010-151 DCTECH flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Cat6a 10GBASE-T distance

Cat6a twisted-pair supports 10GBASE-T over a full 100 m channel, while Cat6 reaches only 55 m at 10 GbE. Choose Cat6a for new 10G copper runs.

OM4 multimode reach

OM4 multimode fiber carries 10GBASE-SR to about 400 m and 40G/100G short-reach links over shorter spans. It is the common choice inside and between data hall rows.

LC versus MPO connectors

LC is a small-form duplex connector used with SFP and SFP+ optics. MPO/MTP is a multi-fiber connector used with QSFP optics for 40G and 100G parallel links.

Passive DAC cable limit

Passive twinax DAC cables carry 10G and 25G signals reliably to roughly 5 to 7 m. Active DAC extends reach to about 10 to 15 m at higher cost and power.

APC versus UPC ferrules

APC ferrules are polished at an angle and UPC ferrules are flat. Mating the two causes high loss and reflection, so fiber patch panels must keep the polish type consistent end to end.

Rack unit height

One rack unit equals 44.45 mm, or 1.75 inches. A 42U rack therefore provides about 1.87 m of usable vertical mounting space for servers and switches.

Circuit derating at 80 percent

Continuous loads should not exceed 80 percent of a breaker rating. A 30 A 208 V circuit therefore delivers 24 A, or about 5.0 kVA, of sustained usable capacity.

Fiber bend radius rule

Installed fiber should not bend tighter than about ten times its cable diameter, and copper about four times. Tighter bends cause loss or break the glass core.

RDIMM versus LRDIMM

RDIMMs buffer address and command signals. LRDIMMs also buffer data, allowing more ranks per channel and higher total memory capacity at slightly higher latency.

Memory channel population rule

Populate all memory channels evenly with identical DIMMs per channel. An unbalanced layout forces single- or dual-channel operation and sharply reduces achievable memory bandwidth.

NUMA remote memory penalty

Each CPU socket has local memory attached to it. A process reading memory owned by the other socket pays higher latency, so latency-sensitive workloads are pinned to one NUMA node.

RAID 5 usable capacity

Usable capacity equals the number of drives minus one, multiplied by the smallest drive size. One drive of capacity is consumed by distributed parity across all members.

RAID 6 fault tolerance

RAID 6 stores two independent parity blocks and survives the simultaneous failure of any two member drives. It needs at least four drives and writes more slowly than RAID 5.

RAID 10 rebuild behavior

RAID 10 mirrors pairs and then stripes across them. A single drive failure rebuilds from its mirror partner only, so rebuild is fast and does not touch the whole array.

SSD DWPD endurance

Drive writes per day states how many full drive capacities can be written daily for the warranty period. Multiply DWPD by capacity by 365 for annual terabytes written.

NVMe protocol advantage

NVMe runs directly over PCIe with deep parallel queues and far lower latency than SAS or SATA, which were designed around spinning disks and single queue depths.

UEFI versus legacy BIOS boot

UEFI with GPT partitioning boots volumes larger than 2 TB and supports secure boot. Legacy BIOS with MBR partitioning cannot boot a volume beyond 2 TB.

SAS dual-port drives

SAS drives expose two independent ports, letting two controllers reach the same drive for redundant paths. SATA drives have one port and cannot provide that multipath redundancy.

802.1Q trunk tagging

A trunk port carries several VLANs by inserting a four-byte tag into each frame. The native VLAN is sent untagged, so both ends of the trunk must agree on it.

LACP link aggregation rules

All links in a LACP bundle must match in speed, duplex, and VLAN membership. Traffic is hashed per flow, so a single flow never exceeds one member link.

Jumbo frame MTU requirement

Jumbo frames use a 9000-byte MTU. Every device in the path, including the virtual switch and guest, must use the same value or large packets are silently dropped.

Standard versus distributed vSwitch

A standard virtual switch is configured separately on each host. A distributed switch is managed centrally and keeps port groups identical across hosts, simplifying migration.

VXLAN encapsulation basics

VXLAN wraps Layer 2 frames in UDP port 4789 with a 24-bit VNI, supporting about 16 million segments. It needs a routable IP underlay between tunnel endpoints.

IP hash teaming requirement

Route based on IP hash balances traffic by source and destination address pair. It requires a matching LACP or EtherChannel configuration on the physical switch.

iSCSI network best practice

Place iSCSI traffic on a dedicated VLAN or separate physical uplinks, keep jumbo frame settings identical end to end, and enable CHAP authentication between initiator and target.

BPDU guard on server ports

BPDU guard disables a port that receives spanning tree messages. Enabling it on edge ports facing servers prevents an accidental loop from a misconfigured host bridge.

Usable hosts per subnet

A /24 subnet provides 254 usable host addresses, /25 provides 126, and /26 provides 62, after reserving the network and broadcast addresses in each subnet.

Physical layer first troubleshooting

Start diagnosis at the cable, link LED, and transceiver, then move up to speed and duplex, VLAN and IP, routing, and finally the application. This isolates most faults fastest.

CRC errors on an interface

Rising CRC or frame check errors usually mean a damaged cable, a dirty or mismated optic, or a duplex mismatch rather than a software defect. Replace the physical path first.

Server power fault checks

Confirm both power supply LEDs are lit and both feeds are live, test BMC access, then read the system event log. The log usually names the failed supply or regulator.

Correctable versus uncorrectable ECC

Correctable ECC errors are logged and scrubbed automatically but signal a degrading DIMM. Uncorrectable errors trigger a machine check and typically require replacing the module.

Firmware baseline before deployment

Apply the vendor support matrix combination of BIOS, BMC, NIC firmware, and OS driver. Mismatched versions cause link flaps, boot failures, and unsupported configurations.