Part of: OCUDU on COSMOS · 4G / 5G & O-RAN · Step 2 of the OCUDU learning path · Previous: ocudu-soft · Next: ocudu-72
Put a real 5G Standalone cell on the air: the OCUDU base station drives a USRP N310 radio in band n41, the Open5GS core runs beside it, and a commercial Quectel RM520N-GL modem on a second node registers, gets an IP address, reaches the internet and measures its throughput.
This is the second step on the OCUDU path. The gNB, the core and the attach milestones are the ones you saw in ocudu-soft; two things are new. The radio is real, so the gNB becomes a real-time program: it must deliver samples to the N310 on a fixed schedule, and when it cannot, the radio underflows and the modem sees a degraded or dropped link. And the UE is a commercial modem with its own firmware and defaults, which exposes interoperability details a matched software UE never shows, such as the PDU-session type it asks for.
Concepts
- USRP N310 — a software-defined radio. The gNB streams baseband samples to it over the server's data network and it converts them to RF.
ru_sdr— the part of OCUDU's configuration that drives an SDR through the UHD driver: radio type, address, master clock, sample rate and gains.- COTS UE — a commercial off-the-shelf modem, here a Quectel RM520N-GL on USB. It is controlled with AT commands and carries data over a QMI interface (
wwan).- PDU session type — a UE asks for an IPv4, IPv6 or dual-stack IPv4v6 session. The OCUDU gNB accepts IPv4 only, so the core must hand out IPv4 sessions.
- TAC — the tracking area code. The gNB must advertise the PLMN and TAC the AMF serves (
00101, TAC 7), or NG Setup is refused.
| Link | Between | Carried over |
|---|---|---|
| Radio | gNB ↔ N310 | baseband samples over the gNB server's data network (UHD, 61.44 Msps) |
| Air | N310 ↔ modem | band n41, 40 MHz, 30 kHz subcarrier spacing, TDD, centred on 2654.55 MHz |
| N2 — control | gNB ↔ AMF | NGAP over SCTP port 38412 on the gNB server's loopback interface |
| N3 — user data | gNB ↔ UPF | GTP-U on the same loopback interface |
| N6 — data network | UPF → internet | the core's ogstun interface (192.168.100.1/22) and a NAT rule on the gNB server |
| UE data | modem ↔ UE node | QMI data call on the modem's wwan interface, with a route scoped to the UE pool |
The core runs on the gNB server, exactly as in the no-radio tutorial; only the path between gNB and UE has changed. There is no fronthaul network here: the link to the radio is sample streaming, not O-RAN 7.2.
Where the radio's address comes from. You never type the N310's address. The parameter ocudu_usrp_node names the radio's inventory node and ocudu_usrp_plane the network it streams on; with ocudu_usrp_addr: auto every mode looks the address up in the COSMOS inventory when it starts. auto means "ask the testbed, now". To pin an address instead, set ocudu_usrp_addr in your --params file.
| Domains | sb1 |
| Nodes | 2, named in OMF_NODES |
| On the console | python3, ansible-playbook, ssh |
cosmos-tutorial check ocudu-ota verifies, on each node:
Not in the package - you provide:
You will also need:
sb1 for the gNB server, the N310 and the UE node, and SSH access to the sb1 console — see Getting started and SSH access;The N310 is shared by every sb1 user. Nobody else may stream from it during your run, so coordinate before you start. No licence is involved.
On the sb1 console, name the gNB server first and the UE node second, and ask whether the tutorial can run. The check is read-only.
export OMF_NODES="<gnb-node>,<ue-node>"
cosmos-tutorial check ocudu-ota
It ends with RUNNABLE or NOT RUNNABLE and the reasons: the OCUDU build on the gNB server, a 5G modem on the UE node's USB, no other gNB running, and the N310 answering from the gNB server. If the gNB server lacks the image, load it first:
omf load -i ocudu.ndz -t <gnb-node> -r 0 -o 2400 -b 600
omf tell -a on -t <gnb-node>
Do not use a gNB server that still carries isolcpus from an O-RAN 7.2 run: this gNB is not pinned to isolated cores, so it gets only the remaining ones and stalls. A fresh ocudu.ndz carries no isolation; grep -o 'isolcpus=[^ ]*' /proc/cmdline on the node must print nothing.
export OMF_NODES=<node>,<node>
cosmos-tutorial check ocudu-ota
cosmos-tutorial run ocudu-ota # ansible
cosmos-tutorial run ocudu-ota --mode manual # manual
cosmos-tutorial jupyter ocudu-ota # jupyter
cosmos-tutorial run ocudu-ota --mode parity # parity
cosmos-tutorial run ocudu-ota --mode teardown # teardown
All three modes read one parameter file, use one log parser, and write the same JSON result into ~/cosmos-runs/ocudu-ota/<UTC>-<mode>/ on the console. OMF_NODES is required and ordered: first the gNB server, then the UE node. The N310 is not in OMF_NODES; it is named in the parameters because it is a shared radio that the tutorial power-cycles from the console.
The playbook runs a preflight check, power-cycles the N310 from the console and waits until it can be claimed, starts Open5GS, provisions the modem's SIM with an IPv4-only session, renders the gNB configuration from the template, and starts the gNB and an iperf3 server as named systemd units (ocudu-gnb, ocudu-iperf). It then checks the gNB host's real-time health, configures and registers the modem, brings up the data call, measures ping and throughput, and writes the result.
The manual mode runs the commands below for you, from the console. To follow them yourself, start on the console with the tutorial's directory and the radio's address:
D=$(cosmos-tutorial which ocudu-ota)
USRP=$(python3 $D/cosmos_addr.py --value auto usrp sdr1-s1-lg1 --plane data2) # the N310's address, from the inventory
Step 1 — Power-cycle the radio (console). A freshly booted N310 is unclaimed and starts clean. Turn it off, wait until it stops answering, and turn it on again. Then wait, on the gNB server, until UHD finds it and can claim it.
omf tell -a offh -t sdr1-s1-lg1
omf tell -a on -t sdr1-s1-lg1
ssh root@<gnb-node> "uhd_find_devices --args addr=$USRP"
ssh root@<gnb-node> "uhd_usrp_probe --args addr=$USRP"
uhd_find_devices must list type: n3xx and claimed: False. claimed: True means another process, possibly another user's, holds the radio: do not start the gNB against it.
Step 2 — Start the core and provision the modem's SIM (gNB server). The core must be up before the gNB, because the gNB sends its NG Setup Request only once, at start-up. The launcher starts the network functions, creates ogstun with NAT and writes every COSMOS SIM into the core. The tutorial's script then makes sure the modem's SIM has the internet DNN and forces that DNN's session to IPv4 (type 1), leaving the SIM's keys and other DNNs alone.
scp $D/ocudu-ota_sub.js root@<gnb-node>:/root/
ssh root@<gnb-node>
/root/open5gs/start.sh
mongosh open5gs --quiet --eval 'var IMSI="001010000000032", DNN="internet"' /root/ocudu-ota_sub.js
ss -lnp -A sctp | grep 38412 # AMF listening for the gNB
Step 3 — Render the gNB configuration (console). The tutorial fills ocudu-ota_gnb.yml.tmpl from the parameters: N310 device arguments with the 122.88 MHz master clock and a frame size of 7900, 61.44 Msps, band n41, 40 MHz, PLMN 00101, TAC 7, the gains and the log file. Do not use the image's own /root/ocudu/gnb-ocudu-40.yml: it is an X310 file with another PLMN and TAC 1, and the gNB exits with CU-CP failed to connect to AMF.
python3 $D/ocudu-ota_lib.py render-gnb $D/ocudu-ota_params.yml $D/ocudu-ota_gnb.yml.tmpl gnb-ocudu-ota.yml --usrp-addr $USRP
scp gnb-ocudu-ota.yml root@<gnb-node>:/root/ocudu/gnb-ocudu-ota.yml
Step 4 — Start the gNB (gNB server). The gNB connects to the AMF, opens the radio and puts the cell on the air. It runs as a named systemd unit so the teardown can stop exactly this process. Keep its standard input open with sleep infinity |: the gNB exits when standard input closes. Its own log, /tmp/ocudu-gnb.log, carries gNB started, NG Setup and every RF failure; the console output goes to a separate file.
systemd-run --unit=ocudu-gnb --collect --working-directory=/opt/ocudu/build bash -c 'sleep infinity | /opt/ocudu/build/apps/gnb/gnb -c /root/ocudu/gnb-ocudu-ota.yml > /tmp/ocudu-gnb-stdout.log 2>&1'
grep -aE 'NG Setup|gNB started|Cell was activated' /tmp/ocudu-gnb.log
ss -np -A sctp | grep -E 'ESTAB.*38412'
Step 5 — Check the gNB host's real-time health (gNB server). The fix-up script moves the radio NIC's interrupts onto that NIC's own NUMA node, then measures the rate of Real-time failure in RF lines over 10 s. On an N310 a start is either clean or floods at thousands of failures per second and never recovers; on a flood the script releases the radio and starts the gNB again, up to five times. It refuses to run on a host with isolcpus. Then start the iperf3 server on the core's gateway address.
scp $D/ocudu-ota_gnbfix.sh root@<gnb-node>:/root/ # from the console
/root/ocudu-ota_gnbfix.sh $USRP /tmp/ocudu-gnb.log /opt/ocudu/build/apps/gnb/gnb /root/ocudu/gnb-ocudu-ota.yml /tmp/ocudu-gnb-stdout.log
systemd-run --unit=ocudu-iperf --collect /usr/bin/iperf3 -s -B 192.168.100.1 -p 5201
Step 6 — Configure and register the modem (UE node). ModemManager is stopped so it does not hold the AT port. The helper finds the Quectel's own QMI device and network interface (a UE node can carry two modems), then sets SA-only mode, the band mask 41, the APN internet (IPv4) and the PLMN lock 00101, and waits for registration. The individual AT commands are explained on The commercial UE.
scp $D/ocudu-ota_modem.py $D/ocudu-ota_iperf.sh root@<ue-node>:/root/ # from the console
ssh root@<ue-node>
systemctl stop ModemManager
read -r QMI WWAN <<<"$(python3 /root/ocudu-ota_modem.py resolve auto auto)"; echo "$QMI $WWAN"
python3 /root/ocudu-ota_modem.py configure /dev/ttyUSB2 00101 41 internet
python3 /root/ocudu-ota_modem.py wait-reg /dev/ttyUSB2 240 00101
Step 7 — Bring up data and measure (UE node). Switch the interface to raw-IP mode, start an IPv4 data call on the internet APN, give the interface its address, and add routes scoped to the UE pool and the remote test target. Never add a default route through the modem: it would cut off the node's management SSH.
ip link set $WWAN down; echo Y > /sys/class/net/$WWAN/qmi/raw_ip; ip link set $WWAN up
qmicli -p -d $QMI --wds-start-network="apn='internet',ip-type=4" --client-no-release-cid
qmicli -p -d $QMI --wds-get-current-settings # note the IPv4 address as $UE
ip addr add $UE/22 dev $WWAN; ip link set $WWAN mtu 1400 up
ip route replace 192.168.100.0/22 dev $WWAN
ip route replace 8.8.8.8/32 via 192.168.100.1 dev $WWAN
ping -I $WWAN -c 10 -i 0.3 -W 2 192.168.100.1 # core gateway
ping -I $WWAN -c 10 -i 0.3 -W 2 8.8.8.8 # the internet
/root/ocudu-ota_iperf.sh 192.168.100.1 5201 $UE 10 -R # DL
/root/ocudu-ota_iperf.sh 192.168.100.1 5201 $UE 10 # UL
ocudu-ota_iperf.sh retries an over-the-air TCP test once if it stalls, and always prints the receiver line, which is the measurement.
The notebook runs on the console and drives both nodes over SSH. It performs the same steps and plots downlink and uplink throughput. It can also be run non-interactively with papermill, which writes an executed copy to a new file.
The parity mode runs Ansible, manual and Jupyter one after the other on the same two nodes, the same radio and the same modem, then compares their results. It passes when every mode puts the cell on the air and registers the modem, local ping loss is at most 20 %, the RF real-time failures stay at most 200, downlink is in 20–400 and uplink in 2–120 Mbit/s, and the modes agree within 60 % (downlink) and 80 % (uplink). The harness runs the teardown on every exit, including a failure or Ctrl-C.
The same cell also serves Amarisoft's lteue software UE driving a second USRP from a UE node loaded with lte5gue.ndz. This path is not part of the package and is run by hand. lteue needs jumbo frames towards its USRP and does not attach on its own: the power_on command over its remote API starts the attach.
cd /opt/amarisoft/lteue-linux-2026-06-12
ip link set DATA2 mtu 9000
setsid bash -c "tail -f /dev/null | ./lteue-avx2 config/ue-ocudu-00101.cfg" >/tmp/ue_run.log 2>&1 &
python3 amari_cmd.py 9002 '{"message":"power_on","ue_id":1}'
ip netns exec ue1 ip -br addr show pdn0
Its subscriber needs the same IPv4-only session. Amarisoft licences are one seat per tag on the whole testbed: stop lteue when you finish.
After Step 2, the AMF is listening for the gNB:
LISTEN 0 0 127.0.0.5:38412 0.0.0.0:* users:(("open5gs-amfd",...))
After Step 4, the gNB has registered with the core and the cell is on the air:
[NGAP ] [I] "NG Setup Procedure" finished successfully
[GNB ] [I] ==== gNB started ===
ESTAB 0 0 127.0.0.5:38412 127.0.0.1:... users:(("open5gs-amfd",...))
After Step 5, the fix-up reports a clean start. A short burst of RF failures in the first fraction of a second after Cell was activated is normal and is not counted.
After Steps 6 and 7, the modem is registered on the home network, holds an IPv4 address from the pool, and both pings succeed:
+C5GREG: 0,1
IPv4 address: 192.168.100.2
10 packets transmitted, 10 received, 0% packet loss
In +C5GREG: 0,1 the second field 1 means registered on the home network; 2 means still searching. The last verified parity run (2026-10-30, sb1) measured a downlink of 103–113 Mbit/s, an uplink of 48–50 Mbit/s and a ping of 26–32 ms. A parity run ends with PARITY: PASS and the two spread lines inside their tolerances, followed by rig released.
Change any of these with cosmos-tutorial run ocudu-ota --params my.yml. An unknown key or a wrong type is refused. cosmos-tutorial params ocudu-ota prints the file.
| Key | Default | Meaning |
|---|---|---|
ocudu_sdr_node |
sdr1-s1-lg1 |
|
ocudu_usrp_node |
sdr1-s1-lg1 |
The radio's inventory node: its USRP address is looked up in the COSMOS inventory at run time. |
ocudu_usrp_plane |
data2 |
The radio's network plane the gNB node streams on (data2 = the N310's data address). |
ocudu_usrp_addr |
auto |
auto = from the COSMOS inventory (the radio node's ocudu_usrp_plane address); an address pins it. |
ocudu_usrp_iface |
auto |
auto = an E810 port that reaches the radio, else the kernel's pick; a name pins the port (ocudu-ota_radiolink.sh). |
ocudu_radio_cycle |
True |
|
ocudu_gnb_bin |
/opt/ocudu/build/apps/gnb/gnb |
|
ocudu_gnb_cfg |
gnb-ocudu-ota.yml |
RENDERED from ocudu-ota_gnb.yml.tmpl each run; relative = in the run's node dir /var/tmp/cosmos/ (an absolute path is used as is) |
ocudu_device_args |
type=n3xx,addr=@USRP_ADDR@,master_clock_rate=122.88e6,recv_frame_size=7900,send_frame_size=7900 |
N3xx device args: 122.88 MHz master clock, frame size 7900 (8000 silently kills UL on N3xx); srate 61.44 = 122.88/2 for 40 MHz. |
ocudu_srate |
61.44 |
|
ocudu_tx_gain |
40 |
TX gain: 25 left the modem at RSRP -106 / SINR 3 and a link that died mid-iperf (2026-10-23, DL 44 / UL 59). 50 gave -79 / 30 and DL 110, but OVERDRIVES the nearby modem: ~10 % of UL grants never arrived (DTX) and UL fell to 30-54. 40 is the sweet spot: -89 / 20, DL 111 / UL 60, no DTX. |
ocudu_rx_gain |
40 |
|
ocudu_ssb_block_power_dbm |
-1 |
Advertised SS/PBCH block power. |
ocudu_pci |
1 |
|
ocudu_dl_arfcn |
530910 |
DL carrier centre (NR-ARFCN): 2654.55 MHz. |
ocudu_bw_mhz |
40 |
|
ocudu_core_start |
/root/open5gs/start.sh |
NFs + ogstun 192.168.100.1/22 + NAT + subscriber_db.csv |
ocudu_gnb_log |
ocudu-gnb.log |
srsRAN log.filename (rendered): "gNB started" + RF failures; relative = in the run's node dir |
ocudu_gnb_stdout |
ocudu-gnb-stdout.log |
the gNB's stdout/stderr (banner, UHD lines, fatal errors); relative = in the run's node dir |
ocudu_band |
41 |
n41, 40 MHz (106 PRB @ 30 kHz). |
ocudu_plmn |
00101 |
|
ocudu_tac |
7 |
TAC MUST match the AMF's tai list or NG Setup is rejected with "unknown-PLMN-or-SNPN" and the gNB exits ("CU-CP failed to connect to AMF"). |
ocudu_imsi |
auto |
auto = read from the modem's SIM (AT+CIMI); a number here is only a cross-check |
ocudu_dnn |
internet |
|
ocudu_ue_pool |
192.168.100.0/22 |
|
ocudu_ue_gw |
192.168.100.1 |
ogstun on the gNB node = the LOCAL data-plane target |
ocudu_at_port |
/dev/ttyUSB2 |
|
ocudu_qmi_dev |
auto |
auto = the Quectel's own QMI device and netdev, found from /dev/quectel-qmi or /dev/quectel-at (a UE node can carry two modems: node1.rrail has a SIMCom on cdc-wdm0/wwan0 and the Quectel on cdc-wdm1/wwan1). |
ocudu_wwan |
auto |
|
ocudu_nr5g_bands |
41 |
the cell's band ONLY: with "41:77:78" the modem camped on a stronger nearby n78 |
ocudu_remote_target |
8.8.8.8 |
|
ocudu_ping_count |
10 |
|
ocudu_iperf_secs |
10 |
|
ocudu_iperf_port |
5201 |
|
ocudu_radio_timeout_s |
240 |
|
ocudu_gnb_timeout_s |
90 |
"gNB started" / "Cell was activated" + NGAP SCTP ESTAB |
ocudu_reg_timeout_s |
240 |
To change a parameter, put only the keys you change in a file of your own and pass it to any mode. A key the tutorial does not have, or a value of the wrong type, is refused before anything starts:
cat > my.yml <<'EOF'
ocudu_tx_gain: 35
EOF
cosmos-tutorial run ocudu-ota --params my.yml
ocudu_tx_gain and ocudu_ssb_block_power_dbm belong together. The modem estimates its path loss as the advertised SSB power minus the RSRP it measures, and sets its uplink power from that, so change both by the same amount. To change the tutorial itself, cosmos-tutorial copy ocudu-ota puts an editable copy in the current directory, which then takes precedence over the installed one.
gNB started say the core accepted the gNB; +C5GREG: 0,1 says the modem completed random access, the RRC connection, registration and authentication; the IPv4 address says the PDU session was established. The flow is on 5G SA fundamentals.| Symptom | Cause | Fix |
|---|---|---|
| Uplink far below the expected value and changing between runs; one uplink slot per frame fails while the downlink is fine | an outside n41 signal at about 2560–2610 MHz in sb1 overlaps the cell's uplink slots when the carrier sits near it |
keep the shipped ocudu_dl_arfcn: 530910 (2654.55 MHz); a carrier you choose yourself must stay clear of that block |
gNB: NG Setup … unknown-PLMN-or-SNPN, then CU-CP failed to connect to AMF |
the gNB's PLMN or TAC differs from the AMF's (00101, TAC 7); the image's gnb-ocudu-40.yml has another PLMN and TAC 1 |
use the rendered gnb-ocudu-ota.yml (Step 3); restart the gNB, because NG Setup is sent only once |
Registration succeeds but there is no PDU session; the gNB logs Unsupported PDU Session Type: ipv4v6 |
the modem asked for a dual-stack session and the gNB accepts IPv4 only | run ocudu-ota_sub.js for the SIM and DNN (Step 2), then restart the modem's registration |
[RF] Real-time failure in RF: underflow; the modem reports CQI 1–2 or drops mid-registration |
the radio NIC's interrupts land on the cores running the physical layer, or the start flooded | ocudu-ota_gnbfix.sh (Step 5) re-pins the interrupts and restarts a flooded start; systemctl is-enabled ocudu-irq-affinity must say enabled |
The fix-up stops with this host has isolcpus=… |
the gNB server still carries CPU isolation from an O-RAN 7.2 run | re-load ocudu.ndz, or remove the isolation as in Removing isolation and reboot |
| One mode's downlink test runs long at a low rate and its uplink test cannot connect | a real-time slip on the gNB host (Received late DL_TTI.request, Downlink data late in the gNB log), not the modem |
re-run the parity; keep other load off the gNB server |
uhd_find_devices shows claimed: True |
another process holds the N310 | find out who before you start; never take the radio from another user |
The gNB aborts with free_buff_pool::pop … Assertion |
the radio's streaming is wedged, for example after another experiment switched it | power-cycle the radio from the console (Step 1), then one clean start |
The modem detects the cell but never camps (SEARCH) |
nr5g_disable_mode=1 disables SA on this modem |
set it to 0; lock AT+COPS=1,2,"00101",12 (the helper does both) |
| The modem never sees the cell, or camps on another cell | its band mask was left at n78 by a 7.2 run, or allows n78 next to n41 | band mask 41 only (ocudu_nr5g_bands); the helper sets it, reboots the modem and checks it stuck |
The PLMN lock fails with +CME ERROR: 13 |
the lock was sent before the SIM reported +CPIN: READY |
wait for +CPIN: READY; ocudu-ota_modem.py configure does |
Registered, PDU up, an address on wwan, and every ping lost; ip -s link show $WWAN shows TX errors rising by one per packet |
the modem's USB data endpoint is stuck | USB device reset of the modem: python3 /root/ocudu-ota_modem.py usbheal (configure does it when it sees TX errors); see Detecting a dead USB data path |
| The node's management SSH is lost after the data call | a default route via wwan |
use only the scoped routes of Step 7 |
The gateway answers but 8.8.8.8 does not |
the core's NAT rule is missing | re-run /root/open5gs/start.sh, which adds MASQUERADE for the pool |
A manual AT session right after a run reports device reports readiness to read but returned no data |
the teardown hands the modem back to ModemManager, which holds the AT ports | systemctl stop ModemManager first |
Preflight: OMF_NODES selected 1 node(s) … but this tutorial needs exactly 2 |
a node is Unknown (UNREACHABLE) in omf stat and does not answer SSH |
check that it is on and booted, power-cycle it with omf tell, and report the chassis manager to the testbed operators |
lteue aborts with Someone tried to claim this device again |
the previous lteue did not release its USRP yet |
stop it with TERM, not KILL, and wait about 20 s before relaunching |
The modem rows are explained in more depth on The commercial UE.
After any mode, run the teardown from the console. On the UE node it stops the data call, clears the routes and hands the modem back to ModemManager; on the gNB server it stops the ocudu-gnb and ocudu-iperf units and the Open5GS network functions. The radio is left powered for other users.
OMF_NODES=<gnb-node>,<ue-node> cosmos-tutorial run ocudu-ota --mode teardown
After the manual steps, systemctl stop ocudu-gnb ocudu-iperf and /root/open5gs/stop.sh on the gNB server and qmicli -p -d $QMI --wds-stop-network=disable-autoconnect on the UE node do the same. Never stop the gNB with pkill: pkill -f apps/gnb/gnb matches your own shell's command line, and a name match would also stop another user's gNB. Then power both nodes off from the console:
omf tell -a offh -t <gnb-node>,<ue-node>
Package cosmos-experiments-ocudu 0.1.0-31 - on every console it is released to (cosmos-tutorial list shows the installed version). The code is installed on the console in /usr/share/cosmos-experiments/ocudu-ota/. To change anything, cosmos-tutorial copy ocudu-ota puts an editable copy in the current directory, and cosmos-tutorial run ocudu-ota then runs that copy instead of the installed one (cosmos-tutorial which ocudu-ota shows which will run). Repository (GitLab account needed, not needed to run or change it): packages/cosmos-experiments-ocudu/usr/share/cosmos-experiments/ocudu-ota.
Last verified 2026-11-19: ansible mode PASS from the 0.1.0-31 code, on sb1 (2× Intel Xeon Gold 6226, 24C/24T, N310 in n41 40 MHz, Quectel modem): DL 88.9 / UL 39.1 Mbit/s, ping 44 ms. Before that, 2026-10-30: three-mode parity PASS from the installed cosmos-experiments-ocudu package, cold start (all devices off), on sb1 (2× Intel Xeon Gold 6226, N310 and Quectel modem): DL 103–113 / UL 48–50 Mbit/s, ping 26–32 ms. Results by build are on OCUDU performance by build.
References: srsRAN Project documentation (ru_sdr and cell_cfg) · Open5GS documentation · UHD manual (device arguments, master clock rates, frame sizes) · Quectel RM520N series · Amarisoft Tech Academy (lteue) · 3GPP TS 23.501 (PDU session types) and TS 38.104 (band n41) · iperf3 · on this wiki: OCUDU on COSMOS · The commercial UE · Preparing a real-time RAN host · SDR hardware · Open5GS.
Author(s): COSMOS team · Package: cosmos-experiments-ocudu (cosmos-tutorial run ocudu-ota) · Tags: ocudu, srsran, open5gs, 5g, sa, sdr, usrp, n310, quectel, ota