Networking the Clouds

thoughtlessskytopNetworking and Communications

Oct 29, 2013 (3 years and 9 months ago)

66 views

Cross Stratum Optimization (CSO):

Networking the Clouds

Young Lee

Huawei

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

2

Agenda


CSO Concept and Architecture Overview


Optical as a Service (OaaS)

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

3

Background


Most Internet Services are offered by “servers/hosts”.


Economies of scale is pushing single servers to server farms,
co
-
located CDN centers, and extremely large data centers.
Virtualization has provided a key enabler and unit of
computing, the VM.


Data Centers have emerged as the primary physical
infrastructure for Content Delivery Networks and Cloud
Computing.


Cloud Computing offers a variety of services: SaaS (Salesforce,
Google Apps), PaaS (Google App engine, Amazon CloudFront
CDN), IaaS (Amazon EC2, Microsoft Azue)

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

4


Page
4

Key Issues with Cloud Computing

Among the top ten obstacles for cloud computing according to "A view of cloud
computing," Communications of the ACM, vol. 53, p. 50

58, Apr. 2010:









We see that over half have significant involvement with the
network.


Without a close coordination with the network, cloud
computing remains as “best
-
effort” clouds at best


There is a great need for “network
-
aware” cloud and “cloud
-
aware” network architecture definition and solutions from
both top down and bottom up approaches.

1. Availability/Business Continuity

6. Scalable Storage

2. Data Lock
-
In

7. Bugs in Large
-
Scale Distributed Systems

3.

Data Confidentiality and Auditability

8.
Scaling Quickly

4
.
Data Transfer Bottlenecks

5. Performance Unpredictability

9.
Reputation Fate Sharing

10
.
Software Licensing

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

5

Networking the Clouds


Recent Amazon Cloud Failure


“Best
-
effort” Cloud is not enough


Resiliency and reliability


Network performance is key to the success of any
cloud
-
based application (Level 3)


Network performance enables application performance


Application should determine the solution


Users must have increased control and visibility into that
network

(Source: P
-
OTE2011 LR event, May 2011)


June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

6

What is Cross
-
Stratum?

Application Stratum


Distributed Resources: servers,
content, data sets, computing power,
cache/mirror, storage, VM


Uses Network Resources (LAN/WAN)


Different QoS requirements for each
application

Network Stratum


Bandwidth, Connections, Links,


Connection Processing (Creation,
Deletion, Management)


Admission Control, Resource
Reservation


Applications uses resources in IP,
MPLS, and/or OTN, DWDM/Fiber

Network Stratum

Application Stratum

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

7

Page
7

Cross
-
Stratum Optimization Objectives


Exchange either or both network
capabilities or application
demand/resource information


Exchange topology and/or traffic
-
engineering related information
between the layers
(virtualization/abstraction)


Initiate service instantiation of
application to network with profile
exchange (provisioning)


Exchange application/network
congestion/failure information
(monitoring)

Provisioning

Utilization

monitoring

Capabilities

Topology

Fault

monitoring

Application Stratum

Network Stratum

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

8

Page
8

Open CSO Interface


Application Stratum


Client/Customer identification of
some type, e.g., IP address.


Server types and identification


Application data flows and QoS
requirements; These may be
statistical in nature and vary over time


Server load and fault conditions


Network Stratum


Network topology, client & server
locations within that topology


Network capabilities and capacities
with respect to QoS, B/W, latency,
etc…


Network load and fault conditions


Application

Network

Open

Interface

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

9

Page
9

CSO Architectural Entities


Application Control Gateway (ACG)


Accesses application related data and
processes


Communicates with NCG


Provides information
abstraction/virtualization and access
limitations to outside entities



Network Control Gateway (NCG)


Accesses network related data


Communicate with ACG


Communicate with network processes
such as admission control, resource
reservation, connection processing, etc.


Provides information
abstraction/virtualization and access
limitations to outside entities


Application

Network

Application
Control Gateway

Network Control
Gateway

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

10

Application

Page
10

Information Sharing and Privacy


Information Privacy


Limit information flow; use
abstraction/virtualization, don’t
provide full access to
management or control planes.


Application CSO Processes


May need some type of network
information, e.g., some type of
topology information


Network CSO Processes


May need some type of
application information, e.g.,
client and server identification


3
rd

Party CSO


Uses both application and
network info

CLO

Processes

3
rd

Party

CLO Processes

Network

CLO

Processes

ACG

ACG

ACG

NCG

NCG

NCG

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

11

Optical as a Service (OaaS)


OaaS is based on the CSO concept applied to GMPLS
-
based MPLS
-
TP/OTN/WSON networks integrated
with multi
-
layer PCE capability.


June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

12

Motivation for OaaS


Cloud Computing and Data Center Networks are changing the landscape
of network architecture and the concept of services.


Highly concentrated facilities

require multiple lambdas to the network.


US Government recent report on the consolidation of its data centers will
change the network/cloud infrastructure.


Carrier’s new Data Center/Cloud Computing RFI changes the access network
architecture and open new optical service concept.


Focus on mission
-
critical applications beyond “best
-
effort” Cloud


Dynamic high bandwidth


Highly concentrated service access


Stringent latency/jitter


Guaranteed performance objective at all cost


High security and dedicated resources


Example: medical surgery, high
-
end financial trading, sporting events, and live
concerts, bulk data transfer, disaster recovery, etc.


Recent progresses in packet
-
optical networks such as
MPLS/GMPLS/WSON /OTN/MPLS
-
TP/PCE help enable the Optical as a
Service (OaaS) vision to reality.


June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

13

Evolving relationship between

Data Centers and Networks

AS

(L2)

Server

Racks

DWDM

Co
-
Lo

Data Center

Router

(CE)

Metro

DWDM

DWDM

DWDM

PE

P

AS

(CE)

Server

Racks

Co
-
Lo

Data Center

OTN

PE

OTP

ULH

ULH

ULH

ULH

P

PE

ULH

ULH

ULH

PE

MPLS TP

ULH

(a) Legacy

(b) Emerging

Lambda/OTU/Ethernet



Significant changes in network access, e.g., from a drop
-
off of an optical metro ring (a wavelength or two), to


an end destination in a long haul DWDM system (many wavelengths, multiple fibers).



Emerging applications require high bandwidth such as sporting events, live converts, 3D video applications,


remote medical surgery and so on.



Network capability that provides flexible, dynamic and reliable transport needs to be considered for mission
-


critical applications

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

14

OaaS Architecture

Application Plane (Resources in DC)

Control Plane

Transport Plane
(MPLS
-
TP/OTN/WSON)

ACG

SCG

ANI

TED

NRDB

MRT

MIB

SCI

CCI

CCI: Connection Control Interface

SCI: Service
-
Control Plane Interface

ANI: Application
-
Network Interface

UAI: User
-
Application Interface

SMI: Service
-
Management Plane Interface

Service Plane

User Plane

UAI

Management Plane

SMI

Config

DB

API

Application

Stratum

Network

Stratum

User Profile

User profile defines characteristics of the
users.



User Identifier


Unique User Identifier (virtual ID)


User Device Identifier


Address of the user device(s):
there might be multiple user
devices depending on the
location of the user.


User Device Codec Information

(if
applicable)


User Preference

(if any):


Server location (if known a priori)


User Capability
:


Max bandwidth user device can
handle for both uplink and
downlink

User

Application

Stratum

Profile

Application (Service) Profile


Application profile defines characteristics of the
application


Security
profile: (i) dedicated end
-
to
-
end
VPN
-
like resource allocation; (ii) dedicated
physical resource allocation


Location

profile: locations of both the
clients and the sources


QoS profile
: (i) Delay Tolerance Bound; (ii)
Jitter Tolerance Bound; (iii) Packet Delivery
Ratio Tolerance; (iv) Network Availability,
etc.


Connectivity
profile: (i) P
-
P; (ii) P
-
MP; (iii)
MP
-
MP; (iv) Any cast, etc.


Directionality

profile: (i) uni
-
directional;
(ii) bi
-
directional


Bandwidth

profile: Maximum, average,
and minimum bandwidth requirements
for the connectivity, maximum burst rate,
maximum burst duration, etc.


Duration of service

profile: service time of
the application once setup


Restoration

profile: (i) Reroute required;
(ii) do not re
-
route, etc.


Application

Stratum

Network

Stratum

Profile

Network Capability





Application

Stratum

Network

Stratum

Network

Capability

Depending on the application, its nature, and related quality
of service, the underlying network is required to have
different capabilities. For our purposes here, network
resources and capabilities can be summarized into the
following categories:


Bandwidth capabilities

---

the ability
of the network to meet bandwidth
profile requirements of the
application service.


QoS and SLA

--

the ability of the
network to deliver according to the
QoS profile requirements and the
corresponding service level
agreements (SLA).


Configurability
--

the ability to
reconfigure/re
-
optimize various
aspects of the network and the
timeliness in which changes can
occur.


Adaptability

---

the ability to adapt
changes due to changes of service
demand or application/network
congestion/failure.

Intra
-
DC Load Balancing

(Application Stratum)

End
-
User 1

GLB

1

2

3

GR

GR

1

2

P1

P2


P4

P3

PE

GLB

1

2

GR

GLB

Application

Resource

(server)

CE 4

Access

Network

Carrier

Transport

Network

DC 1

DC 2

DC 3

(3) Inter DC Communication


(exchange server


performance data)

(1)
End
-
user to DC


communication


(request/reply)

(4) DC
-
Carrier

Communication

(NS Query)

Access

Network

PE

CE 4

End
-
User 2

(2) Intra DC

Communication



Direct Access

(Corporate User)

CE 4


Network
-
aware GLB


June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

20

Network
-
aware GLB Illustration

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

21

Application Controller Gateway
Functionality

Interface with User Plane
(UAI)

User profile

Processing

(User Authentication,

Billing, User Preference

Extraction,

Server/VM Selection

Engine

(Assign Server and VM for

the application, Service

Request)

Service Plane


AC Gateway

(Application Profile Creation,

Application ID Assignment)





Application
-
Network Interface (ANI)


Application Resource

Management Engine

(Trace Application resource

such as servers, VM and connectivity

within application space, etc).


ACG

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

22

Service Controller Gateway
Functionality

Application
-
Network Interface (ANI)

Interface with Control
Plane (SCI)


Profile Mapping

(Network Location Derivation,

Parameter mapping: generic to OTN,

Connection
-
Application Mapping)




GMPLS

Signaling Processing

(Formulate GMPLS UNI Message

& Objects)



Network Resource

Estimation

(PCE/PCE+)



NR DB

TED

Path

Resv

OSPF

LSA

Interface with

Management Plane (SMI)

OAM

MIB



Network Resource

Abstraction &

Virtualization/

Correlation



GMPLS

OAM

Config

DB

SCG

Service

Authorization

(Policy, Subscription,

Admission Control)


June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

23

Scenario 1: Resource Reservation

Application
-
Network Interface (ANI)

Interface with Control Plane
(SCI)


Profile Mapping

(Network Location Derivation,

Parameter mapping: generic to OTN)




GMPLS

Signaling Processing

(Formulate GMPLS UNI Message

& Objects)



Network Resource

Estimation



NR DB

TED

Path

Resv

OSPF

LSA

Interface with

Management Plane (SMI)

OAM

MIB



Network Resource

Abstraction &

Data Correlation



GMPLS

OAM

Config

DB

Request

From ACG

Reply

To ACG

1

2

3

4

Service

Authorization

(Policy, Subscription,

Admission Control)


June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

24

Scenario 2: Resource Query

Application
-
Network Interface (ANI)

Interface with Control Plane
(SCI)

Service

Authorization

(Policy, Subscription,

Admission Control)



Profile Mapping

(Network Location Derivation,

Parameter mapping: generic to OTN)




GMPLS

Signaling Processing

(Formulate GMPLS UNI Message

& Objects)



Network Resource

Estimation



NR DB

TED

Path

Resv

OSPF

LSA

Interface with

Management Plane (SMI)

OAM

MIB



Network Resource

Abstraction

&

Data Correlation



GMPLS

OAM

Config

DB

Query

From ACG

Query Reply

To ACG

1

2

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

25

Scenario 3: Network Event Escalation

Application Plane

Control Plane

Transport Plane

ACG

SCG

ANI

TED

NR DB

MRT

MIB

SCI

(UNI)

CCI

CCI: Connection Control Interf ace

SCI: Serv ice
-
Control Plane Interf ace

ANI: Application
-
Network Interf ace

UAI: User
-
Application Interf ace

SMI: Serv ice
-
Management Plane Interf ace

Service Plane

User Plane

UAI

Management Plane

SMI

Config

DB


When a network event occurs, the network applies protection/restoration
schemes associated with the application connection.


When network level protection/restoration does not work, the network will
escalate to service plane which in turn escalates to the application for a possible
change of the resource origin.


The application may provide an alternative server location to the service plane.


The service plane interacts with control plane to find the path that can provide
application to the user.

1

2

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

26

Scenario 4: Application Event Escalation

Application Plane

Control Plane

Transport Plane

Application

Controller

Service

Controller

ASI

TED

NR DB

MRT

MIB

SCI

(UNI)

CCI

CCI: Connection Control Interf ace

SCI: Serv ice
-
Control Plane Interf ace

ASI: Application
-
Serv ice Interf ace

UAI: User
-
Application Interf ace

SMI: Serv ice
-
Management Plane Interf ace

Service Plane

User Plane

UAI

Management Plane

SMI

Config

DB


When there are application level events (server failure, etc.), the application
control gateway tries to find alternative servers in the same host location.


If alternative servers are only available in the remote locations, then the
application control gateway provide such information to the service control
gateway for possible connectivity change for the existing connection.

1

2

June 7
-
8, 2011

Cloud Computing and Cross Stratum
Optimization Workshop, Daejeon, Korea

27

Scenario 5: QoE Degradation Escalation

Application Plane

Control Plane

Transport Plane

Application

Controller

Service

Controller

ASI

TED

NR DB

MRT

MIB

SCI

(UNI)

CCI

CCI: Connection Control Interf ace

SCI: Serv ice
-
Control Plane Interf ace

ASI: Application
-
Serv ice Interf ace

UAI: User
-
Application Interf ace

SMI: Serv ice
-
Management Plane Interf ace

Service Plane

User Plane

UAI

Management Plane

SMI

Config

DB


When the user experiences degradation of QoE for the application, the user can signal this
event to the application control gateway.


If this is due to serve issues, then the application control gateway tries to find alternative
servers in the same host location and switch over to the alternative server to mitigate the
degradation.


If it is not related to server or alternative servers are only available in the remote locations,
then the application control gateway would provide such information to the service control
gateway for possible connectivity change for the existing connection.

1

2

Summary and Conclusion


Application
-
aware network capability and network
-
aware application capability are important
milestones to realize the NG cloud computing.


CSO can meet the major obstacles associated with
Cloud Computing: availability, data transfer
bottlenecks, performance unpredictability, and
scaling quickly.


CSO/OaaS would help the carriers to be the main
player in the NG cloud computing environment.