The final presentation by Diametriq is very interesting because they presented interesting way of mining the control plane. Thee case study presented was of a 'silent roamer' who is not going to spend money while roaming because he is not sure how much money is spent. This can be exploited by the operator to offer flat packages, 1 day pass, etc. to get some revenue from these roamers. Their presentation included some animations that cannot be shown while being embedded. Please download the PPT from Slideshare to view them.
NTT Docomo recently published a new article (embedded below) on congestion control approaches for M2M. In their own words:
Since 3GPP Release 10 (Rel. 10) in 2010, there has been active study of technical specifications to develop M2M communications further, and NTT DOCOMO has been contributing proactively to creating these technical specifications. In this article, we describe two of the most significant functions standardized between 3GPP Rel. 10 and Rel. 11: the M2M Core network communications infrastructure, which enables M2M service operators to introduce solutions more easily, and congestion handling technologies, which improve reliability on networks accommodating a large number of terminals.
Light Reading carried an interesting cartoon on how M2M works. I wouldnt be surprised if some of the M2M applications at present do work like this. Jokes apart, last week the UK operator EE did a very interesting presentation on Scaling the network for the Rise of the Thing.
A question often asked is "What is the difference between the 'Internet of Things' (IoT) and 'Machine to Machine' (M2M)?". This can generate big discussions and can be a lecture on its own. Quora has a discussion on the same topic here. The picture above from the EE presentation is a good way of showing that M2M is a subset of IoT.
Its also interesting to note how these 'things' will affect the signalling. I often come across people who tell me that since most M2M devices just use small amounts of data transfer, why is there a need to move from GPRS to LTE. The 2G and 3G networks were designed primarily for Voice with Data secondary function. These networks may work well now but what happens when the predicted 50 Billion connected devices are here by 2020 (or 500 Billion by 2030). The current networks would drown in the control signalling that would often result in congested networks. Congestion control is just one of the things 3GPP is working on for M2M type devices as blogged earlier here. In fact the Qualcomm presentation blogged about before does a decent job of comparing various technologies for IoT, see here.
I have blogged about about IDC here and here. If the eNB is interested in knowing if the device is having an interference issue it can ask the UE to send this message in the RRC Conn Reconfiguration message. The UE would send the message if it has interference issues.
Inter-frequency handover is a good solution in case the UE is experiencing interference.
To assist the base station in selecting an appropriate solution, all necessary/available assistance information for both time and frequency domain solutions is sent together in the IDC indication. The IDC assistance information contains the list of carrier frequencies suffering from on-going interference and the direction of the interference. Additionally it may also contain time domain patterns or parameters to enable appropriate DRX configuration for time domain solutions on the serving LTE carrier frequency. Note that the network is in the control of whether or not to activate this interference avoidance mechanism. The InDeviceCoexIndication message from the UE may only be sent if a measurement object for this frequency has been established. This is the case, when the RRCConnectionReconfiguration message from the eNB contains the information element idc-Config. The existence of this message declares that an InDeviceCoexIndication message may be sent. The IDC message indicates which frequencies of which technologies are interfered and gives assistance to possible time domain solutions. These comprise DRX assistance information and a list of IDC subframes, which indicate which HARQ processes E-UTRAN is requested to abstain from using. This information describes only proposals, it is completely up to the network to do the decisions.
All UEs are members of one out of ten randomly allocated mobile populations, defined as Access Classes (AC) 0 to 9. The population number is stored in the SIM/USIM. In addition, UEs may be members of one or more out of 5 special categories (Access Classes 11 to 15), also held in the SIM/USIM. These are allocated to specific high priority users as follows. (The enumeration is not meant as a priority sequence): Class 15-PLMN Staff; -"- 14-Emergency Services; -"- 13-Public Utilities (e.g. water/gas suppliers); -"- 12-Security Services; -"- 11-For PLMN Use.
Now, in case of an overload situation like emergency or congestion, the network may want to reduce the access overload in the cell. To reduce the access from the UE, the network modifies the SIB2 (SystemInformationBlockType2) that contains access barring related parameters as shown below:
For regular users with AC 0 – 9, their access is controlled by ac-BarringFactor and ac-BarringTime. The UE generates a random number
– “Rand” generated by the UE has to pass the “persistent” test in order for the UE to access. By setting ac-BarringFactor to a lower value, the access from regular user is restricted (UE must generate a “rand” that is lower than the threshold in order to access) while priority users with AC 11 – 15 can access without any restriction
For users initiating emergency calls (AC 10) their access is controlled by ac-BarringForEmergency – boolean value: barring or not
For UEs with AC 11- 15, their access is controlled by ac-BarringForSpecialAC - boolean value: barring or not.
The network (E-UTRAN) shall be able to support access control based on the type of access attempt (i.e. mobile originating data or mobile originating signalling), in which indications to the UEs are broadcasted to guide the behaviour of UE. E-UTRAN shall be able to form combinations of access control based on the type of access attempt e.g. mobile originating and mobile terminating, mobile originating, or location registration. The ‘mean duration of access control’ and the barring rate are broadcasted for each type of access attempt (i.e. mobile originating data or mobile originating signalling).
Another type of Access Control is the Service Specific Access Control (SSAC) that we have seen here before. SSAC is used to apply independent access control for telephony services (MMTEL) for mobile originating session requests from idle-mode.
Access control for CSFB provides a mechanism to prohibit UEs to access E-UTRAN to perform CSFB. It minimizes service availability degradation (i.e. radio resource shortage, congestion of fallback network) caused by mass simultaneous mobile originating requests for CSFB and increases the availability of the E-UTRAN resources for UEs accessing other services. When an operator determines that it is appropriate to apply access control for CSFB, the network may broadcast necessary information to provide access control for CSFB for each class to UEs in a specific area. The network shall be able to separately apply access control for CSFB, SSAC and enhanced Access control on E-UTRAN.
From a presentation by Peter Zidar in the Small Cells Global Congress 2012.
The above picture shows that even though the amount data traffic carried by dongles is much more than the amount of traffic carried by the mobile phones, the amount of signalling is far higher from the mobiles than that of dongles. This is mainly because the mobiles need to conserve the battery power and for this reason they disconnect from the network as soon as there is no need for exchange of data. Remember the Fast Dormancy issue in the smartphones? If not see this post.
M2M is going to be big. With the promise of 50 Billion devices by 2020, the networks are already worried about the overloading due to signalling by millions of devices occurring at any given time. To counter this, they have been working on avoiding overloading of the network for quite some time as blogged about here.
The feature to avoid this overload is known as Extended Access Barring (EAB). For E-UTRAN, in Rel-10, a partial solution was implemented and a much better solution has been implemented in Rel-11. For GERAN a solution was implemented in Rel-10. The following presentation gives a high level overview of EAB for E-UTRAN and GERAN.
In Rel-11, a new System Information Block (SIB 14) has been added that is used specifically for EAB. Whereas in Rel-10, the UE would still send the RRCConnectionRequest, in Rel-11, the UE does not even need to do that, thereby congesting the Random Access messages.
The following is from RRC 36.331 (2012-09)
Here is my attempt to explain the difference in overload control mechanism in Rel-8, Rel-10 and Rel-11. Please note that not actual message names are used.
As usual, happy to receive feedback, comments, suggestions, etc.
The problem is divided into two parts, the Access network part where the Air Interface is the bottleneck and the core network part which can easily be swamped by the overwhelming amount of Signalling due to more intelligent billing system and always on devices with background applications generating much more amount of traffic as would have on an older system. Lets look at them in turn.
Core Network Signalling Storm:
As I reported earlier, Diameter has been highlighted as a way of salvation for the operators with dozens of use cases but due to its immaturity has caused outages and have given it a bad name. As Connected Planet mentions, "According to one signaling expert, launching the iPhone’s browser, for example, instantly sets off about fifteen individual network signaling requests. Beyond that, 4G network software elements supporting increasingly sophisticated mobile service scenarios “talk” to each other at rates that traditional TDM/SS7-based networks never had to deal with." Hopefully a stable implementation of Diameter protocol will help not only solve the signalling storm but will help generate new models for charging and revenue generation.
A presentation by Ed Gubbins of Current Analysis, comparing the big vendors of Diameter Signalling is available here.
Access Network Signalling Storm:
My thinking is that the Core Network Signalling problem will become an issue some years down the road whereas the Access Network Signalling problem will be seen sooner rather than later. In fact for 3G/HSPA the problem is becoming more visible as the market has matured and more and more users are moving towards using smartphones, Since LTE rollouts are in its infancy (in most markets) the problem is still some way away.
One of the reasons for Signalling storm is the incorrect APN name. I reported earlier about Telefonica's approach to solve this problem by using 'Parking APN', see here.
Also embedded below are couple of presentations from the Signalling Focus day that talk about the problem from Access Network point of view
I got a chance to attend the 'Handling the Surge in Signalling Traffic Focus day' at the LTE World Summit. In fact I got this opportunity through Diametriq, who were the sponsors of this event and were kind enough to provide me a free pass :) As a result, they get a little plug below.
We got off to a flying start with an Introduction to the need of Signaling followed by a brilliant presentation by Martin Pineiro from Telecom Personal, Argentina.
This was the only presentation that looked at the Access Network Signalling. All other presentations focussed on Diameter signaling. Telecom Personal have 4 carriers, 1 is used for 3G and other 3 for GSM.
Above is their revenue share for different services. The data services really took off for them when they offered a flat rate if 1 peso per day for unlimited data.
Their average dongle data consumption is 2GB/month and average smartphone is 200MB/month.
They do have a simple definition of Smartphone, which is a device that produces 10+ packet connections per day. The device that is most popular in their network is Motorola and Apple devices produce highest data load but their comparison of devices from different manufacturers showed they all produced similar signalling traffic.
One final point highlighted was that OS & Apps are not part of test and certification so we should get better understanding of that to help avoid signalling overload in future.
Interesting to hear that they are 40 year old company with 300+ customers in 100+ countries.
There is a shift coming in the usage plans with multi access roaming. Some sessions will go over WiFi and some over the mobile network. Plans with OTT allowance are already here and will be more common. There may be opportunity for end users to earn allowance as part of loyalty scheme. The main thing for operator to think is how to get a revenue share from advertisement.
Diameter 2.0 is coming. The signalling storms, if not handled properly can cause disruption (congestion) internationally, if the interconnect is not handled properly.
Today we use Diameter 1.0, tomorrow it would be Diameter 2.0. Diamater 2.0 us "nervous system" approach.
Diamater is much less predictable than SS7 but this could be because of Immaturity of Diameter.
Real networks like the one above is out in the field. An example of n/w is one with 140 point to point connections.
DRA (Diameter Routing Agent) is a new topology introduced by 3GPP and DEA (Diameter Edge Agent) was introduced by GSMA.
The network does not want to spend million of dollars in one go so they start by deploying individual components first and then depending on the use cases this scales up as they add more components.
Next up was the Panel Discussion:
Diameter is first protocol that has dedicated vendors offering monetisation of protocol as well
Early operators would have deployed Diameter 1.0 so they can evolve by putting DRA for one use case and so on.
When operators want to monetise using diameter, the signalling problems may become worse
Adding VoLTE may increase Diameter Signalling by 3 times
What is meant by monetisation of Diamater is that in SS7, the focus was on reliability, etc. but in Diameter, the operators can leverage PCRF and as a result monetisation. A new use case can also be a OTT proxy that can leverage advertisement revenue.
The forecast for Diameter is couple of 100 million for this year and growing. There are many components including Router, Roaming, Charging, Security, Interconnect capability, Aggregating relationships with small carriers and OTT service providers, etc.
Next up was Marjan Mursec of Telecom Slovenia:
Some interesting facts from them is that they have a public WLAN n/w, GSM with EDGE as fallback and have rolled out HD voice. Their Data usage surpasses voice and Voice and SMS is still growing as can be seen below.
Above shows the data usage increase after they rolled out all you can eat package. They were then forced to introduce fair usage policy.
Their upgrade paths include RAN, Core, Backhaul.
They think they have a big signalling challenge over S1-MME interface. One wrong configured user is sending 4 requests/second. 12,500 users can be enough to reach congestion (ZG: Maybe they should look at PDP Context Parking). Over the S1-U interface, Narrowband users can send 50 packets/sec. 40,000 users at 13.6kbps can saturate the network and the routers will be overloaded.
Interesting to see that GRX is a service in IPX above.
I think the main point of above is that Diameter by itself is not enough and a mechanism like IPX is required for roaming scenario.
For LTE a new service called LTE Signalling exchange (LSX) can be created within IPX. iBasis has just launched Sandbox for testing Roaming, Charging, Interoperability, etc.
Will LSX bring the roaming costs down? Its operators call but it does provide a foundation and in the next 2-3 years, data roaming costs should come down dramatically.
It should be noted that GRX is an IP network without QoS. Its a service within IPX. Security is also a service within IPX and GSMA based compliance should be there for proper and secure interoperability.
Voice over IPX is not of much interest, especially because there is no return of investment and HD voice cant be send over IP.
One question during Q&A was, why not put this functionality in the cloud and avoid complexity of having another physical box in the system. The answer was that CDRB is implemented to be compliant with cloud deployment but operators have not yet taken this step. The customers are deploying physical boxes but shared infrastructure would be much more efficient.
Everyone is talking about LTE-LTE roaming but there is a need for LTE-3G and LTE-2G so some translation may be required between Diameter and SS7.
Diametriq provides a single platform for signalling between any service (2G/3G/4G) and possibility to enhance.
Next up was another Panel Discussion:
One observation is made is that as compared to the ITM Optimisation event, where the operators were more worried about the OTT players eroding revenues, the focus here was that how Diameter can help monetise the OTT services,