White Paper | May 2012
Cloud Media Processing
Embracing Cloud Computing for MRF Media Processing and IMS Applications
By: Adnan Saleem, Chief Architect, Radisys
Ray Adensamer, Senior Product Marketing Manager, Radisys
Cloud computing is gaining momentum as a substitute for traditional Basics of Cloud Computing pg. 2
IT infrastructure, due in large part to its on-demand model for providing Transition to the Cloud—
computing resources. In “Have it your way” fashion, companies and What Does It Mean? pg. 4
institutions can choose from different categories of service, ranging from
basic server instances to complete application delivery platforms, as
Around MPaaS pg. 5
offered by Salesforce.com. Leveraging capacity already in place, businesses
can quickly deploy new applications with less development effort, capital Requirements for Media Processing
investment and ongoing support. as a Service (MPaaS) pg. 9
MPaaS Deployment Categories
Telecom service providers, looking to derive the same benefits, are now
and Examples pg. 10
investigating cloud computing for delivering IP media services across
public and private networks. A significant hurdle is that public Internet Radisys Products for Cloud
networks were not designed to satisfy the telecom industry’s more stringent Communications pg. 13
requirements around real-time performance and reliability. In fact, the
References pg. 14
Internet operates on a best-effort delivery model, which poses issues for
time-constrained multimedia data, such as video and audio. Still, cloud
computing for IP media services using Intel® multi-core server platforms
is a viable option for telecom service providers who are aware of the
technical challenges and take proactive steps to address them.
For those developing cloud-based telecom solutions, this paper examines
some of the major considerations associated with deploying IP media
services, such as VoIP, video, fax and conference mixing.
Cloud Media Processing | Radisys White Paper 2
Basics of Cloud Computing Fully Integrated
According to a US National Institute of Standards and Eg: Conferencing Base Hardware
in Private Cloud Infrastructure
Technology definition, “Cloud computing is a model for
enabling convenient, on-demand network access to a Packaged Infrastructure
Soﬅware as a Service
shared pool of configurable computing resources that (IaaS)
can be rapidly provisioned and released with minimal Media Processing
management effort or service provider interaction.”1 as a Service
The value propositions created by readily-available Soﬅware Platform
as a Service as a Service
computing capacity and “pay-as-you-go” are the (SaaS) (PaaS)
underpinnings of an economic model made possible Eg: Multimedia
Eg: Network APIs
by service-oriented software, virtualization and grid or Conferencing
in Public Cloud
computing technologies, among others. Strong cloud
computing momentum is reflected by IDC’s 2012
forecast—80 percent of new commercial enterprise
Figure 1. Media Processing as a Service (MPaaS)
applications will be deployed on cloud platforms.2
Benefits of Cloud Computing Service Models
Cloud computing leverages Internet networking The cloud is capable of providing services on many
technology and standard servers to create financial levels, three categories of which are represented
and infrastructure benefits for service providers and by the acronyms IaaS, PaaS and SaaS. These service
their customers. Financially, service providers can categories are defined by TechTarget3 as:
achieve significant cost savings by deploying low- • Infrastructure as a Service (IaaS) involves
cost, commoditized general-purpose servers and outsourcing the equipment used to support
through high equipment utilization enabled by software operations, including storage, hardware,
consolidation. The value to cloud service customers is servers and networking components.
paying only for what they use and accessing additional
resources (i.e., overflow capacity) when needed without • Platform as a Service (PaaS) is a paradigm
making an upfront investment. In fact, the model for delivering operating systems and associated
dramatically lowers customers’ capital expenditures, services over the Internet without downloads
minimizes depreciation expenses and improves cash or installation.
flow. Since the computing platforms are already in • Software as a Service (SaaS) is a software
place, customers can deploy services more quickly distribution model in which applications are hosted
than if everything had to be done from the ground up. by a vendor or service provider and made available
to customers over a network, typically the Internet.
With respect to infrastructure, cloud computing is
built with fungible computing resources that can be In acknowledgment of media processing functions
easily re-purposed, thus extending the useful lifetime (see sidebar: What is a Real-time Service?), this
of hardware and software investments. Capacity paper introduces another type of service, referred
scales easily through the addition of new Intel®-based to as “Media Processing as a Service” or MPaaS.
servers, and the homogeneity of the equipment This service model has hardware, an application
simplifies the management of datacenters and central programming interface (API) and application elements,
office systems. Furthermore, compliance strategy so there are delivery mechanisms consistent with
is easier to develop and execute because of the IaaS, PaaS and SaaS, as shown in Figure 1.
consistency of the infrastructure.
Cloud Media Processing | Radisys White Paper 3
What is a Real-time Service?
Real-time telecom services are differentiated by the need to meet time-critical parameters, including
minimal end-to-end delay and jitter, to produce an acceptable quality of experience (QoE). For audio calls,
the rule-of-thumb is the round trip delay cannot exceed 150 millisecond (ms), otherwise users will hear
echos and get irritated. For this reason, VoIP is a classic real-time service.
Video calling and conferencing are similar to VoIP, where delay must be minimized. Video conferencing also
has “lip sync” issues caused by audio and video content being transmitted from source to destination on
often separate RTP media streams; consequently, their paths and delay through an IP network will differ.
To remedy this situation, the streams must be buffered and put back “in sync” at the receiving end by an
IP media server, thus overcoming jitter and sync problems.
In contrast, receiving an email and refreshing a browser screen are examples of a near-real-time service
since users can’t detect or won’t care if it takes 150 ms or longer. Although users might get frustrated
with slow Internet response, in most cases they will still use the service. Similarly, a 1-way video streaming
service is considered near-real-time because the end user doesn’t know when the transmission was actually
sent. Even with a bogged-down public Internet, YouTube or NetFlix streams are usually good enough for
users to consume and enjoy the service.
A real-time service is typically a true 2-way communication experience. In a real-time service, the users
can immediately tell if the service is slow or not working properly. Today, this is evident with video calling
and conferencing, where it is often “hit and miss” in terms of quality when streaming across broadband
and 3G networks.
The following lists examples of real-time telecommunication services:
• VoIP calls
• Video over IP calls
• Multi-media conferencing
• Video content streaming
• Multimedia ringback tones
• Interactive voice and video response (IVVR)
Cloud Media Processing | Radisys White Paper 4
• Media Processing as a Service (MPaaS) provides Tenets Implications
access to specialized platforms, APIs or fully- Abstract Resources Substitute references to physical servers
and hard drives with instances and volumes.
integrated applications that analyze and modify Cloud resources are fungible.4
media data streams. These applications leverage
On-Demand Provisioning Get more resources right when they’re needed.
functions such as audio and video mixing, Give back unnecessary resources.
multimedia transcoding, recording and playing Scalability in Minutes Scale out or in depending on usage needs.
a media stream, detecting the “loudest talker”
Pay per consumption Don’t pay for resources after they’ve been turned off.
or noisy connections, and applying voice or video
quality enhancements. Automation Increase automation using APIs. The cloud provides
access to scriptable infrastructure.
Types of Cloud Computing Table 1. Tenants of Cloud Computing
Clouds are classified as public, private, hybrid or
community, based on the entities providing computing
resources and networks. A public cloud uses the
mainstream public Internet, which is operated by • New on-premises infrastructure deployments
and shared across many external organizations. give way to cloud-based systems located offsite.
This model is typical for hosted conferencing service • Transcoding-intensive digital signal processors are
providers. A private cloud is operated for a single substituted with virtualized COTS infrastructure
organization and is common among enterprises for based on Intel® multi-core processor technology.
data security, corporate governance and reliability
• Large, upfront equipment expenditures are replaced
reasons. A hybrid cloud combines both public and
by recurring fee (rent) payment models.
private clouds. A community cloud is setup for a
common interest group, and the infrastructure is Business Model:
maintained by its members.
• ‘Conferencing as a business’ transitions to
‘conferencing as a service.’
Transition to the Cloud—
• Interoperability requirements expand from
What Does It Mean? dedicated equipment to all computing devices.
Cloud computing is creating a paradigm shift that
is closely aligned to guiding tenets around how Competition:
computing resources are accessed and consumed, • New entrants take advantage of lower barriers
some of which are listed in Table 1. As a result, to entry provided by cloud computing.
customers of cloud services can take advantage of
• Greater reliance on public networks levels the
new features, like dynamic resource allocation and
consumption-based pricing models, that increase their
agility and flexibility. • Improved interoperability enables new partnership
From the perspective of telecom service providers,
an industry migrating to cloud-based services is
likely to experience fundamental shifts related to
infrastructure, business models and competition.
Drawing on insights from Morgan Keegan5 at Equity
Research, some of these changes include:
Cloud Media Processing | Radisys White Paper 5
Overcoming Challenges virtual private networks (VPNs) designed to improve IP
communications quality through better performance,
Around MPaaS lower delay and less jitter than the Internet. When
The motivations for the industry to consider cloud quality of experience is a top priority, it’s imperative
computing, and more specifically media processing as that service providers test the end-to-end network
a service, are usually grounded on the financial benefits. performance of IP media services at maximum load.
That is why cloud computing is also characterized as a The same is true for those deploying IP media services
financial model first, but then also needs to be balanced on cloud infrastructure; it is important to closely look
with technical realities in delivering a real-time at the network performance, along with the costs for
telecommunications service. delivering that performance, with your service provider.
Leveraging cloud computing for IP media services Radisys Solution: Real-time performance is
requires changes to the underlying network architecture designed into Radisys media servers, enabling both
in order to adapt the new ways data and applications DSP hardware-based and software-based versions
will move between customer machines and central in virtualized environments running on Intel® multi-
offices/data centers. For example, public access core processors to satisfy comparable latency
networks need to be upgraded to deliver greater specifications. Employing patented technology,
bandwidth, higher quality of service (QoS) and improved all Radisys media servers (see sidebar: Media
reliability, among other things. Moreover, MPaaS must Servers in the Cloud) implement a real-time control
run on shared computing platforms and meet stringent layer that sits on top of the operating system. This
QoS requirements despite the reliance on public layer enhances DSP, multi-core and multi-processor
networks, which are far less predictable than private architectures by providing real-time optimizations
networks. The following sections describe many of specifically tuned for the deterministic response
the challenges facing networking and telecom solution times needed for IP media processing.
providers in the delivery of MPaaS, and how Radisys
solutions are helping to address them. The Radisys Software Media Server also utilizes
Intel® Integrated Performance Primitives (Intel® IPP)—
Real-time Network Performance an extensive library of multi-core-ready, highly
optimized software functions for multimedia and
Traditional Internet applications do not have strict
communications applications. Intel® IPP functions
real-time performance constraints. Most users wouldn’t
deliver parallel performance beyond what optimized
complain if a web page download took a half second or
compilers alone can deliver, enabling Radisys to
longer, whereas such a delay is unacceptable for audio
execute packet and jitter buffer processing within
and video communications. Based on user perception
the strict 5-millisecond packet processing increments
studies, most people consider the benchmark of 150
required by in high-performance telecommunication
milliseconds (ms) as the maximum tolerable delay for
applications using Real-time Transport Protocol
a satisfactory voice call. If the delay is greater, users
(RTP) media streams.
will notice the delay, which will negatively impact the
communications experience. In addition, the resource manager differentiates
processing tasks (‘hard-real-time’ versus ‘near-
Today, many telecom service providers choose to
real-time’) and distributes them independently.
purchase their own IP media processing equipment
Ensuring the hard-real-time processes have sufficient
and QoS-enabled IP network infrastructure in order
computing power, the resource manager reallocates
to deliver the real-time network performance needed
CPU, DSP, memory and I/O resources as needed.
to ensure an acceptable quality of experience (QoE)
This optimizes application performance based on
for subscribers. Likewise, control over equipment is
the available computing resources, thus reducing
especially critical for service providers who implement
cost for a given performance level.
Cloud Media Processing | Radisys White Paper 6
Availability and Reliability
Representing the gold standard for reliability, carrier- Media Servers in the
grade equipment is known for “five-nines” availability,
which is often bolstered by robust failover. Cloud-based
infrastructure supporting telecom services should and Software-based
integrate the hardware and software components Cloud service providers offering media
capable of maintaining equivalent availability. processing services have the option of deploying
hardware- or software-based media servers,
Radisys Solution: Radisys Software Media Server,
or a combination. For example, Radisys offers
running on the carrier-grade Red Hat Linux operating
both types of media servers designed for
system, can be deployed on a commercial-off-the-shelf
real-time media processing performance.
(COTS) Linux appliance or blade servers, as well as
AdvancedTCA-based computers delivering five-nines • MPX-12000
availability such as Radisys ATCA Compute Processing ˸ Broadband Media
Modules based on Intel® multi-core architecture. Resource Function (MRF)
Resource Allocation ˸ Voice and video over
As mentioned earlier, on-demand provisioning is
a key tenet of cloud computing, which requires an ˸ ATCA-based system can
infrastructure resource manager capable of adding integrate load balancing blades
burst capacity, when needed. Resource managers • CMS-9000
may also be called upon to route workloads to certain
˸ High density, fault-
nodes, perform load balancing and manage a cluster
tolerant and NEBS
of resources. These tasks must be carried out with the
understanding that some computing resources may
perform specific media services, like text to speech, pure ˸ Scalable up to
transcoding or multi-media processing. Some scenarios 22,800 ports
that resource allocation should comprehend include: • Radisys Software Media Servers
• Resource Management—Customers pay for ˸ Runs on COTS Linux servers
what they need per service level agreement; and based on Intel® multi-core
accordingly, the resource manager assigns servers processors, including the
and balances the load among all the resources. Intel® Xeon® processor
For example, a customer requests six servers total, E5 family
where five are dedicated to media services and one
˸ Patented design to maximize real-time
performance under high system load
• Burst Capacity—As an example, audio conferencing
˸ Utilizes Intel® Integrated Performance
has a peak time, typically during work hours. After
Primitives (Intel® IPP)—an extensive library
that time, a customer chooses to release the cloud
of multi-core-ready, highly optimized
resources to lower cost or to enable other services
software functions for multimedia
to reuse the same cloud infrastructure.
and communications applications
• Geographic Resource Management—Services
˸ Ongoing optimization for virtualized
and features are tailored to individual geographies,
making it necessary to manage dissimilar installations.
• Server Redundancy—In the event of server failure,
the cloud maintains a redundant server for failover
to minimize disruption.
Cloud Media Processing | Radisys White Paper 7
Radisys Solution: Radisys continues to improve
our unique application programming interface (API)
that allows cloud providers to monitor media server
resource utilization (e.g., how many resources are Virtualization is commonly discussed in the context of cloud
still available) in real-time, thus greatly facilitating computing. The technology has been around for many years,
load balancing. Radisys also exposes measurements most notably used in data centers where various applications are
and statistics for any running instance. As a result, consolidated onto a single server. Virtualization software, called
Radisys media servers are easy to manage on a per a hypervisor, abstracts a computer’s resources (CPU, memory,
instance basis or cluster basis in the cloud. I/O, etc.) and creates virtual machines that act like independent
computers with an operating system, memory, disk and an
Media Processing in Virtual Machines application. VMware and Linux KVM are examples of hypervisors,
When implementing software-based media servers, also called virtual machine monitors.
virtualization (see sidebar: Virtualization Technology)
Of concern to telecom application developers is the overhead
gives service providers far more flexibility to add/
introduced by hypervisors, particularly when the computer
remove services and features on the fly. This is
switches between virtual machines. This delay can diminish
because the software providing services/features run
deterministic, real-time performance and lead to unexpected
in virtual machines that can be deployed/undeployed
behavior. As a result, developers should take steps to ensure
while the server is running. This capability can be
their software can run in a virtualized environment without
used to satisfy the abstraction and scaling tenets of
perceptible performance degradation. This includes thorough
cloud computing described earlier. However, caution
end-to-end service testing—especially when the cloud media
is needed when using virtualization since it can slow
processing and associated cloud network infrastructure is under
down systems, as well as introduce unpredictable
high system load.
behavior that negatively impacts real-time
performance, especially under high load. Therefore, telecom service providers should weigh the potential
impact, both cost and performance, of a virtualized cloud
Radisys Solution: Radisys Software Media Server
infrastructure. There may be one option where a server instance
has been successfully tuned and tested on VMware,
runs un-virtualized on its own dedicated Intel® computing platform,
running on an Intel® multi-core server. Radisys
and another, where many virtual server instances run on the same
customers have also reported virtualized operation
on Linux KVM. Radisys Software Media server will
soon be properly productized and supported on these
and other virtualization products in the future. Using • Media Plane—The Secure Real-time Transport
virtualization, service providers can easily deploy Protocol (SRTP) supports encryption and message
multiple instances of the Radisys Software Media authentication. Note: SRTP is an adaptation of the
Server to increase media processing capacity. Real-time Transport Protocol (RTP).
Another usage is to enhance reliability by creating
• Control Plane—Internet Protocol Security (IPsec)
a backup copy that is a hot standby media server
and Transport Layer Security (TLS) secure
in a virtual machine.
IP communications by authenticating and
Media, Control and Access Security encrypting packets.
Storing company information in the cloud can be risky, • Operations and Management—Hypertext Transfer
therefore incorporating encryption and authentication Protocol (HTTP) is an application protocol for
technology is critical, particularly for public clouds. distributed, collaborative, hypermedia information
systems.6 Radisys is implementing the HTTP
Radisys Solution: Radisys implements standard combined with TLS, called HTTPS, in the future.
security protocols to protect IP flows.
Cloud Media Processing | Radisys White Paper 8
Service-Aware Load Balancing management task is traffic redirection, which is
and Traffic Redirection prevalent in teleconferencing. For example, a
conference call may have three participants calling
Load balancing is an essential aspect of managing the
into the cloud, and all of the messaging and signalling
workload sent to media servers because it optimizes
needs to be routed to the single node supporting
resource utilization and minimizes bottlenecks that
the call. Multi-party services and collaboration
impact performance. Another important workload
can make traffic redirection more complicated.
Within the cloud, many identical servers may be required to provide services for many thousands
(if not millions) of individual clients. Therefore, a vital component of a cloud service platform is the
module that distributes traffic evenly across a dedicated bank of servers.
Radisys offers two types of solutions for providing the load balancing function in an ATCA chassis. The
first, ‘wirespeed’ load balancing, resides on the ATCA chassis’ internal switch/hub. The second, which
provides an intelligent adaptable load balancing solution, is a software technology called ‘FlowEngine,’
designed to run on the Radisys family packet processing blades based on NetLogic and Cavium processors.
Wirespeed load balancing leverages functions inherent within the switching silicon itself to parse packet
headers for the fields that contain the identity of the particular user (client) or flow and to use the values
found to select a service blade or thread to process the packet. Since the switch is able to perform this
function on packets as they are received directly off the ingress port, there is very little additional latency.
The switch can load balance all external facing ports at the full line data rate.
A potential limitation of switch based load balancing is that the forwarding decision has to be made on a
completely stateless basis. This means the identification of the target for the packet is derived solely as a
mapping (or hash) of the bits in the relevant header fields. The target cannot be specifically assigned when
any one connection is first made. In the context of a media server platform, where conferencing is a common
requirement, it is important that packets belonging to a particular call can be sent to a specific blade. To
address the requirement, FlowEngine-based intelligent load balancing is often a more appropriate solution.
FlowEngine is a software technology that allows Radisys to create application-specific load balancing
appliances on top of an ATCA packet blade. Such intelligent load balancers are able to look-up key packet
header fields or hashes in a table to see if the packet belongs to a known connection or if it belongs to
a new one (a previously unseen user/session), in which case a new forwarding rule has to be created.
The new forwarding connection can be based upon a weighted round robin snapshot of recent server
loadings or can be specifically created as part of the process of setting up the session/call by the control
plane part of the media server platform.
Both switch and FlowEngine-based load balancing provide high availability features to respond to the
failure of a load-balanced service blade. Packets destined for the failed blade can either be redirected to
a standby blade, or if a standby is not available, they can be distributed among the server blades which
remain in service. Once the failed blade is restored, the load balancer can either return packets back to
the original target, or in the case of FlowEngine, minimize further session disruption by using the restored
blade only for new users/sessions as they appear.
Cloud Media Processing | Radisys White Paper 9
Radisys Solution: Radisys media processing applications employing media services hosted in
solutions can be delivered with other Radisys the cloud. In comparison, the protocols previously
products, including ‘wirespeed’ load balancing on discussed are better suited for non-IP-based
ATCA chassis internal switch/hub. Alternatively, a telephony.
Radisys software technology called ‘FlowEngine’
can deliver load balancing on the Radisys family Requirements for Media
of NetLogic and Cavium-based packet processing
blades (see sidebar: Load Balancing)
Processing as a Service
Developer API’s and Interfaces Cloud computing is capable of supporting media
Developers of telecom applications need processing services, due in large part to significant
straightforward mechanisms to invoke media services advances in computing and IP networking technology
that are compatible with a diverse set of application that are enabling a new class of real-time telecom
development and run-time environments. This services. The stringent performance requirements of
typically entails an API that makes it easier to MPaaS can be met, in many cases, by existing cloud
support multiple client platforms, such as Internet network elements. Special considerations, in order
browsers, smart phones, tablets, laptops, etc. of importance, are outlined in the following, where
MPaaS is compared to other applications moving
Radisys Solution: Radisys media servers support to the cloud.
standard-based APIs, including:
• Network I/O and Associated Load/Traffic
• Session Initiation Protocol (SIP) is an IETF-defined Management—High Importance
application-layer control (i.e., signaling) protocol for
creating, modifying and terminating sessions with ˸ Media processing controls and manipulates
one or more participants. These sessions include IP media streams, and like other applications
Internet telephone calls, multimedia distribution involving multiple video sources, it can consume
and multimedia conferences.7 a large amount of network bandwidth.
• VoiceXML (VXML) is the W3C’s standard XML ˸ High-quality network connections are critical
format designed for creating audio dialogs that to avoid network jitter, which can significantly
feature synthesized speech, digitized audio, impair the quality of experience (QoE).
recognition of spoken and DTMF key input, • Compute Resources—Medium Importance
recording of spoken input, telephony and mixed
˸ Media processing is computing intensive,
as are many other applications.
• Media Server Markup Language (MSML), described
• Element Management—Medium Importance
in RFC 5707,9 is designed to provide feature–
rich media processing control in an IP–based ˸ Perhaps more distributed than most applications,
communications network. MSML has been adopted media processing can be complex to manage,
by dozens of vendors and is used to control many especially when a service provider’s solution
millions of IP media server ports in numerous spans in-house service infrastructure and
service provider networks around the globe. outsourced cloud computing infrastructure.
• HTTP RESTful Interfaces, soon to be supported by • Billing—Low Importance
Radisys, are growing in popularity because they are ˸ Billing is obviously an important aspect of a cloud
a good fit for cloud and other Web-based services media processing service. While normally not a
due to their HTTP synergies and client-side scripting function or top concern for the media processing
capabilities (e.g., web browsing). For example, JSON layer, billing grows in importance when the
control functions are well-suited for mobile web application layer elements.
Cloud Media Processing | Radisys White Paper 10
• Storage—Low Importance
˸ Comparatively, media processing typically
requires minimal storage, unless the application
uses a media library, like a multimedia ringback
The cloud service providers best positioned to provide Integrated
SaaS Monitoring Collaboration Finance Apps
MPaaS have high-quality network connections and Content Communication
supporting QoS mechanisms. However, the costs Platform
to deploy high-performance network I/O can be Media
PaaS APIs &
considerable—often much higher than the compute Identity Queue Interfaces
Object Storage Runtime Database
For this reason the CFO, who likes the financial Compute Network Optimized
Block Storage Elements
benefits of cloud media processing, should work
together with the telecom engineers and service
planners to achieve a balance between economic and
quality aspects. They must identify when it makes
sense to migrate media processing from a controlled,
Figure 2. Media Processing as a Service (MPaaS) Aligning with
properly engineered, private telecom IP network, to Common Cloud Computing Service Models
a cloud environment that often has many variables
and unknowns. For example, one strategy might • Platform as a Service (PaaS)
involve a tiered approach such as:
˸ Application developers accessing media
• Economical Telecom Services, which leverage cloud processing services using API’s and standard
media processing and the associated cost benefits, interfaces to enable mobile and wireline
are suitable for lower-end, price-sensitive service communications applications that leverage the
offerings. cloud media platform and associated resources.
• Premium Telecom Offerings, which use existing, • Software as a Service (SaaS)
dedicated data center and IP VPN infrastructure,
continue to deliver the highest quality of experience. ˸ Cloud hosted, fully-integrated communications
services and applications, such as multimedia
MPaaS Deployment conferencing or communications portals, with all
the necessary management and billing services
Categories and Examples included as part of the service.
When MPaaS (Media Processing as a Service) was first
mentioned in this paper, it was suggested this service Media-Optimized Elements
type fits into the most common cloud computing in the Cloud—IaaS Model
service models. This point is illustrated in Figure 2 Today, most telecommunication service providers have
and described in more detail in the following: their own infrastructure, purchasing IP media server
• Infrastructure as a Service (IaaS) equipment and deploying it in their central offices
or datacenters. With cloud computing, it’s no longer
˸ Media processing services running on virtualized necessary to make this large investment; instead,
and media-optimized compute resources, service providers can get media processing capacity
network I/O and storage. from cloud service providers on a pay-as-you-go
basis. This kind of arrangement is likely to employ
a private cloud, and a single cloud service provider
would host some or all of the service provider’s
media processing requirements and capacity.
Cloud Media Processing | Radisys White Paper 11
To better serve cloud computing customers, media Use IaaS Media
server vendors (i.e., equipment manufacturers) are Processing During
optimizing media processing elements to enhance • Provision and utilize media
performance in these environments. Moreover, media- during peaks Consistent End
optimized elements will ease the deployment and Conﬁgure Network • Service Provider AS routes User Experience
for Normal Traﬃc overﬂow calls to cloud • Same service quality,
provisioning of media processing capacity, and leverage • Service Provider media processing features and performance
virtualization technology to increase the flexibility provisions infrastructure • Turn oﬀ when daily peaks 24/7
for normal traﬃc volumes completed
and cost efficiencies availed by cloud computing
infrastructure. For instance, web service provider PSTN
Amazon offers their Elastic Compute Cloud (Amazon Application
Server (AS) Cloud Media
EC210), where media processing elements could be Services 2G
Provider (IaaS) Cellular
optimized for rapid deployment to support on-demand
resizable media processing capacity in the cloud. Service
Example: Cloud Media Processing with VoIP
for Peak Capacity IP Media
A successful service provider offering hosted telecom Smartphone
services has a growing business and needs to with App
increase its capacity. Rather than increase capital
expenditures (CapEx) to fund more infrastructure,
the service provider obtains additional capacity Figure 3. Outsourcing Media Processing to Add Capacity During Peak Demand
from a cloud service provider during peak demand,
just a few hours a day (Figure 3). When evaluating
• Enables tiered-service offerings.
cloud service vendors, the telecom service provider
considered network I/O costs, connectivity quality and ˸ Premium (highest QoS): uses the service
overall network performance with respect to jitter, provider’s internal network and infrastructure.
delay and QoS. Another key aspect was ensuring the ˸ Economy (satisfactory QoS): uses the cloud.
management features could seamlessly integrate
and hand-off calls from the existing infrastructure APIs and Interfaces for Media
to the cloud provider. Services — PaaS Model
Service providers are looking for new ways to offer
The cloud service provider offering included hosted
real-time media processing services, in addition to
media servers whose resources were partitioned as
hosting their own services. One promising avenue is
virtualized instances, allowing them to be provisioned
to open up their platforms to third party developers,
and used, as required, during peak periods. In other
whose applications would interface with and utilize
words, the virtualized instances could be activated
their real-time media processing as a service (MPaaS).
and brought online during peak periods, and later
This business model could be supported by the public
released when demand decreased.
cloud, which could also handle the billing for third parties.
Benefits of outsourcing for service providers:
To pursue this opportunity, service providers must
• Saves capital and operating expenditures expose their media service APIs and interfaces in
(CapEx/OpEx) to enable future growth. languages and formats that can be utilized by the
• Allows media processing resources to broadest cross-section of the software engineering
be gradually migrated to the cloud. industry. The pool of software engineers that develop
applications using SIP or other telecom-specific APIs
make up only a small percentage of the software
Cloud Media Processing | Radisys White Paper 12
engineering industry. Instead, service providers should
embrace more commonly-used Java or HTTP-based Cloud Customer
protocols. Moving forward, there are new standards Customer Care Care Service
under development to better support these business Session Customer Care
models, including JSR-309 for Java-based media
control development and RESTful HTTP interfaces
for control.11 IP Access “Click to Conference”
Internet, IP VPN Request
Example: Cloud Media Processing LTE Mobile
for 3rd Party Developers Conferencing
A web-based contact center determines it could
dramatically increase the efficiency of its support Conferencing Session Cloud
team using real-time audio/video communication MPaaS
features. For instance, customer contact begins
with an over-the-web chat session, but then, using
a “hotkey,” the customer can quickly start a 1-on-1
real-time conversation with a customer service Figure 4. Cloud Media Processing for 3rd Party Developers in LTE Mobile Network
representative or product specialist. Tasked with
creating this feature, an application developer
Telecommunication Applications as
reprograms the contact center’s call flow to
incorporate the new real-time audio/video
Cloud Service Offerings—SaaS Model
communication features, which the contact center After making significant investments in unified
accesses in the cloud, as illustrated in Figure 4. communications (UC) infrastructure, some
enterprises would like to transfer some of the
Example: RCS Video Services support responsibilities to a third party. This presents
(Rich Communications Suite-enhanced) opportunities for cloud service providers with
The GSMA driven RCS-e industry initiative is gaining offerings similar to hosted unified communication
widespread momentum within operators, device service offered by a telecom service provider using
manufacturers, and application service providers. a subscription or pay-as-you-use billing model. One
An important component of RCS-e is multimedia/ example is a hybrid, enterprise-class deployment
video share and video calling services operated and model consisting of enterprise-owned equipment
provided through carrier networks. Interoperability for headquarters or large branch locations, and
across a diverse set of handhelds, media codecs/ cloud-based telecommunication services for smaller
formats, bandwidth optimization, and inter-carrier locations or remote workers.
media connects and interworking are well identified
The key is for service providers to recognize and adapt
challenges. Hosting RCS-e video as part of a Cloud
their offerings for hybrid deployment environments.
Service adds significant value to overcome the
There may also be mid-sized enterprises that decide
interoperability issues where the Cloud Services can
to outsource their entire communication services to a
be shared and the required Media Processing within
cloud service provider instead of buying and operating
the Cloud would provide seamless interoperability
their own UC infrastructure. A hosted conferencing
across RCS and non-RCS handhelds as well as
provider could also manage the application and media
inter-carrier networks, in real-time.12
resources at the customer premises.
Benefit of open platforms for MPaaS providers:
• Generates new services and associated revenues
driven by the imagination of application developers
around the globe.
Cloud Media Processing | Radisys White Paper 13
Cloud Enterprise HQ Location
2G Circuit Access Service
Cellular PSTN, 2G Mobile
Servers UC Desktop
IP Access UC Platform Desktop
Internet, IP VPN IP Phone
Smartphone LTE Mobile TDM
Figure 5. Enterprise UC Integrated with Cloud-based Communications Service
Example: Enterprise UC Integrated with Radisys Products for
Cloud-based Communications Service
An enterprise maintains a large multi-site UC
infrastructure with on-premise PBXs deployed in large Radisys is the leading supplier of media processing
locations. It has become expensive to backhaul branch technologies and solutions for the telecommunications
locations and remote workers, and a cost-effective industry. Radisys media servers have a proven track
option is to use conferencing services in the cloud. record in supporting IP-based media processing
Therefore, cloud-based telecommunication services in 3G networks and are already working in 4G/LTE
may be required to: deployments. A natural progression is the cloud,
where Radisys has expanded its product offerings
• Provide WAN interconnection of remote workers to include optimized cloud-based media processing.
and branch locations.
• Provide backhauls and interconnects to workers The Radisys Software Media Server (SWMS) is a
in larger enterprise locations. Linux-based SIP media server developed for enterprise
and IMS audio/video media processing applications.
Service providers who want to offer cloud-based The software is designed to install and operate on
host audio and video conferencing can easily Intel® multi-core COTS hardware servers, such as
build a solution by running the Radisys SIPware blades in a shelf, and also run in virtualized machines.
Reservationless Conferencing solution on Radisys The product will evolve to support RESTful HTTP/
media servers. JSON and Java interfaces for 3rd party developers.
Benefit of supporting unified communications Alternatively, Radisys hardware-based media servers,
for cloud service providers: the MPX-12000 and CMS–9000, have a very large
• Creates more demand for media servers capacity, providing the scalability, performance,
already deployed in the cloud. and reliability for large-scale hosted deployments.
Their exceptional processing power and I/O
throughput deliver very high performance for
XML–based IVR and messaging applications, while
expanding multi–service versatility for multimedia
conferencing, IP Centrex, ringback tones, IP contact
centers, video communications and complex audio/
video transcoding and transrating.
Cloud Media Processing | Radisys White Paper 14
When these media services are combined with the 6
Source: The Internet Engineering Task Force (IETF),
Radisys SIPware Reservationless Conferencing http://www.ietf.org/rfc/rfc2616.txt.
solution, service providers have a complete hosted
audio and video conferencing solution built on cloud-
Source: The Internet Engineering Task Force (IETF),
computing architectures and technologies. This www.ietf.org/rfc/rfc3261.txt.
portfolio enables hosted providers to increase their 8
Source: W3G, http://www.w3.org/TR/voicexml20/.
service revenues, and service providers to reduce
infrastructure costs when outsourcing their telecom 9
Source: The Internet Engineering Task Force (IETF),
applications or media processing infrastructure. https://datatracker.ietf.org/doc/rfc5707/.
For more information, visit
Source: “The NIST Definition of Cloud Computing,”
by Peter Mell and Tim Grance, October 7, 2009, 11
For more information about GSMA OneAPI
www.nist.gov/itl/cloud/upload/cloud-def-v15.pdf. and HTTP/REST interfaces, visit https://gsma.
Further information on RCS-e and interoperability
Source: TechTarget website,
specifications are available from GSMA/RCS:
Fungible implies computing resources can
be repurposed or re-targeted, as needed,
to support other applications and services.
Source: Report titled “The Video Cloud,” by Morgan
Keegan of Equity Research, November 28, 2011, pg 1.
5435 NE Dawson Creek Drive
Hillsboro, OR 97124 USA
503-615-1100 | Fax 503-615-1121
About the Intel® Intelligent Systems Alliance: From modular Toll-Free: 800-950-0044
components to market-ready systems, Intel and the 200+ global www.radisys.com | email@example.com
member companies of the Intel® Intelligent Systems Alliance
provide the performance, connectivity, manageability, and ©2012 Radisys Corporation.
Radisys and Trillium are registered trademarks of Radisys Corporation.
security developers need to create smart, connected systems. *All other trademarks are the properties of their respective owners.
Learn more at: intel.com/go/intelligentsystems-alliance. May 2012