There's a lot of industry activity (and hype) around NFV, SDN and Cloud RAN. In parallel, there's a separate initiative to put more application intelligence at the network edge. Why would you want to do that and how does this relate to Small Cells? I spoke with Brian Naughton of IBM, Art King of SpiderCloud Wireless and Caroline Chan of Intel to improve my understanding of this novel approach.
A quick recap of SDN and NFV
There's a lot of activity throughout the telecom industry to allow use of generic server hardware platforms. Many products that previously required custom-designed hardware (telecom appliances), such as voice switches (MSC) or text message routers (SMSC) are now merely software applications. Virtualisation allows these to scale and/or share servers spread across multiple data centres.
These initiatives are aimed at the communications infrastructure itself, reducing cost, improving efficiency, scalability and resilience.
The mass take-up of Client/Server Smartphone Apps
Quite independent from SDN/NFV is the architecture of where and how end-user applications are deployed. We've seen the enormous computing power and memory capacity of our smartphones and tablets that allow them to host and run many different types of Apps. These are usually coupled with a cloud server that makes best use of both local and centralised functions. A small number of standard software environments (iPhone and Android) provide the scale to enable commercial App development for even the smallest niche markets.
What is Edge Cloud Application Computing?
There are numerous network applications which are not directly standard telecom or radio functions, but add-ons that provide additional or customised services. In a similar way to smartphone Apps, these can be managed by the Enterprise IT department or provided as a service by a mobile network operator or hosted cloud partner.
Just as software developers don't want to produce separate variants for each type of smartphone, there is also a drive to standardise and simplify enterprise network application environments. As with Smartphone Apps, some of the intelligence might be located in the cloud and some locally – it depends on the specific application.
The three drivers for Edge Cloud Apps
There are three main reasons why you might want to have the application running locally on an enterprise server rather than remotely in the Cloud:
- Latency: Where you need to turn equipment on or off quickly, responding in a very short time.
- Robustness: Activities that need to continue even when the connection to the cloud is broken.
- Data filtering and consolidation: Sensors, video cameras etc. in the building might generate huge amounts of data, of which only a small proportion is useful information. Filtering data streams down into key information events saves bandwidth and improves system efficiency.
What does IBM provide to support Edge Cloud Computing?
Brian Naughton of IBM told us "From an IBM perspective, we don't know or care about radio access products or small cells. Our expertise is in Enterprise Applications that manage our Customer's business processes. More and more of our Customer processes require the real time analysis of data from a plethora of machines or devices on site, e.g. analysis of surveillance cameras or tracking consumers to send offers etc.
"For the reasons above, it makes sense to run some applications or parts of them locally onsite. Applications originally developed for centralised operation can be deployed and reconfigured for deployment at the edge.
"IBM brings a suitably scaled solution to the party. It's a software appliance extension of the IBM Smart Cloud designed to be deployed in a store or Customer site connected to local devices/machines. The appliance determines which Cloud applications can be downloaded and run locally. These application are connected to the IBM Cloud allowing 3rd party Application developers to build innovative offerings that integrate directly with an Enterprise building or site.
"There are already many vertical markets, such as retail, hospitality and stadiums that can benefit from this style of solution. Adding value by integrating mobile Applications to local point of sale systems, streaming video directly to fans in a stadium, or using small cells to track real time location."
SpiderCloud's Services Node, which encompasses a services module powered by an Intel Xeon processor, can host and support multiple apps as virtual machines.
Some examples of enterprise applications available for their Services Node include:
- Security: Threat detection and prevention by blocking malicious packets originating from mobile devices before they reach the core network
- Location services: IBM have demonstrated handset-to-location video and advertising push services for use at venues and shopping malls
- Data Caching: Saguna can show backhaul bandwidth savings and user experience benefits using a content cache for large campus, venues and shopping malls
- PBX Integration: Tango Networks and Druid Software PBX for Hospitals
Would this lock-in each Enterprise to a single network operator?
Art King, director of enterprise services and technologies with SpiderCloud, points out that there is no technical reason why Enterprise small cells cannot support multiple operators. SpiderCloud's latest dual-band Radio Node includes two independent radio functions, so a system could (in theory) be configured to support two or more network operators. However, SpiderCloud's operator customers want to be able to differentiate with small cell systems and services, so they see no demand for multiple operator support from mobile operators.
Being able to support a wide and growing selection of third party applications combined with the ability to customise or develop on request brings more than just a basic vanilla product option.
His view is that many enterprise and stadium owners have been teased by the DAS offer of fully multi-operator/neutral host. While these provide wireless connectivity, there's no option for added value enterprise applications.
What does Intel provide for these solutions?
Caroline Chan points out that Intel's contribution here is providing the silicon and supporting development platforms for Small Cell vendors such as SpiderCloud to host and support these Edge Cloud Apps. Although this can scale from a single Small Cell upwards to the largest data centre, it probably makes best sense to use a 1U sized server box for each building.
The use of Small Cells in remote and rural applications may be particularly relevant, where backhaul connections are expensive, high latency, low bandwidth but some local service level needs to be maintained during outages. Oil rigs, remote mines, and cruise ships spring to mind but there may be many other contenders.
For example, Intel has one partner who used their reference design for Small Cells deployed in remote African villages. Backhaul links are low bandwidth and high latency. Data caching and local switching made the system operation much more efficient, lower cost and responsive.
Disclaimer: Spidercloud is a sponsor of ThinkSmallCell