eMBMS offers LTE service providers an effective way to lower cost per bit when delivering the same content simultaneously to multiple end users.
As the multicast standard for Long Term Evolution (LTE), Evolved Multimedia Broadcast Multicast Service (eMBMS) allows multimedia content to be sent once and received by many end users. This “one-to-many” distribution mode can be a valuable alternative to unicast when a large number of users are interested in the same content.
For example, during live streaming of major sports or news events, unicast must send the same video to every user individually. But multicast takes advantage of the inherent broadcast qualities of wireless networks to send the video only once to reach an equal number of end users.
In these types of scenarios, multicast makes more efficient use of the available spectrum and reduces cost per bit. At most, it will consume resources equivalent to that of the worst-performing link in a sector. Therefore, the most common uses for multicast are likely to include distributing video, music, software, news, weather, ads and other data to a mass audience. The content can be live or preloaded for later usage, which has the potential for additional cost savings.
More flexible spectrum usage
The eMBMS standard is supported in 3GPP R9, and initial deployments are expected to begin in 2012. With eMBMS, LTE networks will be able to support broadcast and multicast along with unicast, and the same frequency layer can be used for all these distribution modes.
Therefore, to allocate resources appropriately, eMBMS standards have specified a high-level mechanism to reserve network resources for multicasting throughout a session — and to release them when it ends. That is, the necessary resource blocks will be “borrowed” from the total available spectrum for the duration of the eMBMS session, and then returned when it concludes.
The response time for these reserve-and-release capabilities will depend on each vendor’s solution. But the key point is that eMBMS standards have established flexibility for spectrum usage and have eliminated the need for dedicated spectrum, which was an issue prior to 4G networks. For example, MBMS for UMTS and MediaFlo by Qualcomm both require dedicated resources. This makes it harder for Mobile Service Providers (MSPs) to meet customer demands for anytime, anywhere video consumption, and neither MBMS nor MediaFlo has been a huge success. Now eMBMS in LTE offers a more flexible option.
The underlying reasons for greater efficiency
In a wireless network, it might initially appear that unicast will use fewer resources than multicast when transmitting a particular video. In fact, the one-to-one and one-to-many modes each use roughly the same bandwidth (resource blocks) within a sector.
That’s because at the physical layer, due to the nature of wireless transmissions, a one-to-one video transmission is actually broadcast throughout the entire sector — instead of reaching only a single customer, as occurs in wireline networks. As shown in Figure 1, the intended wireless recipient tunes in to the transmission, while other mobile devices are designed to ignore it, because it is not addressed to them.
Figure 1: One-to-one is actually broadcast but made inaccessible to all but the recipient
In one-to-many transmissions, on the other hand, a large number of mobile devices can tune in and receive the video from a single transmission. So, if eight mobiles within a sector all want the same content, multicast can transmit it just once. But unicast would have to transmit it eight times — using on the order of eight times the resources that would be required in multicast.
An illustration of cost savings
If a MSP could consolidate demand for the same content — and use eMBMS to deliver it to multiple users at the same time — the cost per Mb could be significantly reduced, as shown in Figure 2. The extent of the cost reductions depends on two factors:
- The number of simultaneous users in a sector receiving the same content – In the cost model shown in Figure 2, this concurrency is called Sector Copy Factor (SCF), and modeling shows the savings that multicast enables when SCF equals 4 and 40.
- The fraction of the entire network traffic that is multicast – If a higher proportion of traffic is multicast, then the net spectrum required is lower. Less equipment will be required to operate the network for the same number of megabits delivered.
Figure 2 provides a relative cost comparison with and without eMBMS under different assumptions for the proportion of traffic being multicast. eMBMS-derived discounts were applied to the unicast cost data to calculate potential savings. The throughput for eMBMS was based on the worst-performing link in the sector, because, at most, multicast consumes resources equivalent to the worst link. Accordingly, eMBMS throughput is assumed to be 59 percent of the average unicast throughput, when using the same amount of spectrum.
Figure 2: Cost reductions with multicast
Better reception and throughput
To provide better reception and throughput, eMBMS can operate on a single frequency across a group of cells, known as a Single Frequency Network (SFN). A specific set of frequency resource blocks is reserved for the SFN, so performance remains consistent when an end user moves from one cell to another. And because the cells are synchronized, interference is prevented. Users get better reception at cell edges, because their mobiles simply pick up the combined signal.
Figure 3 shows how cells can be grouped administratively into clusters of SFNs, which can be useful for serving geographically fragmented markets.
Figure 3: Administratively linked SFNs
When to use multicast
Broadcast is an appropriate distribution mode when an audience for the specified content is large but concentrated geographically. Unicast can be efficient when the audience is narrow and sparsely distributed. Multicast provides an alternative for distributing content to meet “clustered” demand. When considering which clusters can support a profitable business case, the multicast deployment decision can be assessed by market and by region.
Mobile service providers are currently planning eMBMS solutions for live streaming, as eMBMS sessions can be set up dynamically and resources can be shared efficiently with unicast sessions. However, multicasting also has potential for use with other applications — particularly when it is combined with preloading of content, which can enhance the overall cost reductions.
Multicast in LTE networks
In an operational network, it is relatively easy to add multicast functionality with minimal equipment. The required elements to support eMBMS are shown in Figure 4, which provides an overview of the eMBMS logical architecture. A brief explanation of key functions follows this illustration.
Figure 4: eMBMS logical architecture
Broadcast/Multicast Service Center (BM-SC): The BM-SC schedules an MBMS service, announces the service to user equipment, authorizes users, allocates bearer service identification, and initiates and terminates MBMS bearer resources. It may optionally be the direct interface point for content providers. This entity also terminates the SYNC protocol over the M1 interface that is required to synchronize the radio interface transmission of the same data from all eNBs in an MBSFN area.
Multimedia Broadcast Multicast Service Single Frequency Network (MBSFN) area: This area consists of a group of cells coordinated to achieve an MBSFN transmission, which is a simulcast transmission technique that sends identical waveforms at the same time from multiple cells. This coordinated transmission is seen as a single transmission by a mobile device.
Multimedia Broadcast Multicast Services Gateway (MBMS GW): This gateway’s function is to send IP multicast packets to all eNBs that are part of the eMBMS service. It performs MBMS session control signaling (session start/stop) toward the E-UTRAN using an interface to the mobility management entity (MME).
Evolved Node B (eNB): eNB is an existing element in LTE. To support eMBMS, an eNB is upgraded to support SYNC protocol with BM-SC. The eNB joins IP multicast, terminates the multicast control channel and indicates multicast session start/stop to mobile devices (UE).
Multi-cell/multicast Coordinating Entity (MCE): This is a logical function that could reside in another network element, such as an eNB. It performs admission control, allocation of radio resources throughout the MBSFN, and MBMS session control signaling — and makes decisions on radio configuration.
Mobility Management Entity (MME): MME is an existing element in LTE and is involved in the signaling path to eNBs. The BM-SC signals to the eNB through MME.
SYNC protocol: This helps the eNB identify the timing for radio frame transmission and detect packet loss.
M2 and M3: These are signaling interfaces on the control plane, and M1 is a user plane interface to the eNB carrying IP multicast.
Conclusion
eMBMS multicast capabilities can provide valuable alternatives to unicast for distributing many types of live and non-live multimedia content. They take advantage of the inherent broadcast qualities of wireless networks to send content only once to reach multiple end users, thereby making more efficient use of the available spectrum and reducing cost per bit. In addition, eMBMS sessions can be set up dynamically — and share resources with unicast sessions — which eliminates the need for dedicated spectrum.
To contact the authors or request additional information, please send an e-mail to techzine.editor@alcatel-lucent.com.
Which parameters affect the response time’s value? Is BM-SC in charge of the reserve-and-release procedure?
The question being asked is in the context of the spectrum reserve-and-release capabilities in an eMBMS network.
The MCE logical entity is responsible for admission control and the allocation of the radio resources used by all eNBs in the MBSFN area. Availability of information about current resource allocation in all eNBs is one obvious factor that would determine the response time. There are other factors, but it would be implementation dependent and would vary from vendor to vendor.
So, when a mbms service finishes,the mce inserts unicast data where was that service allocated?
As far as radio resources are concerned MCE can reserve and release bearers for eMBMS only. The release could be used up by unicast (via RRM) but MCE does not take part in unicast bearer allocation. 3GPP 36.300 says:
(MCE functions include) Besides allocation of the time/ frequency radio resources this also includes deciding the further details of the radio configuration e.g. the modulation and coding scheme.
Could Common Subframe Allocation (CSA) and/or Spreading Fractor (SF) parameters which the response time’s value affects? Or is it more linked with the UE’s behaviour?
The radio resource allocation itself is quite fast. The factors to consider are MCE’s coordination with different base stations for single frequency network which implies signaling delays. In addition the UEs need to be made aware of the new allocation which also requires signaling.
Parameters such as CSA pattern, CSA period, MSP and MSI are pre-defined by the operator or MCE decides these values? Is it possible that eNBs cannot allocate MBMS data because unicast data is using almost all the radio resources? Thanks.
>> Is it possible that eNBs cannot allocate MBMS data because unicast data is using almost all the radio resources?
Yes it is possible.
Without going into a vendor’s implementation details let me try to answer your
questions. Whether CSA pattern etc. are predefined or not depends a lot on the service complexity, advantages/cost of having such feature, solution features available from vendors for the offered services, etc. In other words it can implemented/operated in many ways and a solution is usually altered to fit a particular operator’s service package. For
example in “Further
Efficiencies with eMBMS Preloading” we show a use case where 1Mhz (not necessarily contiguous) is pre-allocated during off-peak hours (subject to operator policy precedence) for a set
of services. In this case it “seems” optimal to pre-configure the CSA pattern, CSA
period etc. Even in this case to figure out which optimal frequency slices to allocate and what CSA pattern should be used is done by complex modeling work.
Once the MCE allocates the radio resources, is it in charge of informing to MBMS GW about which services should the MBMS GW forward to eNBs? For example, if MCE deactivates an MBMS service, the MBMS GW should stop transmitting the data from this service. Is it right? Thanks.
MCE’s role is limited to what is specified in the standards like admission control, etc. Peeling deeper is all implementation details that is vendor specific.
I’m wondering if MCE allocates MBMS subframes for all MBMS services. For example, MCE allocates nine MBMS services for 3 different MBSFN areas where in each area is transmitted only three of these MBMS areas. It means MBSFN area 1 (S1..S3), MBSFN area 2 (S4..S6) and MBSFN area 3 (S7..S9). It means that an eNB which belongs from Area 1 will use MBMS subframes from S4 til S9 for unicast data or for extra subframes for S1 til S3? I hope it is clear. Thanks a lot.
Sorry, it is not clear. For example what are the assumptions in your question? Are these 3 MBSFN areas topogically non-intersecting? i.e. no common sector or eNB.
The standards does not restrict 1 MCE for the entire network as you seem to be alluding to. So if you assume 1 MCE per MBSFN area for simplicity that would answer many questions.
I mean only one MCE controlling all three MBSFN areas. Several eNBs and cells form an unique MBSFN. My assumptions are that MCE is connected directly to several eNBs, these eNBs are splitted in different MBSFN areas. Is it possible to use only one MCE for an entire network (in this case 3 MBSFN areas)? I know that more than one MCE is possible. How one eNB will manage MBMS subframes (from an MBMS service)which are not running in its MBSFN area?
One MCE per MBSFN can elegantly solve the resource tracking issues. I do not see anything in the standards restricting one MCE for the entire network. But be aware that such solution would require the resource allocation logic to be more sophisticated.
Is MBMS multicast mode supported by EPS?
If it not supported how UE could receive multicast service.
>Is MBMS multicast mode supported by EPS?
Yes it is supported e2e. MBMS GW as shown above in Fig 4 is at root of multicast tree
Are BM-SC, MBMS-GW, MCE physical entities?
If they are not physical entities where are they located in these phyisical entities of LTE architecture: E-UTRAN S-GW and PDN-GW?
BM-SC, MBMS-GW, MCE are logical entities. They may be located anywhere as long as they conform to the functional architecture. The choice of location may give some advantage/disvantage. For example MCE may be located in eNB as described above or it could be centralized and located with MME or exist as a separate entity to serve a cluster of eNBs.