VMworld 2013
Alka Gupta, VMware
Sanjay Aiyagari, VMware
Allon Dafner, Amdocs
Iain Woolf, Alcatel-Lucent
Artur Tyloch, Nokia Solutions and Networks
Learn more about VMworld and register at http://www.vmworld.com/index.jspa?src=socmed-vmworld-slideshare
Boost Fertility New Invention Ups Success Rates.pdf
VMworld 2013: Network Function Virtualization in the Cloud: Case for Enterprises
1. Network Function Virtualization in the Cloud: Case
for Enterprises
Alka Gupta, VMware
Sanjay Aiyagari, VMware
Allon Dafner, Amdocs
Iain Woolf, Alcatel-Lucent
Artur Tyloch, Nokia Solutions and Networks
TEX5466
#TEX5466
2. 2
Disclaimer
This presentation may contain product features that are currently
under development.
This overview of new technology represents no commitment from
VMware to deliver these features in any generally available
product.
Features are subject to change, and must not be included in
contracts, purchase orders, or sales agreements of any kind.
Technical feasibility and market demand will affect final delivery.
Pricing and packaging for any new technologies or features
discussed or presented have not been determined.
3. 3
Agenda
Cloud Services in Telecom
Network Function Virtualization
• Overview
• VMware Value Proposition
vSphere Carrier-Grade Hypervisor
• Real-time Features
Partnerships and Use Cases
• Amdocs
• Alcatel-Lucent
• Nokia Solutions and Networks
4. 4
VMware: The Industry-Leading Virtualization Software Company
$4.61 billion in revenues
for 2012
More than 500,000 customers
99% of the Fortune 100
99% of the Fortune 500
99% of the Fortune 1000
100% of the Fortune Global 100
99% of the Fortune Global 500
Over 55,000 partners worldwide
~13,800 employees worldwide
5. 5
The Broadest, Most Mature Ecosystem for Virtualization/Cloud
10,000+
Service Provider Partners
2,200+
Technology Alliance Partners
55,000+ Total Partners
Distributors, ISVs, OEMs, Solution Providers,
Service Providers, System Integrators
93,000+ VMware
Certified Professionals
6. 7
Virtual Reality
9%
14%
22%
33%
42%
51%
59% 60%
80%
% workloads
on virtualized infrastructure
Source: Gartner “Magic Quadrant for x86 Server Virtualization Infrastructure” by Thomas J. Bittman, George J. Weiss, Mark A. Margevicius, Philip
Dawson, June 11, 2012
2005 2006 2007 2008 2009 2010 2011 2013
80%
70%
60%
50%
40%
30%
20%
10%
0
2015
8. 9
The importance of Cloud Services to Operators
Eagerly seeking services: A strong desire to make money out of
cloud market
Traditional sources of revenue (communications services) are
declining, especially in developed markets
According to Heavy Reading study early 2012 with 149
operator respondents:
• ~60% of operators are familiar with cloud
• ~80% of operators have a cloud evaluation team in place
• 55% already purchased cloud technology
• >40% of operators have already launched multiple cloud services
Cloud services are a big driver for SDN and NFV
9. 10
Future of Telco Business Linked to Three Emerging Technologies
Virtualization
• Abstracting networking software (functions) from dedicated hardware and
running in virtual machines on COTS platform
• NFV initiative specifically is about consolidating many network equipment
types onto standard high volume servers, switches and storage in a cloud (ie
elastic, resource pooled, orchestrated ) environment
Software-Defined Networking
• Creates network abstractions to enable faster innovation
• Using virtualization techniques, e.g. network overlays to create programmable
data center networks
• Using protocols/technologies such as OSS/orchestration and management
systems to create programmable WAN
Cloud
• Layer of automation/orchestration on top of virtualization that provides NIST
cloud characteristics
• Distinction emerging between “resource cloud” and “application
management cloud”
12. 13
ENTERPRISE
SERVICES
Evolved Packet Core
(PGW/SGW/MME/GGSN)
M2M SERVICES
VOLTE SERVICES
(IMS,TAS,RMS)
SERVICE
MONITORING
Network Infrastructure
(Access, Transport, Routing/Switching)
SMART NETWORK
POLICY CONTROL`
Common HW Infrastructure distributed across
data centers, where NW nodes become
Applications
13. 14
• Virtualization of Mobile Core Network and IMS
• Virtualization of CDNs (vCDN)
• Virtualized Network Functions and Cloud Computing Infrastructure
• Virtualization of Mobile Base Station
• Virtualization of Home Environment
• Virtualization of the Enterprise (vEnterprise)
• Virtualization of service chains (Firewall, NAT, security, load balancer,
DPI, 3GPP functions, various forms of gateways, etc)
• Cloud service exposure to a 3rd party
NFV Use Cases (Source ETSI)
19. 22
Real-time ! = Average Latency
• Real-time applications need determinism
• Primary criteria: Maximum Latency (i.e. Low Jitter)
• Average Latency also important, lower is better
• 99%ile Latency: track trend in reducing outliers contributing to
maximum latency
• RT features available in ESXi 5.5
22
20. 23
Average Latencies in ESXi 5.0
Benchmark: Ping (ICMP) Kernel space <-> Kernel space
VM I/O Path: vmxnet3 vNIC/vSwitch -> Intel 82599 10GbE pNIC
Latencies: Average HRT (Half Round Trip) in microseconds
23
10 us 18 us 13 us
Ping Ping Ping
Physical to Physical VM to Physical VM to VM (same host)
Virtual Networking
21. 24
Real-time Virtualization Performance Results
24
ISR
Latency
Interrupt
period
ISRISR
Jitter
• cyclictest –p 99 –a 1 –m –n –D 10m –q
ESXi 5.0 ESXi 5.1 ESXi RT Prototype
Min 5 μs
Max 1676 μs
Avg 13 μs
Min 4 μs
Max 1536 μs
Avg 7 μs
Min 2 μs
Max 23 μs
Avg 4 μs
99%-ile 56 μs 99%-ile 11μs
99.99%-ile 118 μs 99.99%-ile 17 μs
26. 29
We Support Billions of Telecom Customer Experiences a Day
250+ customers 2500+ projects
Caribbean and
Latin America
Asia PacificEurope, Middle East
and Africa
North America
29. 32
Amdocs Online Charging System
OCS is part network, part IT, part business
These are not packets, this is money
It is challenging:
Highly sensitive to latency
Network grade high availability
Highly tuned and hardware optimized
A lost event, is a lost dollar
30. 33
Virtualizing Amdocs Online Charging System
SUCCESS
• Virtualized OCS results are on par
with the physical deployment
A joint VMware and Amdocs certification was held
• Physical environment compared to virtualized
• Production like, high loads
We had some concerns
• Virtualization may add latency overheads
• Separates OCS from physical hardware
• Real time aspects of continuous processing
!
31. 34
VMware Certification Results
Comparing virtualized OCS and physical
is comparing apples to apples
-10%
0
1
2
3
4
5
6
7
8
90%
95%
99%
99.5%
Latency (ms)
Baseline (physical) VM
Sub Second
Failover
Productivity
32. 35
Virtualized OCS LIVE in Production
A virtualized OCS is running in production in a large Tier-1 North
American carrier
36. 39
From Controller to Radio Network Platform
Virtualization
One application
Dedicated hardware
Multi - Tenancy
• Cloud in a box
Hardware independent scalable platform
supporting multiple technologies
• Flexibility
From small single technology to large
multi-technology controller
• Cloud on a disk
Software only solution can be
deployed in existing data center
• Seamless capacity
and scalability
Cloud architecture
SW
HW
RNC
RNC RNC
RNC
37. 40
Wireless Cloud Element Feature Roadmap
Leverage benefits for Network Function
Multi-Tenancy – GSM, W-CDMA,
and LTE applications share
physical resources
Horizontal Scalability – number
of CPUs controls capacity
(flexible configurations)
Elastic Capacity – the capacity of the
system can change quickly according
to need
Reduced Power Usage – as capacity
breathes throughout the day hardware
can power up / down
Geographic Redundancy – a fault in
one location does not disable the system
Zero Downtime Maintenance – VMs
move to other hardware for maintenance
Sub-Urban
Dense Urban
Rural
As traffic shifts during the day ...
... server
load is
uniform
RNCRNC
38. 41
Transformation from 9370 to WCE RNC
The WCE virtual RNC needs to scale
horizontally to much larger capacity
than possible with the 9370
Two problems must be addressed:
• N2 Communications
• Scope of Failure
Reorganization of existing 9370
code allowed for the creation of
a component responsible for cells
and another responsible for user
equipment (UEs)
As UEs (and UMUs) are
independent of each allowing
for horizontal scalability
The scope of failure is greater
reduced as UMUs don’t
require sparing
WCE vRNC Virtual Machines
Current 9370 RNC Application
RAB
U-Plane
GTP/RLC
MAC
TMU
U-Plane
RPM-Cell
C-Plane
NoB CP
C-Plane
UE CP
CMU UMU
U-Plane
GTP/RLC
MAC
C-Plane
NoB CP
U-Plane
RPM-Cell
C-Plane
UE CP
VxWorks on Linux
VxWorks
CMU
UEU
UEU
UEU
UEU
UEU
UEU
UEU
UEU
UEU
UEU
UMU
UEU
UEU
UEU
UEU
UEU
UEU
UEU
UEU
UEU
UEU
UMU
CMU
CMU
CMU
Cells
(spared)
Mobiles
Scale to
10000s of Cells
Scale to
100000s of Mobiles
39. 42
Wireless Cloud Element – 3G RNC Architecture
RNC Components:
Spared Roles
(SWACT in a few seconds):
• CMU (N+M) – Cell Management
• 3gOAM (1+1) – OAM termination
• PC (N+1) – Public / Private
network gateway
Unspared Role
(Return to capacity ~ 60s):
• UMU - UE Management (UE Call)
& RAB UE Packet processing
Platform Components:
• vCenter Server – VM Management
(VMware)
• LRC Mgr – Tenant Management
(unspared)
• Disk Access – NAS Front End to SAN
Each RNC component represents
a single virtual machine
UMU
U-Plane
GTP/RLC
MAC
NI
SCCP
C-Plane
UE CP
CMU
U-Plane
RPM-Cell
NI
M3UA
SCTP
C-Plane
NoB CP
Iub/Iur/IuCS/IuPS
Signaling
PC
PC
NAT
Common
Channel
Traffic
Iub / NBAP
Dedicated
Channel
Traffic
vCenter Server
LRC Mgr
3gOAM
PC
Control
OMU Netconf
Disk
Access
40. 43
RNC Real-time Requirements
A controller’s user-plane
components typically need
very precise timing, e.g.
The RNC must send downlink
packets such that they arrive
at the NodeB within the Time
Of Arrival Window
When the user goes into
Diversity mode the RNC’s
transmit window must
compensate for different
transmission times
Six NodeB may be involved in
a single user’s connection
which currently reduces the
RNC transmit window to 5ms
147 148 149 150 151 152
CFN
CFN
Early OK Late Too Late
TprocTOAWS
(50 ms)
Node B
CFN
Node B
CFN
Node B
SRNC
41. 44
Non-deterministic Timing
Deterministic timing not
possible within a virtualized /
cloud environment
Virtualized Network
Functions must compensate
for timing uncertainty
Data for this chart was
extracted from the virtualized
RNC while running an hour
long traffic test
Note: this system was
running with ESXi 5.1;
significant improvements are
expected from ESXi 5.5
42. 45
Interval Scheduling
Interval scheduling allows
for variable amounts of time
between events
At each event an accurate time
reference is read and the interval
between events is calculated
Based on the elapsed time and the
priority of work that needs to be
done a work schedule is created
that ensures isochronous work is
completed first at the potential
expense of background work
43. 46
Shadow Network Functions
WCE provides the capability to
create a “shadow” of a network
element within the same physical
hardware as the service providing
network element
This shadow can be used for:
• Fast and Graceful Reset
• Software Upgrade
• Software Configuration Change
• Virtual or Physical Machine Change
• Geo-Redundancy
Roll-back to the previous version
is possible after an upgrade
RNC
iBTS
RNC
Core Network
RNC
46. 49
Towards an Improved Telco Cloud
Nokia Solutions and Networks
Artur Tyloch
VMWorld 2013, San Francisco
47. 50
17.950,500no.1 59
120 20112007
Global company with a rich heritage
Ranking, macro
base station vendor
competitive
assessment
LTE supplier
to launched
operator
networks
Employees
globally
Our net sales
in 2012
bn
$
120+
Started Telecoms
experience
in years
Countries
we operate in
We acquired
Motorola’s wireless
network
Focused on Mobile Broadband
49. 52
NSN’s relationship with VMware
• VMware is NSN’s
approved software
supplier since 2011
• vSphere, vCenter..
1
• Licenses and support for
demos and testing
purposes
• Lab installations in
several NSN locations
2
Parties are cooperating on
proof of concept projects
for migrating mobile core
applications to cloud,
involving several major
wireless carriers
3
NSN and VMware
have an ongoing
business
relationship
Cooperation in
proof of concept
projects
‘Create innovative telco grade cloud solutions for wireless carriers’
NSN is VMware’s
Technology Alliance
Program (TAP)
partner
50. 53
The VoLTE race has began already
• 2 times more voice calls per MHz
• 3 times more data throughput per MHz
• A multimedia environment enriches voice and
equips for fight against OTTs
• Operators plan to combine VoLTE with RCS to
offer multimedia, video telephony and video chat
0
10
20
30
40
50
GSMAMR HSPACS
12.2kbps
VoLTE
12.2kbps
Calls per MHz
• Short call setup time
• HD voice service experience (IR.92)
• Guaranteed bit rate
• VoLTE terminal talk time expected to
outperform OTT VoIP ~ by factor 1.5
* IR.92: GSMA IMS Profile for Voice and SMS
0
2
4
6
CS call VoLTE
Setuptimeinsec
Call setup time
improvement
Increase
Spectral
Efficiency
Superior
Voice Service
Quality
Richer
End User
Experience
51. 54
Network Architecture 2020
Telco Networks will undergo a fundamental transformation
Content and application
aware network ensuring
best customer
experience
• Central data centers
for efficient network control
and XaaS offerings
• Common Network & IT data
Local data centers
hosting multiple
intelligent network
applications
Heterogeneous access
combining multiple
technologies and cell
sizes
Network Architecture 2013
Flexibility & efficiency through
virtualization and automation
52. 55
Why core applications in a Telco Cloud?
High Total Cost of Ownership
MSS
MGCF
CSCF
IMS-MGW
HSS
GGSN
PCRF
TAS
CS-MGW
HLR
SGSN
Browsing GW
I-BGF
Serving GW
MME
PDN GW
High complexity with
dedicated hardware/
software for each
function
Proprietary hardware
limits gains from
industry-wide
innovation
Cannot meet
new capacity demands
and new service
launches easily
Target: Core applications in the IT datacenter
• Better Business Agility (- create new services much cheaper and faster -)
• Operational Efficiencies (- remove operational silos -)
• Capex Efficiencies (- leverage latest IT hardware and cloud technology -)
Cloud
53. 56
NSN and VMware initiative towards improved Telco Cloud
Live experiences from a lab verification
Lab verification
Discover and develop telecom requirements
in hypervisor and cloud orchestration layers
• High availability and resiliency (recovery
groups, affinity rules)
• Integrated operations (FM, PM)
• Latency optimization
• Optimization of vMotion to telecom
applications
Major Telco
Cloud
Functions
verified
VoLTE Test
Cases
Verified
Flexible
Setup
Enabling
Efficiency
Automation,
elastic
scaling,
recovery
Cloud Domain Management
NetAct
Cloud stack mediation
Operationon
Demand
Open TASIMS
iNUM
API
VMware ESXi hypervisor
HWResource Pool
VMware
Cloud
Management
(vCloudDirector,
vCenter Server)
OAM
i/f
Voice over LTE services in Telco Cloud
58. Network Function Virtualization in the Public Cloud:
Case for Enterprises
Alka Gupta, Vmware
Sanjay Aiyagari, Vmware
Allon Dafner, Amdocs
Iain Woolf, Alcatel-Lucent
Artur Tyloch, Nokia Solutions and Networks
TEX5466
#TEX5466