5G Fronthaul–Latency and Jitter Studies of CPRI Over Ethernet

Scalable deterministic scheduling for WDM slot switching Xhaul with zero-jitter

“…Regarding the traffic matrix, CPRI or isochronous traffic flows, for the packet size of 1250 bytes and 10 Gbit/s channel capacity, are chosen from the sets 1 = {2.5, 5, } Gbit/s [1] and 2 = { .67, 3.33, 6.67} Gbit/s, that have different basic periods. Under these assumptions, for different simulations, we consider one of the following four traffic scenarios: a) Scenario 1: The traffic profile is "hub-and-spoke" (more precisely we suppose that each bridge node communicates with the master node, and vice versa, and that the amount of traffic exchanged in both directions in this communication is the same).…”

“…]; 2) The cost of WDM-TRX (CONF. 1, virtual channels) is supposed to be • (where parameter ∈ [1,10]). Fig.…”

“…V compares the sources of jitter and latency in N-GREEN and Ethernet. While variable in Ethernet [1], latency is fixed in N-GREEN in transit and at the insertion, and the jitter is different than zero (and limited to ) only for the physical channel scheduling in N-GREEN. When channel is virtual the jitter is equal to zero.…”

“…Although Ethernet is a mature technology, Ethernet-based fronthaul exploiting statistical multiplexing has difficulties to support synchronization constraints, low jitter (<65ns) and latency (<100 µs) for the CPRI (Common Public Radio Interface) traffic [1] in this network segment.…”

Virtualized eNB latency limits

“…For these reasons, new upper layer functional splits have been proposed by 3GPP in TR38.801 and a Next Generation Fronthaul Interface (NGFI) is under definition [6]. However, different functional splits have got different requirements in terms of capacity and latency [7]. This paper evaluates experimentally the latency limits and the fronthaul bandwidths requirements in a NGFI network architecture.…”