A commissioning runbook for connecting a rack PDU to a DCIM platform: the addressing decision, the SNMP read path, the Modbus command path and the handover.
Written by: Newsunn, Senior PDU Product Engineer Last verified: 13 November 2026 Reading time: ~11 min
TL;DR
- The 30 minutes is not the protocol. It is the addressing. Our published controller exposes HTTP, SNMP v1 and v2, Modbus TCP/IP, Modbus RTU over RS-485, FTP and IPv4, with an LCD showing the amperage, voltage, kW, IP address, baud rate and device ID.
- SNMP is the read path and Modbus is the write path, and the two need different addressing decisions. SNMP pulls meters and traps over UDP; Modbus TCP carries commands over TCP to a fixed unit identifier.
- Give the PDU a static address before you give it a protocol. A PDU on the wrong subnet is the most common reason a commissioning session runs over an hour.
- Both protocols need the device addressable and the credentials to match. Our published configuration is done through WEB and SNMP, and the community string has to match the DCIM.
- Write down the numbers at the end. IP address, subnet mask, gateway, SNMP community and Modbus unit ID. Without them the next firmware upgrade is a truck roll.
Thirty minutes is realistic when the PDU arrives with an address plan, and impossible when it arrives as a black box on a bench. Our controller speaks HTTP, SNMP v1 and v2, Modbus TCP/IP and Modbus RTU over RS-485, and the runbook below runs the way the work actually runs: address first, read second, command third, document last. The delays come from three places, and none is the protocol.
What has to be true before the clock starts
Three conditions decide whether this is a half-hour job or a half-day one, and all three are settled before anyone opens a laptop.
The PDU has a reachable address. Our published controller carries Ethernet ports for remote management over a web interface or a command line, and the LCD shows the IP address, the baud rate and the device ID. That display is why this job is fast: an installer at the rack reads the address rather than inferring it. If the PDU is on a bench at the other end of the building, add the walk.
The DCIM side already knows how it will talk to it. Whether the platform polls over SNMP, writes over Modbus, or uses the PDU’s own web interface for a commissioning pass determines which path you spend time on. Both are published on our OEM network PDU product page: users check the function configuration parameters and the power control of the remote device via WEB and SNMP, and firmware is upgraded over network download rather than by replacing installed units.
The PDU’s functions are configured for what the DCIM needs. Our intelligent PDU range splits into four models, and the choice changes what your platform can read and command. Type A gives total metering, total switching, individual outlet metering and individual outlet switching. Type B gives total metering with total switching. Type C adds individual outlet metering. Type D is total metering only. If the DCIM was promised per-outlet data and the installed unit is a Type B, no amount of protocol work will produce it.
That last point surfaces after the platform is configured. Check the model letter on the specification sheet before you check the cable.
Step one: give the PDU an address the network can reach
This step decides the schedule. Our published configuration is done through WEB and SNMP, and the LCD shows the amperage, voltage, kW, IP address, baud rate and device ID, so the address is readable on the face of the PDU without a console cable.
Set a static address inside the subnet the DCIM server actually lives on. A PDU that takes its address by DHCP and lands in a different range than the platform is the classic thirty-minute session that becomes an afternoon. Two rules make this deterministic. Reserve the address in the DHCP server so a firmware restart cannot move it. And confirm the gateway is the one the DCIM server uses to reach the monitoring VLAN, not the office one.
Then confirm the two things that make SNMP and Modbus behave differently on one cable.
SNMP runs over UDP and identifies the device by its IP address, as RFC 1157 defines. Modbus TCP runs over TCP and identifies the device by a unit identifier, the device ID on the LCD. The same PDU therefore has two independent identities, and both have to go into the platform configuration. A Modbus request that reaches the right host with the wrong unit ID returns no useful error, which is why the unit ID belongs in the handover note.
For a serial chain rather than an Ethernet drop, the split appears differently. Our published interface list includes an RS-485 port carrying Modbus RTU, and the baud rate is one of the LCD parameters. On a bus, the unit ID and baud rate must agree across every device on the segment.
Do the addressing on the switch first, then the PDU. Confirm the port is up, the VLAN is right and the address is free before configuring the unit. A PDU that cannot reach the platform looks exactly like a PDU with a bad protocol configuration, and you will spend twenty minutes in the wrong place.
Step two: read the meter over SNMP
The read path proves the integration works, and it is where SNMP earns its place. Our controller publishes SNMP v1 and v2 alongside HTTP, Modbus TCP/IP, Modbus RTU, FTP and IPv4, and the published function set tells you which objects a platform finds.
Every model reports the same core set: total load current, total power in kW, total energy consumption in kWh, working voltage, frequency, and temperature with humidity. Models A and C add the load current and on/off state of each outlet. Models A and B add the smog, door and water logging sensors. Type A alone adds the control side: per-outlet switching, sequencing, scheduling and limits.
That list answers “what should the DCIM show me”, and it is worth settling before the first poll rather than after the first empty panel. A platform that polls successfully and receives nothing is not a protocol fault; it is a model mismatch.
Two settings have to match on both sides. The community string must be identical in the PDU and the platform, and changed from the factory default on any PDU facing an untrusted network; the community concept comes from RFC 3411 and the address scheme from RFC 1912. The polling interval must sit inside what the PDU can serve, because a platform polling every five seconds across a full row saturates the management path first. The interval is your decision.
SNMP traps get skipped and then regretted, and the mechanism is the one RFC 2576 defines for SNMPv2. Our published features list immediate updates about power events via email, SMS text or SNMP traps, and the alarm set covers total load current over its limit, an outlet over its limit, temperature and humidity over their limits, smoke, water logging and door opening. Traps tell the platform of an event it would otherwise notice on the next poll.
Step three: command the PDU over Modbus
Reading proves the link. Command makes it worth having, and it is where Modbus earns its place.
Our published interface list carries both Modbus TCP/IP and Modbus RTU over RS-485 alongside HTTP and SNMP. Modbus TCP fits when the DCIM server and the PDU are on a routable network and the platform must read a register or write a coil. Modbus RTU over RS-485 fits a serial bus, where a row of units is daisy-chained behind one drop.
The command set depends on the model, and the type letter decides what you can do. On a Type A, the published control functions are on and off of the power, on and off of each outlet, setting the interval time of sequential outlet switching, setting the on and off time of each outlet, and setting limiting values to alarm. On a Type B, only total on and off. On Type C and Type D there is no switching at all, because both are monitoring-only models. A platform that can poll a Type C beautifully and cannot switch anything is working as the product is designed.
There is also a published limit condition for the commissioning notes. The limiting range of total load current applies to every model, the limiting range of load current of each outlet exists on models A and C only, and the limiting ranges of working voltage and of temperature and humidity apply on models A and B.
Set the limits before you hand over. The published alarm set fires on total load current over its limit, an individual outlet over its limit, and temperature or humidity over its limit, and those alarms are how a facilities team acts on the data collected. A default threshold will not fire when needed.

The last 15 minutes: what to write down
The session is not finished when the panel populates. It is finished when somebody else can repeat the integration without asking you a question, and that takes six numbers in one place.
The IP address. Not the one you planned, the one the unit is on. The subnet mask and gateway, because a PDU on the right address with the wrong gateway is reachable from the bench and not from the platform. The SNMP community string, written where the platform configuration lives, not beside the IP address. The Modbus unit identifier, because on an RS-485 chain it must agree with every other device on the segment. And the firmware version, since the upgrade path is a network download and the version you leave behind is the one you will be asked about later.
Two more items belong in the note. Which physical model is installed, by its type letter, because the Type A and Type C distinction decides whether per-outlet data appears at all. And which limits are set, because a threshold changed during commissioning and never written down is one nobody can defend.
Our published quality regime is worth mirroring for the acceptance test: a hi-pot test at 2000V, a ground and insulation resistance test, a 48-hour online aging test for zero failure, a load test at 120% of rating, then a function test on the outlets, breakers, switches and remote control. Sign off the same way: every published function should be exercised once.
Thirty minutes is the happy path, and it assumes the PDU arrives with a label carrying the IP address, the unit ID and the firmware version. That label is a factory setting we can provide, and it is the difference between a commissioning session and a commissioning archaeology project.
Send your platform and we will confirm the model and the protocol path
Send us the platform and we will confirm the model, the protocol path and the commissioning sheet before the units ship. Our intelligent power distribution solution page lists the OEM and customization patterns behind it: IP management vertical PDUs, 19-inch 1U intelligent PDUs with six C13 outlets, and 3-phase builds with per-outlet metering. Tell us the protocol, outlet count, phases and power envelope and we will map it onto the published range, with the network settings printed and the checklist for your installer.
Frequently asked questions
How long does it really take to connect a PDU to a DCIM platform?
Thirty minutes is realistic when the PDU arrives with an address plan and somebody has confirmed which protocol path the platform will use. The work is addressing, then the SNMP read path, then the Modbus command path. Delays come from a PDU on the wrong subnet.
Should I use SNMP or Modbus for DCIM integration?
Use SNMP for the read path and Modbus for the command path. Our controller publishes SNMP v1 and v2 for polling and traps, and Modbus TCP/IP or Modbus RTU over RS-485 for register reads and writes. Modbus TCP fits a routable network; Modbus RTU a serial bus where unit ID and baud rate agree.
What is the difference between the PDU IP address and the Modbus unit ID?
They identify the same device two ways. SNMP runs over UDP and reaches the PDU by IP address; Modbus TCP identifies it by a unit identifier shown on the LCD as the device ID. Both go into the platform configuration, and a Modbus request with the wrong unit ID returns no useful error.
My DCIM polls the PDU successfully but shows no per-outlet data. What is wrong?
Almost certainly the model. Per-outlet load current and on/off state exist on Type A and Type C only; Type B is total metering with total switching, Type D metering only.
Do I need SNMP traps if the platform already polls?
Yes, for events you do not want to wait a poll interval to learn about. Our published features list immediate updates about power events via email, SMS text or SNMP traps, and the alarm set covers load current, outlet current, temperature, humidity, smoke, water logging and door opening. Configure the trap destination with the poll.
Can a PDU firmware update be done after it is installed in the rack?
Yes. Our published features list downloadable firmware upgrades over the network rather than replacing products already in the field, and the protocol support includes FTP and IPv4 alongside HTTP, SNMP and Modbus. Record the firmware version at handover.
Which thresholds should be configured before handover?
The limiting range of total load current applies to every model. The limiting range of load current of each outlet exists on Type A and Type C only. The limiting ranges of working voltage and of temperature and humidity apply on Type A and Type B. A default threshold will not fire when needed, so set and record them.
About the author
Newsunn PDU Engineering Team
Senior PDU product engineering team at Newsunn, a designer and manufacturer of rack power distribution units for data centers, server rooms and mission-critical facilities, with more than ten years of PDU design and manufacturing experience, eight R&D engineers, eight PDU design patents and a 30,000 square metre production base in Cidong Industrial Zone, Cixi City near Ningbo port, with more than 200 staff across injection moulding, painting, aluminium machining, assembly, test and packing.
The team supplies standard rack units and fully customized intelligent power distribution systems, including switched and metered OEM programs, with outlets and inputs configured per project.
About the company · Who we are · Contact us
Post time: Nov-13-2026
