ITU Workshop on “Cloud Computing”
(Tunis, Tunisia, 18-19 June 2012)
Emerging Architecture for Cloud
Computing
Monique Jeanne Morrow
Distinguished Engineer and CTO
Asia-Pac
mmorrow@cisco.com
Tunis, Tunisia, 18-19 June 2012
My Co-Authors
Masum Z. Hasan,Sree Gudreddi, Edgar
Magana and Lewis Tucker
Cisco CTO Office
Remember: Network Factored Cloud?
App Tiers in a Typical DC
Dept/Customer 1
Branch
Branch
Internet
MAN/WAN/SP Net
DC
Dept/Customer 2
Web Tier
DMZ
App Tier
Core
Distribution
DB Tier
Aggregation
Storage Tier
Tiered Network:
§
§
§
§
§
§
Storage
SAN/NAS
Access: App tiers reside here
Aggregation, distribution, core
(part of app tiers may reside here)
DMZ
Campus core/MAN/WAN edges
Dept 2
Dept 1
App 6
App 1
Access
DB 2
SAN
DB 1
Outsource
to Cloud
Generic Data Center and Network Drawing
Data Center A ApplicaBon ApplicaBon ApplicaBon ApplicaBon ApplicaBon ApplicaBon ApplicaBon ApplicaBon Virtual Machine (VM) Virtual Machine (VM) Virtual Machine (VM) Virtual Machine (VM) Virtual Machine (VM) Virtual Machine (VM) Virtual Machine (VM) Virtual Machine (VM) ApplicaBon Virtual Server Hypervisor VM/Server Control Server Server Rack Virtual Subnet/
VLAN 2 Virtual Subnet/
VLAN 1 L2 Aggregator Func%onal Servers Func%onal Servers Provisioning, Billing, Security, Load Balancing, Monitoring, AudiBng, Logging, and ETC. Data Center Core/ Gateway Customer Edge WAN Edge WAN Network
Workloads categorisation (and generalisations)
Type of Workload
Example
Implication
Stateful
Shopping cart, collaboration services
Synchronisation
Stateless
HTTP (without cookies)
No synchronisation required
Live
Mission-critical ERP, hosted UC&C services
Performance, distance,
application tolerances
Offline
Document management, archives
-
Bursty
Voting system, VoD, ticket booking
Capacity management
Non-bursty
Data analysis
Time dependent
(predictability)
Desktop as a Service (during business hours vs.
non-business hours)
‘Follow the moon’ migration
Shared
Utility hosting (SaaS)
-
Grid
Grand challenge problems – derivatives analysis
at NAB, SETI
Higher utilisation possible –
requires HPC environments
Transactional
Billing system
Local storage & compute
Batch
Payroll
Storage & compute can be
remote
<TELSTRA DOCUMENT ID>
Putting it All Together: Seamless Cloud
Public / Community
Clouds
Tenant Private Cloud / Intranet
One or more DC
Enterprise Users,
Departments
Cloud
Service
Consumer
Intrane
t
Enterprise IT
Cloud Service
Provider
vNIC2
OS1
DB1
VM13
Seamless Execution and
management as if all
resources are on Intranet
Seamless Cloud:
covers Private,
Hybrid, Multi-SP
Inter-Cloud
Internet /
SP Private
MAN/WAN
Cloud
Service
Provider #K
One or more DC
Seamless Extension
SCL Service
Internet
/
SP
Private
MAN/
WAN
(IP/
MPLS/
Optical)
One or more DC
Cloud Service
Provider #1
vNIC5
OS2
App1
VM55
Internet /
SP Private
MAN/WAN
Cloud
Service
Provider #K
One or more DC
Use case: Distributed Applications on Seamless
[Hybrid/Inter] Cloud
Example: Hadoop MapReduce
Enterprise (onpremises)
Load Data
in DFS nodes
Launch
Map/
Reduce
Data Block Info
(RPC)
Cloud Network
Cloud Provider s
DC
Get processed
Data
(HTTP)
Get processed
Data
(HTTP)
Cloud Management Framework Architecture
§  Cloud Service and Resource
Management
§  Cloud Abstraction
§  Cloud service interfaces to Cloud
Service Consumers
à  Software :Examples: vCloud
Director, Amazon AWS,
OpenStack (Cloud Abstraction)
Compute, Storage, Network
abstraction & Management (config /
provisioning / monitoring),
Orchestration and Automation
à Software
Example: NMS/EMS, Orchestrator/
Management Systems, Libvirt API,
OpenStack NACI for Inter-Cloud,
DQCS
Cloud Service Consumer
(User / Admin/ Tool / Program)
CSP Cloud Management Framework (CCMF)
Cloud Services Layer
SaaS
Cloud Services
API Engine
PaaS
IaaS
(Abstract) Cloud Resource Management
Cloud
Abstraction Compute/VM
à Software and Hardware
Storage
Network Abstraction for
Cloud Interfaces (NACI)
Network
SCL
CCMF/CSP Internal or
CSPßàCSP
Cloud Resource/Service CRUD Realization Layer
Compute/VM
Embedded Management, Control in
devices
Tenant ßà CSP
Interfaces
SP Private MAN/WAN
/ Internet
Storage
Network
Internal Interfaces or Protocols CLI, XML-I, SNMP, etc.)
Physical Compute/Storage/Network Element or Infra Layer
Compute/VMM
Storage
Network
Service Class based DQCS
RFC 4954
"
Application or Service
class based
"
T2CSP: specify service
class (such as
Multimedia Streaming)
when acquiring
compute/storage
resource
CSP-I/NACI:
Realize_QoS (DSCP,
BW, …,
points_in_network)
CSP-CSP
"
"
"
Delegate T2CSP
request
ITU Y.1541
•  IPTD: one way
•  Y.1541 defines IP Delay Variation in terms of
the difference between the minimum and
maximum transmission delays during some
time interval.
•  IPTDmin = Minimum IP transmission delay
•  IPTDupper = 99.9% percentile of IP
transmission delay
•  IPDV = IPTDupper – IPTDmin
RFC 4594, Y.1540/1 and other Recommendations
------------------------------------------------------------------|Service Class |
|
Tolerance to
|
|
Name
| Traffic Characteristics
| Loss |Delay |Jitter|
|===============+==============================+======+======+======|
|
Network
|Variable size packets, mostly |
|
|
|
|
Control
|inelastic short messages, but | Low | Low | Yes |
|
| traffic can also burst (BGP) |
|
|
|
|---------------+------------------------------+------+------+------|
|
| Fixed-size small packets,
| Very | Very | Very |
| Telephony
| constant emission rate,
| Low | Low | Low |
|
| inelastic and low-rate flows |
|
|
|
|---------------+------------------------------+------+------+------|
|
Signaling
| Variable size packets, some | Low | Low | Yes |
|
| what bursty short-lived flows|
|
|
|
|---------------+------------------------------+------+------+------|
| Multimedia
| Variable size packets,
| Low | Very |
|
| Conferencing | constant transmit interval, | | Low | Low |
|
|rate adaptive, reacts to loss |Medium|
|
|
|---------------+------------------------------+------+------+------|
|
Real-Time
| RTP/UDP streams, inelastic, | Low | Very | Low |
| Interactive | mostly variable rate
|
| Low |
|
|---------------+------------------------------+------+------+------|
| Multimedia
| Variable size packets,
|Low - |Medium| Yes |
|
Streaming
| elastic with variable rate
|Medium|
|
|
|---------------+------------------------------+------+------+------|
|
Broadcast
| Constant and variable rate, | Very |Medium| Low |
|
Video
| inelastic, non-bursty flows | Low |
|
|
|---------------+------------------------------+------+------+------|
| Low-Latency | Variable rate, bursty short- | Low |Low - | Yes |
|
Data
| lived elastic flows
|
|Medium|
|
|---------------+------------------------------+------+------+------|
|
OAM
| Variable size packets,
| Low |Medium| Yes |
|
| elastic & inelastic flows
|
|
|
|
|---------------+------------------------------+------+------+------|
|High-Throughput| Variable rate, bursty long- | Low |Medium| Yes |
|
Data
|
lived elastic flows
|
|- High|
|
|---------------+------------------------------+------+------+------|
|
Standard
| A bit of everything
| Not Specified
|
|---------------+------------------------------+------+------+------|
| Low-Priority | Non-real-time and elastic
| High | High | Yes |
|
Data
|
|
|
|
|
-------------------------------------------------------------------
ITU Y.1540/1
Class 3
IPLR/IPTD/IPDV
Loss/Delay/Jitter
Ignore IPER (BER)
Class 0:
.001/100ms/50ms
Class 0
Class 1:
.001/400ms/50ms
Class 3
Class 2
.001/100/U
Class 0
Class 3:
.001/400ms/U
Class 0
Class 4:
.001/1s/U
Class 4
Class 5:
U/U/U
Other recommendations:
Class 1
Class 3
Class 4
Class 4
Class 5
Class 5
Streaming video:
Loss: 2% (2 loss every
100)
Delay: 5s
Jitter: Unspecified
Video Conferencing:
Loss: 1% (1 loss every
100)
Delay: One-way 200ms
Jitter: Average 30ms
Bandwidth: Extra 20% for
burst
Voice:
Loss: <=1%
Delay: One-way 200ms
Jitter: Average 30ms
Bandwidth per call: 21-106
kbps based on sampling
rate, codec, frame/packet
overhead
Conclusions and Recommendations
Now looking at offering Differentiated
Cloud Services
"   Inter-Cloud and so called Federated
Constructs now
"   Prototyping Service capabilities in
progress
"   Cloud Standards Activities very active
"