Skip to content

Portal | Level: L1: Foundations | Topics: Server Hardware, Out-of-Band Management, RAID, Firmware / BIOS / UEFI | Domain: Datacenter & Hardware

Dell PowerEdge Servers - Primer

Comprehensive Reference

For the full operational guide with detailed iDRAC configuration, Redfish API deep-dive, and automation scripts, see Dell Server Management Guide.

Why This Matters

Dell PowerEdge is the most common server platform in enterprise datacenters. If you work in on-prem infrastructure, colo, or hybrid cloud, you will encounter PowerEdge hardware. Knowing the product line, management stack, and storage controllers lets you provision, troubleshoot, and maintain servers without fumbling through vendor docs every time.

The PowerEdge Product Line

Name origin: The "PowerEdge" brand was introduced by Dell in 1996 for their server line. The name suggests "power at the edge" — servers pushing computing capabilities to new frontiers. Dell's server naming scheme is remarkably systematic compared to competitors, and once you decode the pattern you can identify any model at a glance.

Dell names rack servers with an R prefix, a three-digit model number, and an optional suffix. The pattern tells you generation and form factor at a glance.

Naming convention

R <series> <generation> <variant>
│    │          │          │
│    │          │          └─ 0 = base, 5 = high-density/NVMe, d = storage-dense
│    │          └─ generation digit (2=12G, 3=13G, 4=14G, 5=15G, 6=16G)
│    └─ series: 6xx=1U, 7xx=2U
└─ R = rack mount

Generations you will see in production

Generation Era Example Models CPU Platform iDRAC Version
12G 2012-2015 R620, R720, R720xd Intel Sandy/Ivy Bridge iDRAC 7
13G 2015-2017 R630, R730, R730xd Intel Haswell/Broadwell iDRAC 8
14G 2017-2020 R640, R740, R740xd Intel Skylake/Cascade Lake iDRAC 9
15G 2020-2023 R650, R750, R750xa Intel Ice Lake/Sapphire Rapids iDRAC 9
16G 2023+ R660, R760, R760xa Intel Emerald Rapids/Sapphire iDRAC 9

What you will actually see: Most datacenters run a mix. 14G (R640/R740) is still the workhorse in many fleets. 12G/13G hardware is being decommissioned but you will encounter it. 15G/16G is current procurement.

Common models and their roles

Model Form Typical Role
R640 / R650 / R660 1U Compute-dense: hypervisors, Kubernetes nodes, web/app tier
R740 / R750 / R760 2U General purpose: databases, mixed workloads, moderate storage
R740xd / R750xa / R760xa 2U Storage-dense: Ceph OSDs, object stores, large local datasets
R6525 / R7525 1U/2U AMD EPYC variants (same roles, different CPU)

Rule of thumb: 1U for compute, 2U for anything needing more drives, GPUs, or PCIe slots.

The Management Stack: iDRAC

Every PowerEdge has an iDRAC (Integrated Dell Remote Access Controller) — a dedicated BMC with its own NIC, IP, and web interface that works even when the server OS is dead.

Timeline: iDRAC has evolved across generations: iDRAC 6 (11G, basic web UI) → iDRAC 7 (12G, HTML5 console) → iDRAC 8 (13G, improved API) → iDRAC 9 (14G+, full Redfish API, telemetry streaming). iDRAC 9 is the current standard and supports Redfish, RACADM, IPMI, SNMP, and a modern HTML5 web console. Each generation added significant automation capabilities — iDRAC 9 can be fully managed through REST APIs without ever touching the web UI.

What iDRAC gives you

  • Remote console — full KVM access over the network (no physical monitor needed)
  • Power control — power on, graceful shutdown, hard reset, NMI
  • Virtual media — mount an ISO remotely to install an OS
  • Hardware monitoring — temps, fan speeds, PSU status, memory errors
  • Lifecycle logs — hardware event history (disk failures, thermal events, firmware changes)
  • Firmware updates — update BIOS, NIC, PERC, iDRAC itself without booting the OS
  • Inventory — CPUs, DIMMs, disks, NICs, PCIe cards with serial numbers

Accessing iDRAC

Method When to use
Web UI (https://<idrac-ip>) Interactive troubleshooting, one-off tasks
RACADM CLI (racadm) Scripted config, batch operations, SSH-based
Redfish API (https://<idrac-ip>/redfish/v1/) Automation, Ansible, fleet management
IPMI (ipmitool) Legacy, cross-vendor basics (power, sensors)

iDRAC networking

iDRAC has its own dedicated NIC port (or can share a host NIC in shared/LOM mode). Best practice: always use the dedicated port on a separate management VLAN. Shared mode is fragile — if the host NIC goes down, you lose OOB access.

Storage: PERC Controllers

PERC (PowerEdge RAID Controller) is Dell's branded RAID controller. It sits between the disks and the OS.

Under the hood: PERC controllers are rebranded Broadcom (formerly Avago/LSI) MegaRAID controllers. The PERC H730 is an LSI MegaRAID SAS 3108, the PERC H745 is based on the SAS 3516. This means storcli (the LSI/Broadcom generic tool) works alongside Dell's perccli — they are the same binary with different branding. If you encounter a non-Dell server with an LSI MegaRAID, your PERC knowledge transfers directly.

Common PERC models

Controller Generation Key Feature
PERC H710/H710P 12G 6Gbps SAS, battery-backed cache
PERC H730/H730P 13G-14G 12Gbps SAS, supercap-backed cache
PERC H740P 14G 12Gbps SAS, 8GB cache
PERC H745 15G 12Gbps SAS, 8GB cache
PERC H755/H965i 16G NVMe awareness, larger cache

PERC vs HBA mode

  • PERC (RAID mode) — controller manages RAID arrays. OS sees virtual disks. Standard for most workloads.
  • HBA mode (passthrough) — disks passed directly to OS. Required for software-defined storage (Ceph, ZFS). No hardware RAID.

CLI tool: perccli

# Show all controllers
perccli /call show

# Show virtual disks on controller 0
perccli /c0/vall show

# Show physical disks
perccli /c0/eall/sall show

# Check for failed/degraded disks
perccli /c0/eall/sall show | grep -E "Onln|Failed|Rbld"

Typical RAID layouts

Use Case RAID Level Drives Why
OS boot volume RAID 1 2x SSD Mirror for redundancy, small and fast
General data RAID 5 or RAID 6 4-8 disks Balance of capacity and protection
Database RAID 10 4+ disks Write performance with redundancy
Ceph OSD HBA passthrough All disks Ceph manages its own redundancy

Networking

PowerEdge servers ship with 2 or 4 onboard NICs (LOM — LAN on Motherboard), typically 1GbE or 10GbE depending on generation. Additional NICs go in PCIe slots.

Common NIC configurations

Config Typical Use
2x 1GbE onboard Management + basic traffic (older/small)
2x 10GbE onboard (NDC) Production traffic, bonded
2x 25GbE add-in (Mellanox/Broadcom) High-throughput: storage networks, K8s CNI
Dedicated iDRAC port Always for OOB management

Bonding

Production servers almost always bond (aggregate) NICs for redundancy and throughput: - active-backup (mode 1) — one active, one standby. Simple, works with any switch. - LACP (mode 4 / 802.3ad) — both active, requires switch support. Higher throughput.

Physical Layout: What You See in the Rack

A 2U PowerEdge (like an R740) from front to back:

┌─────────────────────────────────────────────────────┐
│  FRONT: drive bays (8-24 slots), power button,      │
│         status LEDs, service tag pull-tab, iDRAC     │
│         direct USB port                              │
├─────────────────────────────────────────────────────┤
│  MIDDLE: CPUs + DIMMs + PERC + riser cards          │
├─────────────────────────────────────────────────────┤
│  REAR: PCIe slots (NICs, HBAs, GPUs), PSUs (x2),   │
│        iDRAC dedicated NIC, serial port, VGA,       │
│        LOM NICs (2-4 ports)                         │
└─────────────────────────────────────────────────────┘

Key things to notice: - Dual PSUs — A-feed and B-feed from separate PDUs. One can fail without downtime. - Service tag — 7-character alphanumeric ID on a pull-tab at the front. This is the server's identity for Dell support, warranty, and parts lookup.

One-liner: Get the service tag from Linux without looking at the physical server: dmidecode -s system-serial-number. This also works remotely via ipmitool or Redfish. The service tag is your skeleton key to Dell support — enter it at support.dell.com to get exact hardware config, warranty status, driver downloads, and firmware updates for that specific machine. - iDRAC direct port — micro USB on the front panel for emergency local iDRAC access via a laptop. - Drive indicator LEDs — green = healthy, amber = fault, blinking = identify/locating.

Lifecycle: From Box to Production

Unbox → Rack & cable → iDRAC IP setup → Firmware baseline
→ BIOS config (boot order, perf profile, VT-x)
→ RAID config (perccli or iDRAC)
→ PXE boot → OS install (kickstart/preseed/cloud-init)
→ Config management (Ansible) → Join cluster → Workload

At scale, you template the BIOS + RAID + firmware config with Server Configuration Profiles (SCP) exported from a golden reference server and applied via Redfish or RACADM.

What Experienced People Know

  • Always check firmware first. Half of "hardware problems" are fixed by updating iDRAC, BIOS, or PERC firmware.
  • RAID rebuild on large disks is slow. 8TB drives can take 12-24+ hours to rebuild. A second failure during rebuild means data loss. This is why RAID 6 or RAID 10 exists.
  • iDRAC access is your insurance policy. Always verify OOB access works before you need it at 2am. Dedicated NIC, correct VLAN, tested credentials.
  • Service tags map to everything. Warranty status, exact hardware config, driver/firmware downloads — all keyed off the 7-character service tag at support.dell.com.
  • Shared iDRAC NIC mode is a trap. It works until the host OS crashes and takes the NIC down with it — exactly when you need iDRAC most.
  • PERC cache policy matters. Write-back cache is fast but needs a healthy battery/supercap. If the cache battery dies, the controller falls back to write-through and I/O performance craters. Monitor it.
  • Label your cables. A datacenter with 200 identical-looking 2U servers and unlabeled cables is a disaster waiting to happen.

See Also

Wiki Navigation

Prerequisites

Related Content