UMTS和EPS网络中在处理下行用户平面时的不同？

爱卫生

E-UTRAN的架构引入了针对位置管理功能的变化和修改。因此，很多报文的处理方式发生了变化。如下图所示：

在UMTS网络中，以高速的HSDPA为例，下行数据被缓存了两次，因此有两个不同的数据重传ARQ（自动重复请求）。外层的循环是由RNC的RLC层功能完成的，而MAC层的Hybrid ARP循环是由NodeB完成的。因此HARQ是离无线接口比较接近的基于链路适配机制的技术。当RLC被配置为AM（确认模式）或UM（非确认模式）时---采用基于IP的业务传输over HSDPA案例 - RLC层也负责数据的加密。

在E-UTRAN中，压缩和加密功能都是由位于eNodeB中的PDCP层完成的。所有的重传机制也都是由eNodeB完成的，数据传输过程仅要求做一次包头压缩和IP报文加密的缓冲。另外，有意思的是，在UMTS/HSDPA案例中，当终端在它的HSDPA会话过程中如果变更了NodeB，那么整个NodeB的数据缓存将丢失。当然，丢失的数据将被重传，这要干洗整个RLC ARQ loop（当被配置为确认模式时）或通过更高层如TCP来重传（如果RLC被配置为非确认模式但使用了TCP的情况下）。相应的代价是数据恢复时间增加了，因为RLC以及TCP层的交互延迟远比NodeB MAC层高得多。这也是RNC中应最小化NodeB中缓存的报文级别原因。

在E-UTRAN中，还要感谢X2接口（eNodeB之间的接口，数据缓存可以直接在source和target eNodeB之间传递，这将最小化在底层无线协议层丢包的可能性。

最后一点需要注意的是在分组域核心网Gateway节点的处理上没有变化。站在用户平面角度看，2G/3G SGSN或Serving Gateway的角色限制在数据报文的路由，而压缩和加密特性仍保留在接入网络侧完成。

翻译而来，附上原文：

The architecture of E-UTRAN has introduced some modifications about the location of the main functions within the network. As a result, there are some differences in the way packets are handled. These changes are presented in Figure 4.2, which focuses on the downlink User plane.

In UMTS, taking high-speed HSDPA as an example, data packets used to be buffered twice, as a result of the separation of the two data retransmission ARQ (Automatic Repeat Request) loops. The outer loop supported by the RLC is located in the RNC, whereas the MAC Hybrid ARQ loop is located in the NodeB, as HARQ is based on link adaptation mechanisms close to the radio interface. When RLC is configured in AM (Acknowledged Mode) or UM (Unacknowledged Mode) – which is the case for IP-based services transported over HSDPA – the RLC layer is also in charge of data encryption.

In Evolved UMTS, Compression and Ciphering are both supported by the PDCP layer, located in the eNodeB. As all retransmission mechanisms are also located in the eNodeB, data packet processing requires only one buffer of header compressed and ciphered IP packets.

In addition, it is interesting to note that, in the UMTS/HSDPA case, the whole NodeB data buffer will be lost when the terminal changes NodeB during its HSDPA session. Of course, the lost data will be retransmitted, thanks to the overall RLC ARQ loop (when it is configured in Acknowledged Mode) or even higher at the TCP application level (if RLC is configured in UnacknowledgedMode and if TCP is used). The price to pay is an increased data recovery time, as the reaction time at RLC and TCP levels is much higher than in the NodeBMAC. This is the reason why the amount of buffered packets at the NodeB level shall be minimized by the RNC.

In the E-UTRAN case, thanks to the X2 inter-eNodeB interface, the data buffer may beforwarded between source and target eNodeB, which helps to minimize the probability ofpacket loss at the lowest radio protocol level.

The last point to note is that there is no difference in the Packet Core gateway node asregards to packet handling. From a User plane perspective, the role of the 2/3G SGSN or the Serving SAE GWis limited to packet data routing, as the compression and ciphering features remain in the Access network.