Multigigabit 2.5G and 5G Ethernet on Existing Cat5e - 夜莺博客

Multigigabit 2.5G and 5G Ethernet on Existing Cat5e

Wi-Fi 6 and 6E access points regularly exceed 1 Gbps of real throughput, which turns the access link into the bottleneck rather than the radio. Replacing every Cat5e run is rarely an option, which is why the 802.3bz standard exists: it reuses the 10GBASE-T signalling design at lower rates so that 2.5G and 5G clear existing cabling at the full 100 m reach.

Speed against cable type

Cable 1G 2.5G 5G 10G
Cat5e 100 m 100 m 100 m Not recommended
Cat6 100 m 100 m 100 m 55 m
Cat6A 100 m 100 m 100 m 100 m

The rule of thumb vendors offer for access links is that switch bandwidth to a wireless access point should be at least 75 percent of the radio bandwidth. That is why a 2.5G or 5G port is sufficient for most Wi-Fi 6 deployments: it removes the bottleneck without requiring a cabling replacement that would cost more than the switch.

Negotiation behaviour

Multigigabit ports autonegotiate down a ladder - 10G, 5G, 2.5G, 1G, 100M in the usual implementation - so a mismatch never breaks the link, it just runs slower than expected. Two consequences follow. First, "the AP is slow" is frequently a 1G negotiation over a marginal cable, not a radio problem, and the switch port counters and link speed are the first place to look. Second, because negotiation falls back silently, a run that is electrically marginal will not alarm; it will simply cap the rate. Certification of the run, or at least a check of what speed the port actually negotiated, is the only way to detect it.

! If a 2.5G-capable port negotiated at 1G, investigate before blaming the AP
show interfaces status
show interfaces counters errors

PoE is the harder constraint

Multigigabit is usually deployed together with PoE because the device at the other end is an access point, a camera or a lighting controller. The power envelope has moved from 15.4 W (802.3af) through 30 W (802.3at) to 60 W and 90 W under 802.3bt Type 3 and Type 4, and the switch provides it over the same pair set it is using for data. Two planning rules matter:

  • Budget at the switch and the closet, not per port. A 48-port PoE++ switch with a 1500 W budget will power 25 devices at 60 W and no more; the arithmetic that matters is the aggregate, not the per-port maximum.
  • Cabling heat matters. Bundled PoE cables in a closed tray dissipate power as heat, and elevated temperature reduces insertion loss headroom - which is precisely the parameter that limits the achievable data rate. Very high power over dense bundles is a real, measurable effect rather than a theoretical caveat.

Where it is worth the money

The strongest case is an access layer carrying Wi-Fi 6 or 6E access points, high-resolution cameras, and PoE lighting, on cabling that will not be replaced this decade. The weakest case is a general-purpose user port: a workstation or a phone gains nothing from 5G, and a plain 1G port is cheaper. A tiered design - multigigabit ports where the endpoints need them, 1G elsewhere, with 10G or 25G uplinks and full PoE budgets - is the usual compromise.

Related reading: Wi-Fi 6E and the 6 GHz band, Wi-Fi site survey methods, and Energy-Efficient Ethernet benefits and gotchas.

原文链接:https://www.cisco.com/c/en/us/solutions/collateral/enterprise-networks/catalyst-multigigabit-switching/nb-06-transform-multigigabit-wp-cte-en.html