-
NET: error factor analysis


Factor analysis
reveals that
non
-
IRM input

to SCA
is main cause of
error

0 1 2 3 4 5 6 7 8 9

Cache ID

1.14

1.12

1.10

1.08

1.06

1.04

1.02

1.0

0.98

Sim

/ Approx Misses

“Approximate
Models for General Cache
Networks,”


Elisha J.
Rosensweig
, Jim
Kurose, Don
Towsley
,
2010 IEEE INFOCOM

Overview


architecture, system, prototype


networks

of caches


challenges


approximation algorithms


n
etwork calculus for cache networks


w
orkload: logical mobility among networks

(
σ
i
,
ρ
i
)
analyses of cache networks

(
σ
i
,
ρ
i
)

bounds
# requests for f
i

over [t
1
,t
2
]:




where
r
i
(t)

= request rate for
f
i

at time
t


Goal:
a network calculus for cache networks:

(
σ
1
in
,
ρ
1
in
)

(
σ
n
in
,
ρ
n
in
)

(
σ
1
out
,
ρ
1
out
)

(
σ
n
out
,
ρ
n
out
)

t
0

t
1

t
0

t
1

f
1

requests

f
n
requests

(
σ
i
,
ρ
i
)
cache networks: observations


n
ot all requests arriving at cache will leave (unlike queue)





t
0

t
1

t
0

t
1

f
1

requests

f
2

requests


s
tream of input requests for one file only generates no
output






“burst” of request for same file generates one output






i
nteractions among files in cache critical






different intuition (from queues) about “performance damage” of
bursts





Building block: miss set,
M
i

m
iss set for f
i
:
set of requests for
c
unique files, other than
i





x
1

requests for
f
1

M
i
(x
1
, ….,
x
n,
c
)
:
max number miss sets for file
f
i
,
given
{
x
i
in
}
arrivals, cache of size c.

p
roperties:


M
i

= min(
x
i
in
, M)


x
i
out

<
M
i
,
and this bound is achievable





c:
cache size

x
n

requests for
f
n

w

. . .

. . .

!!

From
(
σ
i
in
,
ρ
i
in
)

to

(
σ
i
out
,
ρ
i
out
):


(
σ
i
in
,
ρ
i
in
)

(
σ
j
in
,
ρ
j
in
)

(
σ
i
out
,
ρ
i
out
)

(
σ
j
out
,
ρ
j
out
)

f
j

requests

. . .

f
i

requests

If {
(
σ
i
in
,
ρ
i
in
)
}
n
i
=1

and {
(
σ

i
in
,
ρ

i
in
)

}
n
i
=1


are globally
tight

and
ρ
i
in

=
ρ

i
in

for all
i

then
ρ
i
out

=
ρ

i
out


Theorem:


i
out

independent of
{
σ
i
in
}

. . .

From
(
σ
i
in
,
ρ
i
in
)

to

(
σ
i
out
,
ρ
i
out
):


(
σ
i
in
,
ρ
i
in
)

(
σ
j
in
,
ρ
j
in
)

(
σ
i
out
,
ρ
i
out
)

(
σ
j
out
,
ρ
j
out
)

f
j

requests

. . .

f
i

requests

For a cache of size
c
:


ρ
i
out

=
min(
ρ
i
in
,
M
i
(
ρ
1
in
, … ,

ρ
n
in
,c

))

Theorem:

Can calculate
ρ
i
out

from {
ρ
i
in
}


. . .

Can compute as well


i
out


Numerical example

f
0
,f
2

4 files, uniform popularity distribution

f
1
,f
3

c
ache size = 2 at each node

h
omogeneous IRM arrivals, exponential
interarrival

times

Numerical example: bounding
results

bound

simulation

m
iss rate for f
3

Cache ID


ICN: architecture
vs

protocol/mechanism


elucidation, value
-
proposition of design principles,
service/function
modularities


m
odeling networks of caches: develop analytic
models for cache networks (ICN)


equivalent of blocking networks (circuit
-
switched), or
queueing networks (packet switched)?


e
xact in asymptotic regimes?


e
rgodicty
?


Challenges

Overview


architecture, system, prototype


networks

of caches


challenges


approximation algorithms


n
etwork calculus for cache networks


w
orkload: logical mobility among networks

Characterizing mobility among networks


Historic shift from PC’s to mobile/embedded devices


~5B cell phones
(`1B smart) vs
. ~1B Internet
-
connected PC’s


Mobile data growing
exponentially
,

surpassing wired
user
traffic by 2012 [Cisco
]


any evaluation/model (ICN or otherwise) must consider
mobility


“not your father’s mobility:” characterize mobility
among

networks


distinctly different from physical mobility, models


p
hysically mobile users may be stationary (from network
transition POV); stationary users may move among networks
(multi
-
homing, multiple devices)

Characterizing mobility among networks


Measure mobility among networks via IMAP logs


o
nline users periodically “push” (background login, check)
email, and/or intentionally read mail

-
e.g
.,
kurose@cs.umass.edu

generated 7,482 IMAP entries 4/14/13


6/4/
13


track network location from which IMAP accessed


70 users,
4/14/13


7/
5
/13, resident in 183 unique AS
numbers


Where do users (in aggregate) spend time?

Mobile: Verizon, AT&T, sprint)

Work: 5 college

Home: Comcast, Verizon, Verizon, Hughes

Misc
: 172 other networks in trace

Users spend most of their time in small number of networks

Where do users (individually) spend time?

Users individually spend most of their time in small number of networks

Multihoming

COMCAST

VERIZON

Multi
-
homing

Q: How often are users multi
-
homed?


15
-
min subinterval has IMAP access from two different AS’s

Characterizing mobility among networks


o
ther characterizations of mobility among networks


generative mobility
-
among
-
network model:
parsimonious Markov chain model for individual user
transitioning among networks



ICN: architecture
vs

protocol/mechanism


elucidation, value
-
proposition of design principles,
service/function
modularities


m
odeling networks of caches: develop analytic models for
cache networks (ICN)


equivalent of blocking networks (circuit
-
switched), or queueing
networks (packet switched)


e
xact in asymptotic regimes?


ergodicty


t
raffic models (for ICM and otherwise), with mobile users,
content


m
obility the norm


Challenges

Conclusion


architecture, system, prototype


networks

of caches


challenges


approximation algorithms


n
etwork calculus for cache networks


w
orkload: logical mobility among networks

End



?? || /* */

Interesting reading


J.
Wroclawski
, “All
h
at
n
o
answers: Some issues related to the
evaluation of
architecture,” March 2013 NSF FIA meeting, Salt Lake
City UT


D. Clark, “The Design Philosophy of the DARPA Internet Protocols,”
ACM Sigcomm 1988,

Revised with extensive commentary 2013


E.
Rosensweig
, D.
Menasche
, J. Kurose, “On the Steady
-
State of
Cache Networks,”
IEEE Infocom
2013.


E.
Rosensweig
, J. Kurose, “A Network Calculus for Cache Networks,”

IEEE Infocom Mini
-
conference 2013.


E.
Rosensweig
, J. Kurose, D. Towsley, “Approximate Models for
General Cache Networks
,” 2010 IEEE Infocom,
pp. 1100
-
1108


S
. Yang, S.
Heimlicher
, J/.Kurose, A.
Venkataramani
, “User
Transitioning Among Networks
-

a Measurement and Modeling
Study”,
submitted, 2014.

Ergodicity

of
c
ache networks


does steady state performance
depend on initial conditions
(ergodicity)?


s
hown existence of non
-
ergodic

cases (replacement
policy, topology, cache size)


d
erived sufficient conditions
for ergodicity


topology (single
-
custodian trees)


from individual ergodicity
to system ergodicity

A

B

Requests

f
or A

Requests

for B

A

B

B

A


ergodicty

(continued):


s
howed random
replacement:
ergodic


defined class of non
-
protective policies (including
LRU): all
ergodic



content

Ergodicity

of
c
ache networks