Mobile Internet - Backhaul Bottleneck
Operational tools for managing mobile Internet.
The rapid growth of video and other rich-content traffic over 3G/4G backhaul networks is a well-known phenomenon. What isn’t generally recognized is that this traffic often comes in unpredictable bursts that are a challenge to manage.
In order to manage network bandwidth and subscriber-specific voice, video and data flows without causing performance degradation, providers need a means for dynamically allocating bandwidth while maintaining ubiquitous quality of service (QoS) across users.
Deploying managed carrier Ethernet for business services and backhaul capabilities offers more cost-efficient dynamic bandwidth and network visibility for proactive network monitoring and quality assurance in real time.
Carrier Ethernet is accompanied by new monitoring solutions that help to ensure a uniform quality experience by addressing backhaul bottlenecks in real time while minimizing operators’ management operational expenditures.
Frame relay, and to a lesser extent ATM, were the primary packet data service offerings in the 1990s and early 2000s. But in the mid-2000s, Ethernet services became the choice offering for telecom service providers. The reason for this is that Ethernet services offered better bandwidth scalability and flexibility while riding the much better volume cost across both consumer and business markets.
The service provider network infrastructure has been TDM-based since its inception. TDM far predates frame relay and ATM, and with packet data traffic becoming the dominant traffic in service provider networks, TDM is likewise no longer adequate or efficient for packetdominated traffic.
Therefore service providers across the world are evolving to packet-centric infrastructures, and mobile networks are no exception. Mobile networks are riding the same Ethernet wave for the same reasons.
The bottom line is that packet networks are very efficient for packet data because you can perform statistical multiplexing on the traffic. In other words, in most cases you can send higher amounts of traffic without increasing the size of the pipe.
Packet networks can scale quickly and easily, as many of the user Ethernet interfaces today are at least 100 Mbps and quite common at 1,000 Mbps (1 Gbps), with user traffic varying across those rates.
Traffic policing can minimize this phenomenon, but the number of users will still be an issue because it can increase significantly at any given time. The increasingly popular mobile broadband services fueled by smartphones and other devices add further demand and challenges to the packet networks.
John Donovan, AT&T’s CTO, described the challenges for mobile networks in his CTIA Wireless & Entertainment 2009 keynote as:
- Data is much less predictable than voice traffic is, and is much more "bursty," with massive traffic flows up and down that can dwindle to nothing quickly, as opposed to the steady flow of voice.
- We no longer have the luxury of predicting where and when we might see [traffic] growth.
- We have to be able to manage our networks so that everyone gets a good experience.
The challenges can be summarized as:
- The ability to recognize when the traffic goes up or down as it happens in real time.
- The ability to maintain quality user experience in real time.
In order to monitor traffic in real time, even without any notice of spikes, the capability needs to be implemented in hardware at wire speed. But with customer applications evolving so rapidly, hardware also needs to be programmable to accommodate the evolving traffic requirements.
Real-time monitoring and non-intrusive monitoring can support the standard SLA (service-level agreement) metrics: latency, jitter, frame and byte loss, and throughput in real time needs to perform at a per-flow basis. A flow can be an endto-end Ethernet virtual circuit or a priority class within an end-to-end Ethernet virtual circuit. Furthermore, a section – a subset of a flow – can be monitored, especially when a flow traverses over two or more service provider networks.
In order to do so non-intrusively, the system monitors the user traffic inline and via the Ethernet OAM frames.
As far as throughput is concerned, there are other factors affecting it due to application and higher layer behaviors such as TCP. To prove in real time that their subscribed throughput is honored, synthesis traffic using Ethernet OAM frames can be created while adjusting its load in real time without affecting the customer traffic as it increases and decreases.
Once network performance is monitored and reported, an appropriate action can be taken to eliminate or minimize the performance problem.
Service providers can have the ability to set thresholds for the performance objectives. Once the threshold level is reached, an alarm can be generated to alert the network operation center (NOC) as it occurs in real time. Protection switching can be performed to switch the network flow to a predefined alternative path.
In order to obtain a quality experience whenever and wherever possible, the service provider should utilize all resources available at any given moment. With the ability to monitor the bandwidth throughput in real time, service providers are able to observe when and where resources are exhausted and underutilized.
Furthermore, real-time feedback can be provided to those who are in charge of the user policy to throttle user traffic accordingly. As mobile Internet traffic can be the major contributor to the network congestion, a mobile Internet offload is being considered. Again, real-time feedback can be provided to those who perform such tasks more effectively.
The explosion in mobile broadband traffic has posed challenges to service providers’ network infrastructure and user quality of experience. These challenges will only become greater in the foreseeable future, and, consequently, service providers must have ways to monitor traffic in real time and – more importantly – act upon them accordingly again in real time.