A mountain not yet climbed: Machine to Machine (M2M) Diameter signaling management requirements
Date: Thu, 11/15/2012 - 12:50 Source: By Ben Volkow, VP Product Development, F5 Networks
Machine-to-machine (M2M) is one of the hottest topics in the communication industry today. Beyond the buzz, telecom network planners are scrambling to figure out how their networks,whichhave only supported subscribers’voice and data communications until now, will support the operational side of M2M
And they are working fast as M2M technology trials and deployments are starting to appear everywhere; heated discussions on the topic are taking place, and equipment vendors are proposing solutions for this unusual market.
However, before M2M delivers on the great opportunities for communications service providers (CSPs) to generatesignificant revenue streams, vast initial investments are required and huge technical challenges have to be overcome on an unprecedented scale. Unlike existing networks which typically host a few million or tens of millions of subscribers, M2M requires building and architecting networks to scale and support billions of "subscribers," or rather every connected machine.
Among the many technical challenges in M2M, one that CSPs are less familiar with due to its dependence on knowledge of Diameter protocol is network signaling. As a member of the F5 Traffix team who has been developing and designing Diameter signaling products since 2005, I’d like to outline a few of major signaling challenges in migrating to M2M, and how they differ from the typical subscriber-centric LTE network signaling requirements.
M2M signaling versus LTE signaling: A different animal
There are many differences between signaling in M2M networks and LTE networks. Although both are based primarily on Diameter protocol, and both comply with the same 3GPP standards, this is where the similarities end and the differences begin:
1. Signaling volume
When we refer to M2M networks, we are discussing a network in approximately ten times larger than LTE in subscriber size. With signaling already growing today in LTE networks at three times the speed of data, according to many market analysts, the leap in scale of signaling volume represents a challenge never yet encountered by CSPs.
Furthermore, 3G networks are built around voice and data. Machine-to-machine networks do not include voice and will hardly use data (there will be metering, alerts and warnings, but machines rarely browse on the Internet). M2M networks are completely signaling focused. Based on our deep experience in signaling, we foresee there will be two to three times more signaling in M2M networks compared to "regular human" LTE networks. When this increase is added to the already higher LTE signaling volume and the scale of "subscribers," you have an unprecedented quantity of signaling.
2. Signaling behavior
The behavior of human subscribers as it relates to telecom networks is for the most part known to network planners, with predictable peak usage occurring before and after work, on weekends andholidays, etc.However, machines are not humansand requiredifferent network architecture and scaling to support theirbehavior.
For example, consider what happens in a network when one million ATMs (cash machines) simultaneously send updates to a central office at the end of the day, or when ten million electricity meters send an update on the last day of the month all at 8a.m. This will create a tsunami of signaling hitting the network all at once. Now think what happens if the ATMs don’t receive a reply in time; think about the retransmissions (an automatic repeated request) and how all of these will affect the network.
The same goes for connection times. Again, human voice and data behaviors have been tracked, and their signaling has been well architected in a network. We know we will have some connections that are on constantly, sending signaling all the time. We know which connections will be very short and will turn off after a specified period. However, some machines are sending thousands of messages a day in a very rapid manner. These patterns and behaviors vary completely fromtypical network and signaling behaviors known to date.
In short, the learning curve on Diameter signaling is still on the rise for LTE networks. No doubt that learning curve will have to be revisited totally for M2M.
3. Network and signaling infrastructure
In terms of hardware, LTE networks continued to be built in the same way as legacy networks. CSPs can use proprietary hardware or off-the-shelf hardware. In either method, the approach remains similar to how legacy networks were built.
Machine-to-machine networks will use emerging cloud and virtualization technologies. There are both technically and commercially strong reasons for this major shift. From my discussions with CSPs on M2M architecture, moving ahead to cloud and virtualization is the only viable option. Any other way just doesn't make sense.
Cloud and virtualization are game changers and there are many reasons to push these forward. However, they also have signaling effects that need to be taken into account in the M2M network design process. For example,a 4G cloud-basednetwork will be more fragmented with many more dedicated network elements and functionalities.
In the M2M signaling domain,there are many more Diameter routers (Diameter Routing Agents(DRAs) and Diameter Edge Agents (DEAs)). However they need to be software-based and will behave differently. In contrast to LTE, in which DRA core routers need to handle millions of transactions per second, in M2M the Diameter routers will have fewer transactions per router. However, we will see M2M Diameter routers hosted in the virtualized network and be tuned to handle M2M signaling behavior, cope with longer session times and bursts of signaling messages. In short, the management and synchronization of hundreds of Diameter routers is a huge challenge.
4. Legacy/LTE interworking
As LTE networksare being rolled out today, they are still connected to much of the existing legacy network infrastructure. Although several CSPs tell us that they are using brand new LTE network elements for a separate architecture,there will always remain somelegacy databases, legacy charging systems, legacy switches and routers to handle some of the traffic, and some interworking (connectivity)to the legacy management or OSS systems. In other words, there remains plenty of legacy signaling RADIUS, LDAP, CAMEL, MAP, GTP, HTTP, and other protocols.
However, M2M networks are a different story. They are being built from scratch predominantly in a cloud-based virtualized environment. Whereas it makes sense from a technological standpoint to design a separate network to handle M2M traffic, it means that this new, out-of-the-box LTE network based on Diameter signaling will have much more Diameter signaling and much less interworking and legacy signaling compared to the "human subscribed" LTE networks.
There are many more differences between signaling and Diameter in both LTE networks and M2M networks, but this covers the main ones. Services are different, and thecommon use cases are different,but this article highlightsthe conceptual differences that must be addressed at the onset of network planning.
F5 has been working closely with CSPs to address the M2M signaling challenges. During our discussions we understood that addressing the M2M signaling market requires a special re-tuning of current signaling technology to M2M’sspecific requirements and network characteristics.
Here are a few tips to keep in mind when planning a M2M network:
1. You must have a software-based approach. Sticking to hardware only won't work. You need to be able to provide software and work in cloud environment.
2. Addressing the M2M signaling market requires scaling and designing your signaling equipment such as DRAs and DEAs to handle the bursts and signaling scale of M2M. You need to have the right management in place, the right reporting tools, and most of all, the right architecture.
3. Diameter in M2M networks is not SS7, not SIP and not LTE network Diameter and requires different knowledge, tuning and design.
4. DRAs and DEAs in M2M need to be just that. Don't talk about integrated DRAs and PCRFs, DRA and charging systems or DRA on SBCs. Firstly, this is not needed in virtualized environment. Secondly, the challenge is so high that service providers need a real DRA and not a PCRF or SBC (session border controller) disguised in a DRA costume.
CSPs are facing several significant challenges as M2M becomes the next revenue-generation source. Network signaling volume and behavior need to be addressed, as well as the network signaling infrastructure itself. Meanwhile, as LTE networks continue to be developed, they need to retain interoperability with legacy infrastructure elements. As organizations maintain a flexible, software-based approach that can scale to meet bursts of high demand, they will be prepared for the future of continuously growing network traffic without slowing down their business.
About the author: Ben Volkow
Ben Volkow serves as VP of Product Development at F5 Networks, joining the company with F5’s acquisition of Traffix. Ben was CEO and co-founder of Traffix and is considered a visionary and industry expert in signaling and Diameter protocol. Prior to founding Traffix, he filled management roles at several communications companies.