Networks will have a major headache in 2011 and tablets will only make it worse — however wireless Ethernet microwave can ease the pain
With large screen, high performance connected devices emerging in 2010, this trend is likely to increase. 2010 has been the year of the tablet – the iPad tablet – and this is only one of many high performance connected devices we will see in 2011, with a multitude of Android, Linux, Microsoft and Blackberry devices already announced.
With the dramatic growth in mobile data, the divergence of traffic and revenue in mobile networks is well known, but let’s quantify the problem.
Looking at the problem in terms of ‘revenue per MByte’ demonstrates the true scale as shown in Table 1.
Table 1: Revenue per MByte for various services
.
The other significant factor in this equation is the capacity of each service. Whilst SMS and voice only require low capacities, mobile data (particularly mobile broadband for laptop and tablet computers) require data rates that approach fixed broadband. Expectations are in the region of at least 1Mbps per user.
Measured in terms of revenue per bit delivered, SMS is hugely profitable; voice is moderately profitable; mobile data is an expensive drain on resources.
Even with variations in usage, profile and traffic, the trend is clear. The revenue generated per Mb of network capacity has declined with the widespread adoption of mobile data usage, yet the total network cost is increasing rapidly to cope with the demand (Figure 1).
Figure 1.
Source: www.theregister.co.uk/2010/09/29/allot_mobile_data.
Source: www.theregister.co.uk/2010/09/29/allot_mobile_data.
However this problem can also be seen as an opportunity. The opportunity is to embrace the change that is essential and get ahead of the competition.
On the face of it, the obvious solution for delivering high capacity for packet data is a rapid move away from the voice-centric model of the 1990’s – the traditional circuit-switch TDM backhaul – towards a more efficient, internet-centric packet-based Radio Access Network (RAN) and backhaul. An all-IP network has long been seen as the ultimate architecture for an efficient network, but getting to that architecture from today’s hierarchical networks is a very big step. A recent survey1 of mobile operators by Infonetics Research revealed that all were deploying IP/Ethernet backhaul in their network, some in conjunction with TDM backhaul but the majority as a complete replacement for TDM.
The move towards packet RAN has already taken place with High-Speed Packet Access (HSPA) rolled-out across most European Mobile Networks. The move towards packet backhaul, on the other hand, is lagging far behind.
The reason for this lag in the backhaul network is simple. Whilst high capacity data requirements readily move from traditional circuit-switched TDM (e.g. E1 for GSM/GPRS and ATM over E1 for UMTS) backhaul to cost-efficient packet-based Ethernet backhaul, the migration for voice traffic to a packet Ethernet backhaul is not so clear. This is due to the fact that voice traffic is highly reliant on the accurate and stable network synchronisation delivered by the traditional synchronous TDM backhaul network.
The challenge is therefore to cater to the very high capacity demands of mobile broadband packet data whilst also supporting the synchronous requirements necessary for voice. There are a number of possible solutions, each with their own pros and cons:
- Dual backhaul networks: Keep the existing TDM network and adding a second packet network;
- Migrate to hybrid network: carrying a mix of TDM traffic and packet-traffic over the same network;
- Migrate to all-packet RAN: converting all existing TDM RAN to have an IP interface;
- Migrate to packet backhaul, with TDM delivery capability: carrying TDM traffic as packets over a packet network.
Looking at each of these possible solutions in turn, it is clear that whilst some at first glance might be attractive, in the long term, only some make practical and economic sense.
Dual backhaul networks
This immediately solves some of the problems: it maintains a reliable network for voice – one that quite likely has already been optimised, and adds a second network for the high-capacity data without impacting the original backhaul network. But the cost penalties cannot be overlooked. The cost and complexity of running two networks, in terms of network management, equipment, spectrum, site rental, fibre rental etc, means that the original problem, how to reduce the cost of delivering the required network capacity, had not been solved.
Hybrid network
For each hop of the backhaul network, the existing TDM link could be replaced by a hybrid link that combines the TDM and packet traffic then separates it at the other end. This preserves the timing information of the TDM traffic, and goes some way to offering the capacity for packet traffic, but it requires the duplication of switching and cross-connect equipment at each nodal point in the network – driving significant costs and complexities. In addition, this in effect creates separately managed and protected network layers, adding significant operations costs.
It also means there is no statistical aggregation of voice traffic towards the core of the network – one of the key cost benefits of all-IP networks. This solution may achieve the goal of delivering both types of services, but will not achieve the cost per bit economics required to be profitable when delivering high bandwidth services.
Migrate to all-IP network
An all-IP network might seem to be the ideal solution, but the steps to getting there, can be substantial. There is a huge number of TDM-based GSM & UMTS/HSPA basestations already deployed, and the cost of replacing these with packet-based RAN equipment is too high to take in a single step. Therefore all-IP networks might be the goal, particularly with LTE networks now already being rolled-out, but from a cost and management consideration, intermediate steps are likely to prove more financially viable, unless a greenfield network is being deployed.
Migrate to packet backhaul, with TDM delivery capability
Considering the impact in moving to an all-IP network, an intermediate step is very attractive. Having a packet backhaul network delivers the benefits of statistical multiplexing offered by aggregated IP traffic and yet through the use of complementary technologies allows time-sensitive voice traffic to also benefit from such an efficient backhaul. A suitable technology is Pseudowire End-to-End Emulation, which allows TDM traffic to be transported over packet. With pseudowire, the backhaul operator can deploy a single end-end packet network, eliminating costly TDM connections and equipment at intermediate sites. This provides a CAPEX benefit, but more importantly, a significant operations benefit too. But can pseudowire meet the stringent timing requirements of TDM?
Pseudowire is often deployed at the base station using a cell-site router, which interfaces both TDM and Ethernet traffic with IP backhaul. According to research by Infonetics, operator spending on cell-site routers increased by 136% in 2009. It is important to note that TDM E1 links are not only used to deliver raw information, but also to deliver timing information to the base-station – an essential requirement for GSM and UMTS/HSPA networks. In an IP-based network, the ability to deliver TDM data and synchronisation imposes these network requirements:
- Need for minimum end-to-end latency and effective QoS management;
- Optimised bandwidth utilization (minimising the overhead to carry TDM traffic);
- Managed jitter and a variation in packet delay within predefined limits;
- Maintain clock synchronisation between the two ends;
- Maintain packet order.
Defining these criteria quantitatively, and ensuring each is met, is part of the network engineering function but let’s expand them a little further to ensure that a high capacity packet network is going to be suitable for existing TDM-based networks.
Ensuring QoS
In a packet network, delays occur when multiple packets contend for the use of links and other resources. When this occurs, it is important that the time-sensitive packets are given the highest priority and therefore sent first. If there is insufficient capacity, then the lowest priority packets should be discarded. If this occurs too often, then the network isn’t dimensioned correctly and it does not have sufficient capacity. Applying QoS to the IP packet network is essential to overcome problems of contention and suitable QoS mechanisms would be through the use of priority bits in VLAN tagging, 802.1q Ethernet priority tagging, DSCP or MPLS, for example. To ensure consistency throughout the network, end-to-end management of the packetized TDM traffic simplifies network management.
Tagging and managing TDM traffic as high priority in this way not only minimises the latency through the network, it also minimises the variation in latency. If the variation in latency were to become excessive, this would lead to a loss of synchronisation for the TDM traffic.
Avoiding excessive packet delay variation
Whilst effective management with QoS can minimise the variations in latency in a network, there will inevitably be some, leading to jitter at the receiving end. This jitter makes it difficult to deliver the constant bit rate TDM traffic, but is easily managed through the use of jitter buffers – in the same way that streaming media players buffer data before playing it. However buffers introduce end-to-end delay, so there is a trade-off between size of buffer and overall latency: the larger the buffer, the greater the tolerance to jitter, but the greater the latency. This trade-off is an important engineering optimization.
Maintaining clock synchronisation
E1 links deliver TDM traffic at a constant bit rate, delivering not only the data but also a clock that is in sync at the two ends of the link. Because an IP network is not clocked in a synchronous way, the clock from the TDM traffic is lost. However, pseudowire employs clock recovery mechanisms to reconstruct the original clock from the data, thereby effectively delivering a constant bit rate stream over an asynchronous network. Timing recovery with pseudowire can deliver accuracy up to ±0.016 ppm and meet industry standards for jitter and wander masks, easily meeting 2G/3G basestation requirements.
Applying TDM transport requirements to Ethernet microwave
Wireless Ethernet-based backhaul is an attractive technology choice due to its flexibility in deployment, its efficient high capacity data-transport capability, its low cost and rapid deployment and its resilience in backhaul environments. Having defined a mechanism for carrying synchronous TDM circuits over a packet network, it is then important to look at the features and capabilities of an Ethernet microwave system to see how they map onto a pseudowire backhaul network (Table 2).
Table 2. For a larger image.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
