Overload Control in MME

The MME serves as the initial control-plane anchor for a UE entering the core network. Given this central role, it is important to design MME congestion and load-control mechanisms so that they maximize the number of users the network can serve, preserve overall network integrity, and ensure that emergency calls and other high-priority services remain available even under load.

Overload Control via the eNodeB

When network load exceeds capacity, the MME can invoke overload control, using internal triggers to detect the overload condition and signalling it to its serving eNodeBs over the S1 interface, so that the eNodeBs themselves reduce the volume of traffic reaching the core. This happens in four steps:

    • Detection — the MME's own internal algorithms determine that it is entering an overload condition.
    • Notification — the MME signals Overload Start to its serving eNodeBs, indicating the degree of overload and the GUMMEI(s) it wants protected. This identifies the overloaded MME itself (an MME may have more than one GUMMEI), not any alternative MMEs — the MME does not need to tell the eNodeB about other MMEs in the pool, since the eNodeB already knows the full pool configuration from S1 Setup, and its own NAS Node Selection Function (NNSF) simply redirects new UEs to a different, already-known MME instead.
    • Enforcement — each affected eNodeB begins overload supervision, reducing incoming traffic destined for the named GUMMEI(s) by rejecting non-emergency, non-high-priority call originations toward it, while continuing to allow emergency and high-priority traffic through.
    • Recovery — once the MME's overload condition clears, it sends Overload Stop to the affected eNodeBs, which then lift the restrictions for the GUMMEI(s) named in that message.

    Overload Triggered by User-Plane Nodes

    MME load can also increase due to unusual events originating from the user-plane nodes — the SGW and PGW — rather than from UE-side signalling alone.

    Overload triggered by SGi-side disruption and Overload control by PGW

    A problem on the SGi interface (the PGW's connection to external IP networks, suh as the internet or an IMS network) can disrupt user-plane traffic; when this happens, the affected bearers or PDN connections may need to be modified, deactivated, or re-established, and each of these actions generates control-plane signalling that ultimately passes through the MME. A localized fault on the user-plane side can therefore translate into a spike in MME signalling load, even though the MME itself is not the source of the original problem.

    If the PGW experiences disturbance affecting the PDN served by a particular APN, it can begin rejecting PDN connection or additional bearer activation requests for that APN. When this happens, the PGW may request the MME to apply additional backoff time to UEs associated with that APN — the MME then rejects any further PDN connection requests for it, and may supply the affected UEs with a backoff timer for session management procedures such as PDN connection or bearer activation requests. In severe cases, the MME can escalate this further, combining the backoff timer with active PDN deactivation to reduce load on the affected APN. Managing this risk means the network must account for — and, where possible, limit — the signalling that SGW/PGW-side issues can trigger toward the MME.

    Overload triggered by DL Data Notification and Control:

    MME can also control the volume of DL Data Notification triggers sent by the SGW, which would otherwise cause paging of idle UEs and increase network. Based on its own load threshold, the MME can request the SGW to reduce DL data notifications for low-priority traffic — traffic priority itself being determined at the SGW by operator policy and the ARP value of the associated bearer. The MME provides the SGW with a throttling factor and a throttling delay; the SGW applies this overload control accordingly, and resumes normal operation once the delay period expires.

    Congestion Control via Backoff Timers

    Beyond the mechanisms above, the MME can apply backoff timers directly to UEs, as part of mobility management and session management procedures. These timers can be applied either generally (across all APNs) or scoped to a specific APN experiencing problems — the underlying mechanism (reject the request, attach a backoff timer) is the same in both cases; only the scope differs.

    Mobility management backoff:

    when the network is congested, the MME can reject a UE's Attach Request and include a backoff timer in the rejection. While this timer runs, the MME rejects any further Attach Requests from that UE — tracked per UE — except for detach requests and high-priority or emergency access attempts, which are never blocked. This restriction remains in effect until the UE is either paged by the network or moves into a new PLMN outside its equivalent PLMN list; for a UE that remains connected, Tracking Area Update and handover commands continue to be followed as normal even while the timer is running. This mechanism is particularly valuable in situations where repeated reconnection attempts by UEs would otherwise worsen an already-congested network condition.

    Session management backoff:

    the same reject-and-backoff approach applies to session management procedures such as PDN connection establishment or bearer activation/modification — whether the underlying congestion originates within the MME itself or is signalled from the PGW (as described above). Here, the timer is tracked per UE and per APN, so the restriction applies specifically to that subscriber's access to that particular network, without affecting the UE's other, unaffected connections. Beyond simply rejecting new requests, the MME may go further and actively deactivate an existing PDN connection associated with the congested APN.

    Cause Code Values

    EPS can also provide various responses to UEs, known as Cause Code Values — for example, indicating congestion, or that CS service is temporarily unavailable — so that the UE can act accordingly and he rate at which devices access Network can be regulated.

    Summary

    Together, these mechanisms — eNodeB-level overload control, Cause Code Values, SGW/PGW-triggered signalling management, and mobility/session management backoff timers — give operators a powerful set of options for managing network resources under load. However, many of these procedures depend on correct implementation and standards compliance by the UEs themselves to be fully effective.

Leave a Reply

Your email address will not be published. Required fields are marked *