Transcript QoS Support in 802.11 Wireless LANs

Optical testbed
Malathi Veeraraghavan
Univ. of Virginia
[email protected]
• Talk outline
• Item I in Yr. 1 Plan
• Existing optical testbeds – centralized control
• Opportunity for creating scalable testbeds – distributed
control
• Why is scalability important?
• Plans
• Equipment to purchase
• What we plan to implement
• Wide-area circuits
1
Item I in Year 1 Plan
• Set up a Gigabit Ethernet/Ethernet
over SONET CHEETAH testbed,
develop software to demonstrate
dynamic circuit setup and release.
– What equipment should we purchase to
first create this testbed?
– What software should we develop?
CHEETAH: Circuit-switched High-speed End-to-End ArcHitecture
7/17/2015
2
Any existing optical testbeds?
• Research optical testbeds
–
–
–
–
7/17/2015
Canarie - Canada
Starlight – Illinois
SURFnet – Netherlands
UKLight – UK
3
Canarie optical testbed
7/17/2015
4
Canarie’s optical testbed
Bill St. Arnaud
• What are the network switches?
– Answer: Cisco 15454 MSPPs
• Goal: popularize user-controlled lightpaths
(UCLP)
– Lightpath: GbE signal - Ethernet-over-SONET (EoS) - GbE signal
• Experiments supported:
–
–
–
–
7/17/2015
CRC/U. of Ottawa
U. of Quebec and Montreal (UQAM)
Carleton Univ.
U. of Waterloo
5
Multi-Service Provisioning Platform
(MSPP), e.g. Cisco’s 15454
PC
WAN access
10/100M
Ethernet
Control
1Gbps
Ethernet
Crossconnect OC12/OC48/OC192
(VT1.5 or STS1)
SONET card
• Functionality
– Cross-connect Ethernet signal to equivalent SONET signal
– Deployed in enterprises today: “multi-service” - telephony and data
7/17/2015
6
Canarie’s optical testbed
• What are the network switches?
– Answer: Cisco 15454 MSPPs
• Goal: popularize user-controlled lightpaths
(UCLP)
– Lightpath: GbE signal - Ethernet-over-SONET (EoS) - GbE signal
• Experiments supported:
–
–
–
–
7/17/2015
CRC/U. of Ottawa
U. of Quebec and Montreal (UQAM)
Carleton Univ.
U. of Waterloo
7
Research efforts: provisioning
• Most of the Canarie-supported research efforts
are focussed on provisioning
• Difficult task: many steps
– Create external databases to manage routing
information: to reach a destination host, what is the set
of switches to traverse?
– Create external databases to manage inventory
information
– Inter-domain issues (multiple carriers) – policy manager
– Then set up circuit with TL1 commands
• Lately, efforts at integrating provisioning efforts
with GridFTP
7/17/2015
8
UQAM Architecture
Figure 1: UPLA layer model
Figure 2:UPLA installation in multiple domains environment
Legend
Lightpath request
BGP router
Inter-domains
layer
Custmer
ServiceAgent
OXC
Inter-domain layer
Service location search
WSIL
Intra-domain layer
Intra-domain layer
Intra-domain layer
Device layer
Device layer
Device layer
IntraASRegistry
Domain 1
Domain 2
Domain 3
IntraAS Service
Intra-domain
layer
policies
Policy Manager
•
–
Policy
Directory
LP Service
Inter-domain layer
LPO Registry
devices layer
LPO I/F
•
•
7/17/2015
Omar Cherkaoui
Intra-domain layer
–
–
TL1
End-to-end lightpath manipulation : orchestrate the
end-to-end lightpath searching and concatenating
them.
Single-domain lightpath creation and setting up.
Respecting the domains management in allocating
lightpath.
Device layer
–
–
Network equipments
Equipment control module
9
WSDL
client
interface
UPLA Signaling process
WSDL
LP InterAS
Controler
LDAP (
Reference to
AS's WSDL )
IntraASRegistry
VPN InterAS
Controler
Service Agent
Telnet,
SNMP
GRID
APPLICATION
Border Router
Border Router
ONS 15454
WSDL
WSDL
ONS 15454
LPO I/F
LPO I/F
LPO Factory
LPO Factory
AS 2
AS 1
IntraASServer
IntraASServer
Admin
Policy
Manager
VPN Service
VPN Service
LDAP
PET
PET
GUI
Admin
Policy
Manager
LPService
Policy
Management
GUI
LPService
Policy
Management
LDAP
7/17/2015
10
Omar Cherkaoui
UPLA Signaling time
Average exec.
time (ms)
Operations
Minimal exec.
time (ms)
Maximal exec.
time (ms)
Searching & temporary
reserving in each domain (1)
3358
1623
7086
Reservation confirmation in
each domain (2)
2475
777
5884
2 phases reservations time
per domain (1+2)
5833
2400
• Signaling time = Number of domain x 2.400 (second)
• The causes of moderate signaling time:
12970
– Webservices signaling
• Communication is over Http
• XML processing when parse message
7/17/2015
– Admission control processing and domain management verification
– Resource seeking processing
Omar Cherkaoui
11
CRC/U. of Ottawa: UCLP E2E Example
Vancouver
Ottawa
6
1
4
2
6
6
SCS
SCS
SCS
JavaSpace
JavaSpace
JavaSpace
LPOS
Jini SAP
5
3
Grid SAP
7/17/2015
Michel Savoie
12
CRC/U. of Ottawa: Sequence to create
e2e lighpath using webservices and Jini
1.
2.
3.
4.
User A accesses the Grid Service Access Point (GSAP) and is authenticated by a
user name and password. The GSAP creates an instance of a User Functions
(UF) module that uses the functions provided by an instance of the Jini SAP
(JSAP) which was obtained earlier from the Jini Lookup Service in Federation A.
User A sends a request to set up an E2E connection between User A in
Federation A and User C in Federation C. The User Functions (UF) module will
call the ConnectionRequest method of the JSAP(A) and pass the connection
information that was entered by the user.
The ConnectionRequest method of the JSAP(A) will then find the route and the
LPOs required to get from Federation A to Federation C
–
Once finished, the JSAP(A) will have all the LPOs required to make the
connection
The JSAP(A) will then download the Lightpath Object Service (LPOS) from the
JLS(A) in Federation A and pass the information about LPOs and the proxies to
the JavaSpaces they came from to the makeE2EConnection method
7/17/2015
Michel Savoie
13
Sequence (cont’d)
5.
6.
7.
8.
The LPOS(A) will then lookup the proxy to the class implementing the
Connection Services (CS) interface from the JLS(A)
–
The results of this lookup will be the proxy to the Switch
Communication Service (SCS) in Federation A
The LPOS(A) will then lookup the proxy to the Jini Lookup Service (JLS) in
Federation B from the JLS in Federation A
–
The result will be the proxy to JLS(B)
Using the JLS in Federation B, the LPOS(A) will use step 5 to get the SCS
for Federation B
The LPOS(A) will then use the SCS(B) which will then talk TL1 to the
physical switch to make the cross connect between LPO1 and LPO2 at the
Switch in Federation B
7/17/2015
Michel Savoie
14
Sequence (cont’d)
9.
Once all the cross connects are established, a new LPO
is created (LPO3 (A-C)) to reflect the concatenation
between LPO1 and LPO2 which is then put into the
JavaSpace at Federation A
10. The LPOS(A) will then call the SCS(A) to make the
cross connect between User A and LPO3 at the Switch
in Federation A
11. The LPOS(A) will then use step 6 to find the proxy to
the JLS in Federation C and steps 5 and 10 to make
the cross connect between LPO3 and User C in
Federation C
7/17/2015
Michel Savoie
15
Starlight
• Starlight and OMNInet: optical vs.
photonic
• Photonic Interdomain Negotiator (PIN)
signaling – their own design
• Domain routing database stored outside
switches
• Use TL1 commands to program switches –
Nortel, Glimmerglass, Calient and IMMI
• Centralized approach
7/17/2015
16
Problems recognized
by these efforts
• Scalability
• Using network management systems outside the
network elements to manage
– routing data
– inventory data
takes significant effort – Operations costs
• Other problems
– Interoperability of different vendors’ equipment
• Hence a startup such as Elematics
– Inter-carrier issues
• My pet peeve:
– Call setup delays of in the order of seconds too high
7/17/2015
17
Scalability “charges”
• Levied against Optical/TDM networks:
– “Widely recognized that the current GLIF optical/TDM
networking model does not scale beyond a limited number
of sites” Internet 2 talk dated 10/15/2003
– “While such circuit-switched networks may not
necessarily be suitable for deployment of the scale of
the Internet, they are still viable candidates for
specialized deployments for connecting a small number of
DOE large-scale science nodes” Report of DOE Workshop on
Ultra High-Speed Transport Protocols and Dynamic Provisioning
for Large-Scale Science Applications, April 10-11, 2003,
Argonne, IL, dated Oct. 27, 2003.
GLIF: Global Lambda Integration Facility
7/17/2015
18
Problems recognized
by these efforts
• Scalability
• Using network management systems outside the
network elements to manage
– routing data
– inventory data
takes significant effort – Operations costs
• Other problems
– Interoperability of different vendors’ equipment
• Hence a startup such as Elematics
– Inter-carrier issues
• My pet peeve:
– Call setup delays of in the order of seconds too high
7/17/2015
19
Inventory problem
• TL1 command to set up a crossconnect through a 15454
• Command:
– ENT-CRS <STS_PATH>:[<TID>]:<FROM>,<TO>:<CTAG>::[<CCT>][::];
• Example:
– ENT-CRS-STS1:BODEGA:STS-5-1,STS-12-5:116::2WAY;
• TID: unique name for the system
• From and To: Access Identifiers – to identify timeslots on
interfaces
– STS-1 on the card in Slot 5
– STS-5 on the card in Slot 12
• CTAG: unique identifier used to match response with request
• CCT: Crossconnection type: e.g., 1WAY or 2WAY
7/17/2015
20
Signaling approach to
connection setup: distributed
• Call setup request carries destination IP
address D + bandwidth B + incoming timeslot/l
– Lookup routing data table (same function as in an IP
router)
• find outgoing interface O to reach destination D
– Resource allocation
• Allocate bandwidth B on interface O
• Select outgoing timeslot/l
– Program switch fabric
• Map incoming timeslot/l to outgoing timeslot/l
• Send call setup request to neighbor connected
by interface O
7/17/2015
21
Industry answer to
support distributed signaling approach
• IETF GMPLS
– Routing – OSPF-TE
• Routing built into network switches
– Signaling – RSVP-TE
– Link Management Protocol (LMP)
• Inventory data stored in network switches
• Auto-discovery of neighbors
• OIF’s UNI-C, UNI-N, NNI
– Addresses carriers’ inter-domain issues
7/17/2015
22
Actually implemented!
• Not just idle specifications
• Implemented by many switch vendors
• And interoperability-tested by an
OIF-sponsored effort led by Univ. of
New Hampshire
– Demo’ed at OFC2003
7/17/2015
23
Interoperability
Participating Companies
• Alcatel (UNI-C, UNI-N, E-NNI, NMS/EMS)
Edge device - router
• Avici (UNI-C)
• Ciena (UNI-N, E-NNI)
• Data Connection (UNI-C, UNI-N, E-NNI, NMS/EMS)
• Elematics (UNI-N, E-NNI)
• Mahi Networks (UNI-N, E-NNI, NMS/EMS)
• NEC (UNI-C, UNI-N, E-NNI)
• Motorola/Netplane (UNI-C, UNI-N, E-NNI, NMS/EMS)
• Nortel (UNI-N, E-NNI)
• Sycamore (UNI-N, E-NNI, NMS/EMS)
• Tellabs (UNI-N, E-NNI, NMS/EMS)
• 7/17/2015
Tellium (UNI-N, E-NNI, NMS/EMS)
24
UNI/NNI Signaling Display
7/17/2015
25
From keynote at Opticomm,
Dallas, Oct. 03
• Rajiv Ramaswami, CTO , Optical Systems Group,
Cisco, Keynote
• UCP (Unified Control Plane) Benefits
– Superfast Provisioning
• Enables E2E circuit setup without SP intervention while
reducing provisioning times
• Enables future bandwidth on demand applications as policy &
billing standards mature
– Enhanced Scalability
• Network level: Support for thousands of nodes, links and
circuits per inter-connected network
• Lightweight EMS: Move from EMS based (centralized)
provisioning to node level (distributed) provisioning using
signaling
7/17/2015
26
UCP Benefits contd.
• Interoperable vendor implementations
– Reduces EMS/NMS integration / interoperability issues
– UCP/GMPLS – A Driver for Evolution
• Build Network as a Database
– Simplify provisioning by driving intelligence (topology,
circuit inventory and link characteristics) into the NEs
with updates to EMS (CTC/CTM)
– Enable migration from an NMS based network database
to NEs based network database, retrievable on demand
by NMS
• Deliver Advanced Benefits
– New services & features (Ethernet,OVPN & Storage)
not possible today…
– Reduce costs, increase revenues, address scale of
growing networks
7/17/2015
– Enable multi – network/vendor/SP interoperability
27
Revisit problems identified
by Canarie and other groups with
centralized solution
• Scalability
• Using network management systems outside the
network elements to manage
– routing data
– inventory data
takes significant effort – Operations costs
• Other problems
– Interoperability of different vendors’ equipment –
– Inter-carrier issues
• My pet peeve:
– Call setup delays in the order of seconds too high
7/17/2015
28
Open questions with
distributed approach
• SONET XCs typically sold in largesize configurations to carriers
– Hard to get the attention of vendors for
purchases of ones and twos
– Expensive
• What is the call setup delay?
– in the order of ms?
7/17/2015
29
Outline
• Item I in Yr. 1 Plan
• Existing optical testbeds – centralized control
• Opportunity for creating scalable testbeds –
distributed control
• Why is scalability important?
• Plans
• Equipment to purchase
• What we plan to implement
• Wide-area circuits
30
Why is scalability so
important?
• Research testbeds such as ESnet,
CA*net 4 may not see scalability as an
important goal
– Reason: if main application is to transfer
very large files and support other
eScience apps such as visualization,
deployments will be small-scale
7/17/2015
31
Why is scalability important?
• Beyond the obvious reason of growth
• Interaction between traffic load,
utilization, costs and file sizes
– The higher the traffic load, the higher
the utilization
– Higher the utilization, lower the costs
– The smaller the sizes of files
transferred on circuits, the higher the
load
7/17/2015
32
Aggregate utilization in a
circuit-switched network
(1  Pb )  
 m / m!
ua 
, where Pb  m
k
m
  / k!
k 0
: traffic load
m: number of circuits
Pb: call blocking probability
For a 1% call blocking probability Pb = 0.01

1
10
100
7/17/2015
m
4
17
117
ua
rho/m <1
24.8% rho/m ~ua
58.2%
Pb = 0 ua = rho/m
84.6%
(M/M/m/inf case) – Pb=0
33
should we give high-BW circuits
or low-BW circuits from util
point of view?
• Per-circuit BW is high, m is small
• Rimax = 10Gbps m =1
• to get high util. in this setting,
– because rimax is high, service time is
low, mu is high
– to get rho = lamba/m*mu to approach 1,
lambda has to be high. Crossover file
size should be low
7/17/2015
34
• Per-circuit BW is low, m is high
• Rimax = 100Mbps m =100
• to get high util. in this setting,
– because rimax is lower, service time is
higher, mu is smaller
– to get rho = lambda/m*mu to approach 1,
lambda has to be same. Crossover file
size has to be same
7/17/2015
35
effect of m
• this means if m is small, one can run
at high util. by increasing rho so that
rho/m is close to 1. But then Pb will
be high. My guess is Pb increases with
rho faster at low m than at high m.
• Compare with M/M/1/k where a
packet loss is comparable to call
blocking
7/17/2015
36
effect of crossover file size
• the smaller this is the more it allows
to build switches with fewer
interfaces. This is the advantage of
h/w sig. – it is not being able to
increase load. whatever the reqd. load
to achieve high util. based on the
crossover file size we can adjust the
no. of interfaces aggregating to a link
7/17/2015
37
File sizes
• Use of e2e circuits for file transfers: typically
limited to “large” files
• What is the drawback of using e2e circuits for
small files?
– e.g., in a path with a 50ms r.t. propagation delay, if we
transfer a 100KB file over a 100Mbps path, transfer time
is only 8ms. Circuit utilization is 8/(50+8) = 13.7%
• Two opposing factors
– If the crossover file size (beyond which circuit setup is
attempted) is increased
• per-circuit utilization increases
• traffic load decreases (Pareto distribution of file sizes),
which means aggregate utilization decreases
7/17/2015
38
Pareto density function
P( f ) 
 
 1
(  0, f    0)
f
 : shape parameter
 : scale parameter
7/17/2015
39
Impact of data rate increase
• As data rates increase, the crossover
file size will increase
– propagation delays stay unchanged
• Therefore to achieve high utilization
on the CS network
– decrease signaling message processing
delays
– engineer large-scale networks
7/17/2015
40
Plot of utilization u with
rc= 100Mbps, k=20
7/17/2015
Pb=0.3
Pb=0.01
41
Summary
• Background on optical testbed
creation
– Circuit-switched networks
• Networking a.k.a. resource sharing 
dynamic circuit setup/release
– Centralized management approach or
distributed signaling approach
• Deployment of testbed in wide-area
will be addressed in afternoon talk
7/17/2015
42
Why routing decision module
• Next few slides explain the need for
this module
7/17/2015
43
CHEETAH concept: Leverage
presence of Internet path
1. Use Internet path for initial short message
exchanges prior to actual data transfer
–
e.g., URL from client to server – or get file request
–
–
If call is blocked, fall back to Internet path
Engineer network for high utilization at the cost of
blocking
2. Run circuit-switched network in call blocking
mode
3. Use Internet path for reverse direction error
control and flow control messages and some
retransmissions
7/17/2015
44
Should the app. attempt a
circuit setup or not?
• Mean delay if a circuit setup is attempted
E[Tcheetah ]  (1  Pb )( E[Tsetup ]  Ttransfer )  Pb ( E[T fail ]  E[Ttcp ])
Pb: call blocking probability in the circuit-switched network
If circuit setup fails, fall back to Internet path
7/17/2015
45
Routing decision

if E[T

] attemptcircuit setup
if E[Tcheetah]  E[Ttcp ] use theT CP /IPpath
cheetah] 
E[Ttcp
 E[Tsetup]


if 
 E[Ttcp ]  Ttransfer use t heT CP /IPpat h
 1  Pb

 E[Tsetup]


if 
 E[Ttcp ]  Ttransfer at t emptcircuit set up
 1  Pb

7/17/2015
46
Numerical results
link rate = 1Gbps
7/17/2015
Tprop = 0.1ms
Tprop = 50ms
47
Crossover file sizes
When rc = 100Mbps and Tprop = 0.1ms
Measure of loading on
ckt. sw.
network
Pb = 0.01
Pb = 0.1
Pb = 0.3
TCP/IP path
rc = 1Gbps,
Tprop = 0.1ms
7/17/2015
Ploss = 0.0001
Ploss = 0.001
Ploss = 0.01
22MB
9MB
1.2MB
24MB
10MB
1.4MB
30MB
12MB
1.8MB
48
Additional background
• What is a circuit switch
• SONET hierarchy
• GFP and VC
7/17/2015
49
Circuit switching
Unfolded view of switch
TDM or WDM
demultiplexers
TDM or WDM
multiplexers
Output line cards
Input line cards
(buffers)
(processors to examine
headers + buffers)
Circuit switch: “Position” based switching
Space
(Interface)
7/17/2015
Time
Frequency
(Wavelength)
Packet switch: “Header”
based switching
50
SONET hierarchy
•
•
•
•
•
OC1: ~51Mbps
OC3: ~155Mbps
OC12: ~622Mbps
OC48: ~2.5Gbps
OC192: ~10Gbps
7/17/2015
51
Multi-Service Provisioning
Platform (MSPP)
WAN access
PC
10/100M
1Gbps
Ethernet
Ethernet
• MSPP as a circuit switch:
Crossconnect
(VT1.5 or STS1)
SONET card
– “Space” on one side (Ethernet)
– “Space” and “Time” on the WAN side
• PLUS: Physical layer modification from Ethernet PHY to SONET PHY
• Virtual
7/17/2015 concatenation: 100Mbps Ethernet mapped to 2OC1 instead
52 of
to an OC3