Engineering

The multi-gig Ethernet migration: Motivations and implementations



Supplier “push” and consumer “pull”. When their efforts coordinate, moving the market in the same direction, the result can be meaningfully impactful.

Toward the beginning of last month’s introductory small form-factor pluggable (SFP) module tutorial, which was followed by a series of module teardowns, the first of which also appearing in September with another arriving later this month, I also noted the following observation.

I’ve subsequently become intrigued (also with pending editorial-coverage consequences) with the increasing (and dramatically so) cost-effectiveness of mainstream network switches, routers, and other devices based on 2.5 GbE technology.

Foreshadowing

I followed that statement with several paragraphs’ worth of past and present product examples suggestive of the dramatic advancements in performance and feature set at various price points in recent times. And I concluded that section of the piece with the following “teaser” prose.

Why 2.5 GbE (along with, to a lesser extent, 5 GbE) has become mainstream is a topic for another post another day (soon). Similarly, I’ll save for the near future more discussion on why 10 GbE SFP+ ports are appearing on mainstream gear like this.

That time is now. I’ll start with a brief reintroduction to the two unmanaged-switch case studies I highlighted last time. This one has eight 2.5 GbE RJ-45 ports, alongside two 10G SFP+ ports, and sells at Amazon for only a bit more than $40.

And this one, with four fewer 2.5 GbE ports, is priced $10 less than its larger-allotment sibling.

Admittedly, these specific devices are trendsetters from price standpoints at their associated feature sets. But not by much versus alternatives, if at all. Further to that point, neither Davuaz nor NICGIGA is a widely known brand, at least right now. But even established network equipment suppliers are selling now similar-featured products in the same price ballpark.

And admittedly, too, “vanilla” 1 GbE 5- and 8-port switches are still less expensive than these higher performance alternatives (does anyone else also remember when even those garnered significant markups over their 10 and 10/100 Mbit precursors?). But only by a factor of 2x, if that. Versus the 10x (or more) price premiums that network equipment manufactures were charging only a few years back. What’s changed, and why? Glad you asked.

Supplier “push”

The 1 GbE networking equipment market is effectively saturated at this point, at least from a wired-connectivity standpoint (hold that thought for wireless subsystem comments to come later in this piece). The only notable motivations for consumers to buy new gear are if existing hardware dies and needs to be replaced, or if it runs out of spare ports for newly added LAN clients.

What do you do, then, if you’re an equipment developer, or a supplier of silicon and software building blocks for that equipment? You rely on the longstanding “bigger numbers is better” marketing mantra in striving to convince consumers to upgrade their existing gear.

Take MaxLinear, for example. Ahead of the 2023 Computex conference, the company introduced a family of chipsets containing up to eight 2.5 GbE PHYs per device along with two optional 10G SerDes transceivers capable of implementing SFP+ ports, and supporting both unmanaged, “smart” and fully user-managed switch configurations.

At that same show, the company demonstrated functional prototype networking hardware based on those same chipsets. Subsequent teardowns suggest that system designs based on single-chip MaxLinear offerings have captured a not-insignificant percentage of the cost-effective 2.5 GbE equipment market, joining legacy multi-chip implementations leveraging Realtek and other IC suppliers.

These low-cost switches reportedly don’t hold up solidly to the heavy, multi-protocol and multi-speed packet payloads created by “power user” testers, although the degree of correlation (if any) between such worst-case traffic and more modest typical real-world data profiles also bears consideration. And I’ve also seen some online commentary suggestive of high operating temperatures and associated questionable long-term reliability, particularly of the devices’ SFP+ subsystems. To at least some degree, this discussion is likely reflective of the equipments’ fan-less designs; online forums’ comment traffic domination by those with an “axe to grind” versus those with more positive experiences should also be factored into any determination.

Consumer “pull”

I plan to upgrade my LAN from 1 GbE to a mix of 2.5 and 10 GbE spans and am accumulating equipment in anticipation of this pending transition. I’ve also set aside several other pieces of gear with my own teardown aspirations in mind. But regarding my LAN-update plans, you might reasonably ask “why”? After all, as my recent Comcast debug coverage series made clear, I don’t enjoy higher-than-1 GbE broadband downstream speeds and won’t for the foreseeable future. And putting aside my longstanding “why not” early-adopter curiosity tendencies, of course …

Part of the answer, likely unsurprisingly, is that WAN-sourced and -destined traffic is only a portion of the overall packet profusion on a typical LAN. Client-to-client traffic flows also bear consideration. To wit, I’ve happily already confirmed that although 10 GbE technically requires Cat6 Ethernet cable at a minimum, the several-dozen foot long archaic Cat5e span running from the furnace room LAN nexus to my office above it seemingly reliably handles 10 GbE speeds, too.

Still, unless I were shuttling ultra high-resolution video and other large-payload files and streams across the LAN (as I actually already do to a degree and intend to do more of in the future), and/or if the network was in use by a large family versus just myself and my wife, a 1 GbE LAN would continue to suffice. I might still be tempted to upgrade, but at lower equipment prices (therefore supplier profit margins) than otherwise if my update need was more acutely felt.

The solution to this dubious-demand predicament is multifold. For one thing, present company excepted, an increasing number of broadband customers do have multi-gig WAN tiers available to them courtesy of DOCSIS 3.1 Extended. Plus, in addition to Ethernet cable, several other easier-to-implement LAN connectivity options are also now capable of multi-gig speeds. Take MoCA, for example. My multiple recent editorial mentions of it are non-coincidental; in fact, it was the key impetus that sent me down this multi-gig “rabbit hole” in the first place.

A few months back, I was poking around the storage closet and came across the remaining two MoCA 1.1 units that had survived their product-generation siblings’ lightning-induced demises. I decided to put them on the to-donate pile and then recalled that I also had two sets of two MoCA 2.0 transceivers in my possession…which led me to then remember that these had also subsequently also been technology-superseded, by MoCA 2.5 successors supporting (as the spec version number suggests) up-to-2.5 Gbps transfer rates.

Initial skepticism as to the viability of such offerings for the mainstream volume consumer market led to surprise once I did a bit of research and realized how inexpensive 2.5 GbE switches and the like had become.

Or take Wi-Fi. Right now I’m running a Google Nest Wifi Wi-Fi 5 (802.11ac)-based mesh wireless network, with eventual plans to migrate to either a Wi-Fi 6 (802.11ax) successor comprising open source-converted, therefore mesh-capable, Linksys LN1301 (MX4300) routers, or a Wi-Fi 6E topology made up of Google Next Wifi Pro nodes.

Although, as Wikipedia notes, Wi-Fi 5 only “has multi-station throughput of at least 1.1 gigabit per second (1.1 Gbit/s) and single-link throughput of at least 500 megabits per second (0.5 Gbit/s)”, Wi-Fi 6 and 6E theoretical speeds are much greater; “Though the nominal data rate is only 37% higher than that of Wi-Fi 5, the throughput increases by at least four times, making it more efficient and reducing latency by 75%. The quintupling of overall throughput is made possible by higher spectral efficiency.” Not to mention Wi-Fi 6E’s added 6 GHz spectrum access.

That all said, both Wi-Fi 6 generations have already been superseded by Wi-Fi 7 (IEEE 802.11be) for high-end gear, complete with claims that “in a single band, throughput reaches a theoretical maximum of 23 Gbit/s.” And initial Wi-Fi 8 (802.11bn) hardware is already being demonstrated, complete with integrated 10 GbE RJ-45 wired ports.

Wikipedia wisely notes, regarding Wi-Fi 6 peak bandwidth estimates and equally relevant to other Wi-Fi generations, that “actual results are much lower”. But as I said earlier, don’t underestimate “bigger numbers are better” allure.

What about 5 GbE, and why 10G SFP(+)?

You likely already noticed that both switches showcased earlier included not only multiple 2.5 GbE RJ-45 ports but also two 10G SFP+ module cages.

The latter’s existence is more of a “stretch” to rationalize, particularly in consumer applications. This is likely why although their presence is common, it’s not pervasive, and some 2.5 GbE switches make do with one SFP+ port as an interim step. They’re often referred to as “uplinks”, reflective of their employ as connections to higher-layer network gear gear such as routers.

That said, I’m not personally aware of any consumer (versus enterprise) routers with SFP connectivity. However, they can also more generally find use in high-speed tethering multiple switches together to expand the locational topology if, as noted earlier, wired client counts grow in the future. And if you just need another RJ-45 port in your switch, a SFP+ Ethernet adapter will neatly solve your issue, too.

And more generally, what about 5 GbE? Right now, gear prices versus both 1 GbE and 2.5 GbE alternatives remain high, with associated equipment availability correspondingly limited (both as measured by overall volume and supplier and product options). That said, just as is happening now with 2.5 GbE vs legacy 1 GbE, I have no doubt that a few years down the road, once 2.5 GbE demand starts to level off with increasing market saturation, silicon suppliers and their system partners will decrease costs and fire up the 5 GbE hype machine next.

Have you already begun phasing in 2.5 (or 5, for that matter) GbE and 10G SFP+ equipment to your residence or business? Or do you have plans to do in the future? Please share your motivations, hesitations and planned timeframes in the comments!

—Brian Dipert is the associate editor, as well as a contributing editor, at EDN.

Related Content

The post The multi-gig Ethernet migration: Motivations and implementations appeared first on EDN.



Source link