When launching new services, mobile operators often worry whether these services will inhibit existing infrastructure or services. In particular, USSD services are often suspected of depleting radio channel capacity and inducing voice channel outage. Clearly, a service that exposes the operator’s core asset to risk needs to be carefully scrutinized before launch.
This article will discuss these concerns and the possible risks.
The following USSD services are technologically possible:
USSD Phase 1 Service (GSM 02.90) – a service launched by a mobile subscriber (phone-initiated session or USSD Pull)
USSD Phase 2 Service (GSM 03.90): Launched by a mobile subscriber or initiated by the network (USSD Pull & Push)
USSD Phase 1 and 2 services can occur simultaneously with a phone call (Call Related USSD Services) or without a voice connection (Call Independent USSD Services).
Call related USSD services use FACCH (Fast Associated Control Channel) for data transfer, call independent USSD services use SDCCH (Independent Dedicated Control Channel).
USSD supports a maximum of 160 bytes of user data per message (the upper limit may be less than 160 bytes depending on the underlying protocol layer).
The speed of the FACCH channel is approximately 140 bytes per second and that of the SDCCH channel is approximately 83 bytes per second.
Therefore, any USSD message can be delivered in less than 2 seconds.
Call-related USSD messages are less interesting for commercial use, since the user is highly unlikely to call and use USSD. However, it is possible for a USSD network-initiated message to be delivered while the subscriber is speaking. However, the risk of breaking a voice channel is minimal, which is supported by the fact that FACCH is not a dedicated channel but an associated one, which means that it is implemented by stealing bits from the associated voice channel, if exists. If such bit stealing is allowed frequently and for long periods of time, it will certainly degrade speech quality to some extent, but it will not prevent the subscriber from speaking. In the case of a significant deterioration of the speech quality, the situation can be improved by the operator by rearranging the priorities of using the FACCH channel.
The situation is more complex in the case of USSD Call Independent services. It is actually a channel resource used for many purposes. These include call setup (including authentication), call completion, location updates, and informing the network that the subscriber is still within the service area. Consequently, without a free SDCCH channel, a mobile phone cannot access mobile services (ie the subscriber cannot make a call).
In addition, USSD and Short Message Service (SMS) messages are also communicated over SDCCH logical channels. That said, an effective use of SDCCH logical channel resources is crucial for reliable and efficient mobile communication.
According to the GSM 05.02 guidelines, a cell can have up to 8 SDCCH channels. A typical cell configuration has 4 or 8 SDCCH channels.
A USSD session uses one SDCCH channel. It means that 8 simultaneous USSD sessions exhaust the capacity of the SDCCH channel of a cell. Therefore, other subscribers served by this same cell could not make calls or send/receive SMS messages.
Seems like a serious objection to using USSD services. But is not; in real life this situation does not happen.
A proof can be found in a German research article “Performance evaluation of GSM signaling protocols in USSD”. According to the document, new call initiations are more likely to block existing SDCCH channels than new USSD sessions. It is backed by our 4 years of USSD service experience with MTS; actually, we never had problems of this kind.
In fact, problems can arise with the massive use of USSD in a limited geographical location, that is, when the service requests are concentrated in a small number of cells. This could happen within a network-initiated ad campaign at a stadium, for example.
Such situations require extra attention from the operator. First, each operator can decrease USSD session timeouts and therefore decrease the probability of simultaneous service load spikes. A transaction within a dialog session is limited to 30 seconds and the total duration of the session is limited to 600 seconds (actually it can differ on various hardware platforms), so there is room to change these parameters. Timeout timers are defined in GSM 09.02.
Second, it is possible to define priorities for network services. Therefore, phone calls may receive the highest priority, while SMS/USSD services may have lower priority.