Multicore Media Platforms Support Cloud Infrastructure
To adequately address cloud computing requirements means new hardware, new infrastructure, new software… and sometimes even new business models are needed. It’s a “Gigantic Data” dilemma best solved using Intel’s latest solutions.
|Figure 1: The IoT has the potential to further propel the “gigantic data” growth curve sharply upward. Due to the transition to a world with billions of connected, intelligent devices, there is the reality that the massive amount of machine-to-machine (M2M) real-time data will outpace what humans can produce by multiple factors of 10, or a number so large it is hard to fathom.|
The “Internet of Things” (IoT) is an influential force that is transforming the telecom and cloud infrastructure industries. A huge portion of the force of IoT is the amount of content continually added to online servers. Adding to the content explosion, video is becoming the dominant communication vehicle on the Internet. Supporting this claim is Intel who predicts mobile video content will double every year, and that two thirds of all mobile data traffic will be video by 2015. Furthermore, at a recent conference Intel presented the following facts: Each minute, another 30 hours of video is posted on YouTube; Twitter handles 100,000 tweets; Facebook handles 6 million page views; iTunes App Store processes 47,000 app downloads. The number of devices on the Internet already equals the population of the world (i.e., approximately 7 billion). And, that number is expected to double in two to three years with mobile Internet traffic predicted to make a staggering 11-fold increase.
The sheer volume of the data generated by all forms of communication has been appropriately referred to as the “gigantic data” problem. Without question, this reality calls into question whether today’s server computing platforms or cloud infrastructure are up to the task of future IoT requirements (Figure 1). But cloud computing really is different; it’s not just marketing-speak for the same old client/server model. To adequately address cloud computing requirements means new hardware, new infrastructure, new software and sometimes even new business models are needed. Starting with a clear understanding of the data expansion trends affecting IoT requirements for the cloud puts infrastructure solutions into better focus.
Trends Causing Gigantic Data
There are several significant existing and future forecasted market trends contributing to the “gigantic data” phenomenon. While blogs, social networking and media streaming are definitely affecting Web 2.0 applications, much of the growth in Internet usage is coming from gadgets or devices, not people. Unattended “embedded” devices are popping up everywhere that include everything from Internet-enabled soft drink machines to utility meters, traffic lights or a whole group of electricity-driven applications (Figure 2).
|Figure 2: M2M-connected devices like this intelligent Coke machine can predict and respond to a customer’s behavior. This machine was demonstrated onstage by an Intel executive at the company’s IDF2012. (Photo courtesy: Chris A. Ciufo.)|
Today, mobile, IPTV and content delivery providers are pushing more and more content from their networks, which, in turn, are spawning ever greater demand from end-users. Compounding the issue is the increase in 4G LTE service availability along with associated mobile tariffs that also put a strain on these providers. Higher bandwidth service is leading to elevated quality content expectations from consumers that include HD with the ability to access OTT (over the top) and TV Everywhere services such as Netflix, iPlayer and Hulu. Content delivery providers use the latest technology advancements to encode video. They are finding, however, that this is a processor-intensive job especially when outputting multiple videos into multiple formats, and is often used by broadcasters, cable companies, and large production companies who need to encode large amounts of media into multiple formats, all at the same time. This is why it is so crucial to select a computing platform that offers a highly scalable and distributed processor approach that will allow providers to share application workloads.
Moving Mobile Devices to the Cloud
Several challenges confront content providers in supporting mobile devices in the cloud, and these issues are the motivating factors suppliers are using to build powerful cloud computing platforms. Providers must contend with services monetization, increased bandwidth requirements, power constraints, accommodating multiple content delivery standards, expandability, and reducing costly OPEX (operating expenditure).
Another reality is that even with the increase in data handled in the network, the profile of data packets passing through the network will be in sizes different from the traditional mobile device connected to the network. Therefore, what is called for is a more distributed computing approach in the cloud that can enable highly efficient management of the cloud infrastructure. The new structure of IoT introduces Web 3.0 with a simplified and structured interaction for M2M communication without any human interaction.
Part of the solution is efficient transcoding in the cloud, which is seen as the answer to accommodate both power consumption and bandwidth-hungry video content requirements. Tackling energy efficiency, new server platforms need to provide extensive and smarter power management that adapts power consumption to the actual workload as well as dynamically powering up or down processors independently for significant energy savings.
For expandability, cloud infrastructure needs to scale, meaning that it needs to grow without changing its nature. It is important to note that not everything scales—small servers do not necessarily scale into larger server networks. The reality is that scalability that performs in both small and large scale deployments can be difficult to design into microprocessors, systems and networks. From a cloud infrastructure perspective, scalability is by no means assured with most hardware technology, so it is crucial to check the system specs to make sure scalability is an integral part of the platform. And, every part of the cloud infrastructure needs to scale at once—the network, the memory, the performance, the cooling—or the overall system doesn’t work. Upgrading one component without the others doesn’t make sense.
|Figure 3: Security is becoming a top priority, as shown here at IDF2012.|
Accessibility and user privileges, authentication, encryption, malware and OS vulnerabilities through managing OS patches are the top security threats on mobile devices connected to the network (Figure 3). Disaggregation of compute resources in the cloud platform allows a more secure, network infrastructure compared to traditional massive multicore platforms. For example, traditional platforms that use dual socket server architectures and a massive memory plus a software layer for virtualization actually become an additional weak point to securing the network.
A number of hardware companies are touting a new generation of cloud servers that scale down the traditional web/cloud server with reduced power consumption, lower cost, smaller footprint, and less heat dissipation. But beware, some suppliers are offering old technology that they are masquerading under new labels. Mobile providers must be weary and realistic about what modern “cloud-ready” systems can really do, and plan accordingly. Lane Patterson, CTO of US-based Equinix cautions about naive expectations: “The cloud does not automatically back [itself] up, nor provide a tool to automatically mirror the solution or application to another location. Hence, if there is a failure or fiber cut within or to the data center, the user will experience a service outage.”
Multicore Media Platform Solutions
Transcoding requires intensive compute cycles, and is a fundamental necessity in the network, as numerous devices require a multitude of different bitrates, resolutions and codecs. Intel multicore processor architectures, such as the Intel Core i7 and Xeon E3-1200 Processor Series are well-suited to support these types of processor-intensive video encoding. Specifically, the Intel Core i7-3615QE processor is designed for media optimization applications and features Intel Clear Video HD technology and Intel Quick Sync Video 2.0 for improved visual quality, HD media playback, and the ability to decode and transcode simultaneous video streams while freeing up the CPU for other tasks. In addition, low power, high performance Intel processors help developers create platforms that easily scale and share the workloads of web, M2M and mobile applications deployed in cloud infrastructure. And, Intel processors contribute to platforms with an enhanced and comprehensive power management suite that permits dynamic powering up and down when workloads change.
Intel’s latest processor architectures supply the optimal feature set with low power and high performance for new computing platforms that deliver the right combination of energy consumption, size and scalability to enable reliable cloud-based media content delivery and transcoding applications. The company’s processors provide the key foundation allowing media platform developers to design computing resources that are more than traditional servers, but true versatile building blocks that make it easier to create cloud-based networks. Now, network equipment providers (NEPs) who offer hosted services and IPTV, Cable, Cloud and Mobile Cloud service providers can use new media cloud platforms as a framework for optimized streaming content to mobile devices. In addition, the modular design of these media platforms can be the building blocks for new applications used for mobile and fixed video, unified communications and Video Surveillance as a Service (VSaaS).
|Figure 4: The modular design of the Kontron SYMKLOUD Media platform supports running multiple applications, including transcoding, across multiple independent low-power, high-performance processors. It offers a scalable, future –proof solution and features up to 18 Intel Core i7-3615QE Quad-Core Processors with integrated Intel HD Graphics 4000.|
For example, the Kontron SYMKLOUD MS2900 Media platform is based on Intel’s highly scalable and distributed Intel Core i7-3615QE Quad-Core processors (Figure 4). Supporting shared application workloads, it integrates switching, load balancing and processing in a 3-in-1, 2U rackmount platform design that gives cloud service providers the ability to configure clusters of highly dense 42U cabinets that require four to eight times fewer fiber and copper cables. This type of design approach enables improved power efficiency and scalability, and a much lower CAPEX (capital expenditure) for new deployments in the network, without any limitation on Carrier Grade High-Availability and switching capabilities.
Making Mobile Cloud Infrastructure a Reality
To fully support cloud service providers and hosted services with highly reliable server computing platforms and cloud infrastructure design resources requires a shift from legacy, purely processor-driven hardware to more scalable and versatile cloud-enabled Web 3.0 infrastructure equipment. Next-generation data centers requirements will continue to evolve so it is crucial that providers look for new hardware and software solutions that are fully integrated and application ready. These solutions must also provide improved power and cluster management, and deliver cost-effective 5-nines high-availability. The good news is that a series of new Intel-based media platforms are now available. These platforms are truly cloud-worthy solutions service providers can use to prepare for video growth predictions as well as drive the growth for OTT and TV Everywhere, HD formats and devices, mobile video and video surveillance.
Sven Freudenfeld manages Business Development for the Communications Product Business Unit at Kontron focusing on the telecom vertical market, including the AdvancedTCA, MicroTCA, AdvancedMC, Communication Rackmount and Cloud Computing product lines. Sven possesses more than 15 years experience in voice, data, and wireless communications, having worked extensively with Nortel Networks in Systems Engineering, Sanmina-SCI in Test Engineering, and Deutsche Telekom in Network engineering.