The need for fast data transfers has been rapidly increasing. Gigabit network switches are now the norm in homes and businesses. It wasn’t long ago most home wifi routers only had a handful of 100Mbps network ports and WiFi N offering meager 600Mbps. We now have routers with 1Gbps ports for internet access (although many home broadband connections still don’t come close to those speeds), and home wifi advertised as 3600Mbps (with a bunch of asterisks of course). To facilitate those wifi access points that can handle multiple channels servicing WiFi 6 (802.11ax) or even WiFi 5 (802.11ac), “multi-gig” networking is becoming more common on new devices, and we need networking equipment that can handle them.
What is multi-gig and how does it help? In short, “multi-gig” refers to an update to the Ethernet specification that allows 2.5Gbps and 5Gbps connections over existing Cat5e and Cat6 UTP (twisted pair) copper cables. These are the patch cables you use to plug your computer into your router. Cat6 is a higher-standard cable than the older Cat5e (you should be able to read the cable grade printed on the side of your networking cable). When gigabit ethernet was first hitting mainstream, it was recommended to use Cat6 to ensure a high-quality 1Gbps connection, but the goal of the spec was to also allow 1Gbps over the older Cat5e cabling that was more ubiquitous, but with limitations on distance. Multi-gig comes to us in a similar fashion, allowing 2.5Gbps and 5Gbps over existing cabling infrastructure, but with limitations on distance depending on the cable specification. Why the history lesson? It’s to show that multi-gig is another evolution, just like what we’ve seen before, and has some similar limitations (mainly distances) to be mindful of.
EnGenius sent us their ECS2512FP switch to try out and review, and it’s absolutely chock-full of the latest in switching technologies, with multi-gig being just one of the many features it holds. Due to heavy multimedia loads on my network, I had been looking for a managed switch to upgrade my home lab with. My existing 1GbE switch was causing my frequent file transfers to take several minutes which adds up to a lot of wasted time waiting for transfers to complete before being able to continue additional tasks with those files. Looking to cut down on that idle time, I knew I needed multi-gig or 10Gig and I had already been looking at options, so EnGenius was rather timely with asking us to take a look at their offering in this space.
The EnGenius ECS2512FP Cloud-Managed 8-Port 240W PoE++ Multi-Gigabit (2.5G) Switch with 4 SFP+ 10Gbps Uplink Ports
That’s a lot in a title, but sums up the primary features succinctly. The ECS2512FP reads to me as a business-class (or prosumer!) switch designed to be an edge switch for multi-gig WiFi access points, such as EnGenius’s EWS377AP which packs 3600Mbps of aggregate WiFi throughput. This switch has a total power budget of 240W, averaging 30W over the 8 multi-gig PoE ports. Since the ports are 802.3af/at/bt. it means you can have devices up to 60W on a single port (obviously limiting the remaining power budget of course). If PoE isn’t a requirement, do check out the non-PoE version of this switch, the ECS2512. While multi-gig technically goes up to 5Gbps, the ECS2512FP only enables up to 2.5Gbps on its PoE ports (commonly written out as 10/100/1000/2500 Base-T GbE). If you have some ultra-high-end motherboards that feature 5Gbps NICs, they’ll simply fall back to 2.5Gbps, which is still 2.5X the speed of common 1Gbps ports. Thus, with 20Gbps total between the 8 ports, and another 40Gbps potentially from the SFP+ ports, the switch can only physically have 60Gbps of aggregate data communications. However, the switch’s total switching capacity on the spec sheet is 120Gbps, because ethernet communications are full-duplex, so “1Gbps ethernet” is technically 1Gbps up plus 1Gbps down making 2Gbps of theoretical throughput. Speaking of those uplink ports, they’re dual-speed 10G SFP+ ports. EnGenius sells their own SFP+ transceivers, such as the SFP3213-10 10G single-mode transceiver, but since using fiber optics with the factory SFPs is obviously going to work, we opted to test some alternatives instead. The last high-level feature is their full-featured Layer 2+ (or what some companies may call Layer 3 Lite). We’ll talk more in depth on this later.
The ECS2512FP has some key features for enterprise applications, namely VLAN support, static routes, PoE (Power-Over-Ethernet), and 10Gig SFP+ uplinks. The full-featured Layer2 support adds many needed items as well, including spanning tree and loopback detection, link aggregation, QoS, SNMP, and ACLs (access-control lists). You can read through the ECS2412FP whitepaper for a detailed list of specifications.
EnGenius ECS2512FP Internals
The internals of the switch is nice and organized. I didn’t remove the heatsinks on the logic chips so as not to disturb the ship-as condition, but we can spot at least the one thing I was curious about.
Tucked up on the left side of the board is not one, but two Broadcom BCM59121B0KMLG ICs to drive the Power Over Ethernet. The IC is rated at 30W per channel and can drive up to 8 channels, which alone would have given us our 240W power budget for the switch. However, the IC is only 802.3af/at rated, so in order to obtain the 802.3bt for 60W per port, there’s a second Broadcom IC complex (top-left corner of the above picture) to help provide an extra 30W per port. I’m assuming there are other limitations preventing these two chips from providing a full 60W per port for all eight ports, likely in power delivery, but I’m not as good at reading circuitry as some others in the tech media.
The last thing to mention is the dual small fans on the left side of the enclosure. It’s difficult to actively cool switches quietly, and since this switch is designed for a telecom closet or server room, decibels aren’t really an issue. I positioned this switch on top of my home NVR in my home lab, which is also actively cooled, and they both have the same noise level, and are the noise contributors that keep my home lab at -51dBA. It’s certainly quieter than the ~83dBA of some server rooms I have to work in, and is quiet enough that a simple closet door should muffle it quite easily.
The Cloud Management UI
Sadly, “THROUGHPUT” is a measurement of Access Point SSIDs and not switch traffic, hence why there’s no trend data for my Dashboard, but also makes the dashboard landing page practically useless for just this switch. This AP-centric Cloud Management portal reaffirms my original suspicions that this particular Multi-Gig switch was designed with the EnGenius APs in mind. I just wish I could utilize some of the dashboard space to be relevant for the switch if I don’t have the APs.
Best feature of Cloud Management (outside the obvious modifying switch settings while off-site of course!) is the alerts. This setting can be a canary for home internet or power outages, or even problems with the switch like a firmware lockup (which I haven’t experienced with this switch, but it has happened to me with other switches).
This also brings me to another point, you’ll notice there’s “Professional” tier settings which you get if you upgrade from the free Basic to a Pro cloud subscription. These alerts can actually be especially beneficial in a business environment, and you’d usually need an network monitoring tool like Solarwinds or Nagios to get these types of alerts, so they’re certainly appreciated. They’re estimating about $50(MSRP)/year per AP for this Pro license, but as it’s new they haven’t yet settled on the official price just yet. The service still works just fine with the free Basic license however.
Next, the feature I dislike the most: automatic updates using scheduled maintenance windows. Don’t get me wrong, the Network Security Admin inside of me loves this feature. Set and forget! However, even in my home environment, I don’t have a regular time I can point to that I’ll never have something not running. Even 2AM on a Sunday, I’m likely running some batch encoding process that may take all weekend to churn through. I’d much prefer this be an optional thing (even if it’s enabled by default for those that don’t change configs!). But, as you can see from the error message I included on this screenshot, maintenance window configuration is required. It’s ultimately the reason I removed my switch from being Cloud Managed. Now I’ll manually update the firmware when I know they’ve released a new version and I know I don’t have an active task running. This is obviously not a recommendation, but it’s important to me in my scenario.
What will I miss most of the Cloud Management portal? This dashboard. At a glance, I can see everything I need to know: firmware version, uptime, port usage (and speeds) and basic important settings like if QoS and STP is enabled. The web management UI on the switch doesn’t have an analog of this view; the information is scattered across several config pages.
Lastly, I was rather disappointed with the configurability of the switch through the Cloud Management portal. I was expecting something that would allow me to easily define vLANs, assign those vLANs to ports, handle port aggregation, etc, but I didn’t see anything to configure such. Unfortunately, once you’ve set up Cloud Management, you also can’t use the switch’s webUI to make those configurations either (it says the switch is managed by EnGenius Cloud and you’ll only be able to view diagnostic information). While a lot of those configuration options aren’t immediately useful in my home lab, they’re important in an enterprise environment, and I hope they’re just locked behind the Cloud’s Pro license, or are on the roadmap to be put into the Cloud Management portal which is still a fairly new offering in itself. But all the features I’m missing in the Cloud is very editable via the WebUI and Command-Line Interface (CLI) on the switch.
I’d also be remiss if I didn’t mention that the Cloud Management has an API for automating management tasks if you’re handy with scripting such things.
The Web Management UI
Once I logged in to the switch’s WebUI, I was quite thrilled to see the level of configurability. It instantly reminded me of the Cisco SG200-08P switches I manage at work (the one exception is the Cisco device has in its WebUI a screen similar to that switchport view I showed from the Cloud Management portal).
The collapsing left menus organize things nicely and using the WebUI is snappy, which isn’t as common as you might think for a on-switch WebUI.
Rather than enumerate all the numerous features found on this switch, I’m going to include a gallery of the expanded left-menu headings go give you an idea of what is in each section, and then talk about some specific features of note. There’s additional groupings under each heading that give more granular selections, but I won’t be exhaustively including those screenshots though.
Access Control Lists (ACLs) are lists of Access Control Entries (ACEs). The EnGenius ECS2512FP can define entries using MAC, IPv4, or IPv6 addresses. As you can see from the above screenshot, they’re fully-featured as well.
The onboard flash storage contains two system image partitions: a primary (active) and a backup. This feature alone tells you this is an enterprise-class switch with resiliency in mind. Many prosumer and lower-end business switches don’t have this, so it’s certainly appreciated to see it here, as this switch positions itself above such devices.
Another useful feature is cable diagnostics. Individual status for conductor pairs and estimated cable length on a per-pair basis. It’s obviously not as detailed as a Fluke cable tester, but if you’re not getting a cable to work, you can check for a miswire at least.
802.1X network authentication is also supported, providing a great way to keep edge ports secure in a corporate environment.
Lastly, here’s the port statistics that can be used to quickly pick out issues with connected devices. You can quickly spot the 23316 RXErrors on port 4, which is our 2.5Gbps USB NIC I’ve been testing, as well as the 5 RXErrors and 219K of discarded RX packets on port 9, which happens to be our 10G-BaseT SFP+. The twinax cable we’re also testing with is port 10 by the way.
We’re basically guaranteed to work effectively using EnGenius-branded SFP+ modules, but what about some other connectivity options out there? I picked up an HP-branded Intel 560SFP+ 10Gb 2-port adapter to test a common switch uplink and server connection method of twinax, which is a direct-attach copper passive cable with an SFP connector on each end. This saves money on needing to buy separate SFP+ modules and a fiber patch cable as well as the fiber connections that might get dirty or have issues over time. I opted for a 3-meter generic Twinax cable to represent a common intra-rack run length. Next, to give another 10-gig link to test the twinax against, I picked up a OSFPTEK QT-SFP-10G-T, which is a 10GBase-T SFP+ to RJ45 module which provides 10-gig connectivity over standard twisted-pair cabling. I used this to make a standard office-building-type Cat6 run, which means a patch cable on either end of a cable plant, and a patch panel and wall jack in the mix. Since the effectiveness of 10-gig over UTP (unshielded twisted pair) is limited by length (and obviously the quality of cable and connections), I picked a short run, keeping the whole end-to-end length a mere 20ft. Next up, we have a Plugable 2.5G USB-C ethernet adapter. This gives us the chance to test the robustness of a multi-gig add-on solution, as well as using a laptop to check length limitations. Lastly, I have an older ASRock Fatal1ty X370 Professional Gaming motherboard with an onboard AQUANTIA (Marvell AQtion) 5Gb NIC to test with as well.
For data transfer rates, at 10-gig, we quickly hit disk limitations before actual network speed caps. Transferring from a 3-disk RAID5 NAS on our twinax link to an NVMe 4.0 disk on the 10GBase-T SFP+ module, we saw a buffered boost and a leveling off around ~408MB/s, which is technically a great result for sequential read from a 3-disk RAID5, pushing those disks to the max. 10G should hit a theoretical maximum of 1.25GB/s, so we still have plenty of overhead to tap. To do so, we’ve have to move outside of the usual multi-spindle NAS setup to something more exotic like a NAS hosting RAIDed NVMe drives for multimedia editing workloads in order to harness the full bandwidth 10G affords. This makes multiple 10G uplinks for this multi-gig switch ideal, considering if we ran the switch loaded with EnGenius EWS377AP APs, each capable of maxing out their 2.5G connection, we could quickly saturate a single 10G in a dense, busy environment. This is likely why there’s four 10G uplink ports: LACP bonding of the uplinks, making two connection pairs for uplink and switch redundancy.
Our second transfer shows copying from the same 10G twinax-connected NAS to a NVMe 3.0 drive in a laptop, connected via the Plugable 2.5G USB network adapter. Theoretical max of a 2.5G connection is ~312MB/s, so if we drop 10% of that for TCP/IP overhead, we land quite exactly at the ~280MB/s we see from the transfer chart.
Next, I subjected these various links to prolonged workload testing; basically ran multiple instances of scripted handbrake transcodes for a few days on our Threadripper reviewed previously, directing the data source from our twinax-connected NAS and also from a NAS connected with the Plugable 2.5G USB adapter.
This is where we circle back to the Port Statistics page previewed earlier. Port 10, is our twinax-connected NAS, showing a resilient, solid connection with no discarded packets or transmission errors, but only half the traffic as our 10GBase-T-connected Port 9. Even with a short 20ft run using end-to-end Cat6-grade cabling and connectors, we still see discard and error packets. While the link is still quite suitable, it’s not ideal. It would likely be worse if attempted over Cat5e.
Next, we come to the Plugable 2.5G USB adapter on Port 4. I connected it to a NAS and ran a short 7ft Cat6 patch cable directly into Port 4 on the switch, removing any cable plants and wall jacks from the equation. I then ran it overnight as the data source feeding a couple of the simultaneous aforementioned handbrake scripts, but it failed sometime during the process, making the adapter unresponsive in Linux, requiring it to be unplugged and replugged back in to make it accessible again. That instantly makes it a fail in my book, as I view devices that need the increased connectivity speed as data-intensive. However, being a USB device, it’s more likely to be used in circumstances where it provides improved throughput for short bursts of large data, where it will then go back to nearly idle for prolonged periods of time. This could be a media laptop, dumping a day’s worth of video and images to a NAS. With sufficient cool-down time, it may not exhibit a similar lockup.
Lastly, I attempted to link up the AQUANTIA 5Gb on a 21-meter standard run, which includes the two patch cables and the infrastructure run from the patch panel to the remote wall jack. I say “attempted” as I only got the link to stabilize once after many attempts, and upon reboot, would not relink. It would do the usual auto-negotiate loop of initialize and disable that happens when the cable has problems with the speed the system wants to establish. It did this with both cable runs to that particular wall plate even swapping patch cables. The Plugable worked just fine on either wall jack, so I’m fairly certain the AQUANTIA is at fault on this one.
The Port Statistics, especially when errors start showing up, is a valuable tool for network administrators, so I highly appreciate EnGenius putting the ports and trunks into a spreadsheet view for quick and easy reference.
The EnGenius ECS2512FP is a solid device providing up to 60W of PoE power per port, with a 240W power budget, which is more than enough to run the current gen of multi-gig APs on the market. A enterprise-level of quad-10G uplink connectivity shows enterprise-connectivity thoughtfulness, rather than hamstringing the device with just dual uplink ports like some devices tend to. The WebUI on the device is well constructed, feature-rich, and snappy, making the CLI (while highly powerful in its own right) practically unnecessary unless you’re making a lot of pre-scripted changes. I was disappointed with the Cloud Management aspect though. I went in expecting it to be able to set up at least a portion of the switch’s advanced features, but found it able to do not much more than rudimentary settings and ended up defaulting back to local management. The forced maintenance window update scheduling, while only being bit by it once as firmware updates are obviously infrequent, made it undesirable for my use-case.
I had actually been looking for an affordable, but reliable way to upgrade my home lab’s core to multi-gig and could only find various 4-port devices or 8-port with single 10G uplinks in what I would consider a “reasonable” price point for home use. Looking at the non-PoE version of this switch, the ECS2512, provides an ideal assortment of 2.5G links and the generous 4 10G links that I was looking for. Even though I do run a Ruckus R500 with PoE, it would be cheaper for me to use an inline power injector instead of flexing up for the PoE version of this switch. It’s when one runs multiple PoE devices, that the ECS2512FP becomes a great value upgrade over the base switch.
The ECS2512FP readily took our inexpensive, generic 10G SFP+ module and cheap twinax cable, so it’s not locked in to vendor-specific equipment, which is always a plus when you’re looking to shave some costs.
If you have four or fewer 2.5G capable devices now or in your near future, there’s cheaper alternatives out there to get you increased networking speeds, but if you want a managed switch with an advanced feature-set and four 10G SFP+ ports, the ECS2512FP, or at least the non-PoE ECS2512, should be high on your consideration list.
Liked it? Take a second to support Kirk Johnson on Patreon!