GBR vs Non-GBR Bearers in LTE: How Operators Decide

An operator would choose GBR QCIs for services where the preferred user experience is "service blocking over service dropping". GBR bearers are typically used for services where it is essential that the service does not degrade due to resource unavailability once admitted. It is better not to admit such a service at all if guaranteed resource availability cannot be ensured. Examples are conversational voice and conversational video. However, once a bearer is admitted, if the eNodeB can no longer sustain its GBR, it should deactivate the bearer. Other services, such as internet browsing, email and chat programs, normally do not require a constant, fixed bandwidth, and Non-GBR bearers can be allocated to them.

A GBR bearer has a guaranteed bit rate, and the network reserves capacity for it through admission control, whether or not the bearer is currently sending traffic. This reserved capacity counts against what the network can admit for other GBR bearers, so each admitted GBR bearer reduces the room available for new ones. As long as the service sends at or below its GBR, congestion-related packet loss is not expected, because a GBR bearer is only set up if sufficient resources are available. Exceptions such as transient link outages can still cause drops. Non-congestion losses are also still possible, within the packet error loss rate of the QCI. For example, conversational voice allows a packet error loss rate of 10⁻².

The MBR limits the bit rate that can be expected to be provided by a GBR bearer; any traffic in excess of the MBR may be discarded by a rate-shaping function. The MBR of a GBR bearer may be set larger than its GBR, which supports rate-adaptive codecs: the network guarantees a minimum bit rate (the GBR), and the bearer may be allowed a higher rate, up to the MBR. Explicit Congestion Notification (ECN) complements this by letting the access network signal congestion, so that the application can adjust its rate in response.

A Non-GBR bearer has no guaranteed bit rate and no resources reserved for it. As a result, a Non-GBR bearer may experience packet loss under congestion. The availability of radio resources for a Non-GBR bearer depends on the total load of the cell, the QCI of the bearer and the UE's radio conditions.

In EPS, Non-GBR bearers have no per-bearer MBR. Instead, the network limits their combined bit rate using two aggregate parameters. The APN-AMBR limits the total across all Non-GBR bearers of the same APN, and the UE-AMBR limits the total across all Non-GBR bearers of a UE. Any one Non-GBR bearer can use the whole of the applicable AMBR when the others are idle. GBR bearers are outside the scope of both.

Whether a service is mapped to a GBR or a Non-GBR QCI is an operator policy decision. It depends on the expected traffic load relative to the dimensioned capacity. When capacity is sufficient, both real-time and non-real-time services can be carried using Non-GBR QCIs only.

Note: Although TCP-based traffic is assigned a Non-GBR QCI, it requires a much lower packet error loss rate (10⁻⁶) than some services mapped to GBR QCIs, such as conversational voice (10⁻²). The reason is that TCP's congestion control treats every lost packet as a sign of congestion and slows down, and it becomes increasingly sensitive to non-congestion related losses as the end-to-end bit rate increases.

Leave a Reply

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