In general an XB-Engine needs to receive data for a subset of channels from N F-Engines where N is the telescope array size. This is complicated to configure and requires many F-Engines. In order to bypass this, an F-Engine simulator has been created that simulates packets received at the XB-Engine (i.e., packets from multiple F-Engines destined for the same XB-Engine). This simulator benefits from a server with a Mellanox NIC and ibverbs to run. This fsim simulates the packet format used by katgpucbf.
The minimum command to run fsim is:
fsim --interface <interface_name> <multicast_address>[+y]:<port>where
- <interface_name> is the name of the network interface on which to transmit the data;
- <multicast_address> is the multicast address to which all packets are sent. The optional [+y] argument will create additional multicast streams with the same parameters each on a different multicast addresses consecutivly after the base address. <port> is the UDP port to transmit data to.
The data rate per multicast address is adc_rate * N_POLS * SAMPLE_BITS * antennas * (channels_per_substream / channels). With the default arguments, this is 1712000000 * 2 * 8 * 80 * (512/32768) = 34.24 Gbps.
For improved performance, use :option:`!--ibv` to enable ibverbs acceleration [1].
This requires the CAP_NET_RAW capability to run. The easiest way to do it
is with spead2_net_raw.
See the fsim source code (in src/katgpucbf/fsim/) for a detailed
description of how the F-Engine simulator works and the useful configuration
arguments.
| [1] | See the spead2 documentation for information on the requirements (particularly hardware requirements) for ibverbs. |