Introduction
As the cost of networking equipment drops further and further, the accessibility of such equipment to the consumer increases inversely. That is, where only a few years ago, 100Mbit network cards and hubs were all but out-of-reach for the average Joe, today they are common. In fact, the old stalwart of the home network, the hub, is gradually disappearing from business and home networks alike, as it gets replaced by switches offering far better performance at very attractive prices. And, as always, increased demand has created increased supply: There are now a variety of very good, low-cost fast Ethernet switches coming out on the market. We figured that it was time we found out just how good this new breed of switches is, so we took five of them from four popular manufacturers and put them through their paces.

Technology Brief
Before we get to the comparisons, though, you might be wondering what the difference is between a hub and a switch, and what makes one better than the other. Simply put, a hub is much like a cable splitter. It takes in signals from each port and feeds them to all the other ports. This means that the signal from each node (computer, printer, etc) connected to the hub goes to every other node connected to the hub. That ensures that the node that the message is intended for will get it. Unfortunately, it means that all the nodes the message isn't intended for get it too. This makes for a lot of unnecessary traffic across the hub, and on the network. That, in turn, means that the network gets more and more congested as more nodes are added. Since there is no way for the hub to manage the traffic it receives, packets run into each other fairly often (collisions). These collisions fragment the packets involved, so they have to be re-sent, increasing delivery times and thereby lowering the effective speed of the network.

While hubs merely pass packets along the wire, switches are intelligent enough to manage the packets they receive in a number of ways. Without going into OSI model specifics, let's say that switches are able to "look inside" each packet to a certain degree. Inside each packet is the MAC address of the sender and intended receiver of the packet. The MAC (Media Access Control) address is a unique identifying number borne by each node on the network.

By keeping track of these MAC addresses, the switch is able to tell on what port each network node resides. For example, a packet comes in on port 2 with destination address X and source address Y. The switch immediately knows that the address Y is on port 2 as the packet has come in to that port. In the meantime, a packet comes in to port 5 with the destination address Z and the source address X. The switch now knows that X resides on port 5, and thus has the destination port of the original packet from port 2 (MAC address Y). Theoretically, this chain of events only needs to happen once for each MAC address, because each switch has an address table built into it which holds the information for future reference.

Technology Continued...
In addition to reducing traffic on each port, switches that conform to the IEEE 802.3 specification (which is all of the ones we tested) are able to further reduce the number of collisions by using CSMA/CD (Carrier Sensing Multiple Access/Collision Detection). This feature is a media-access method which allows the switch to, among other things, check the line for traffic before it sends data. If it senses that there is traffic on the line, it waits for the line to clear before sending data. CSMA/CD also enables the switch to listen to each packet that comes across it, and discard damaged or fragmented packets, thus reducing traffic even further.

One final note about switching technology: All of the products featured in this round-up are 'store-and-forward' type switches, as are virtually all new switches today. Store-and-forward switches take in each packet completely before sending it out to the destination. This allows the switch to analyze the packet (to see if it might be a fragment from a collision, for instance) and decide if it should be sent or dropped. The other predominant type of switching technology is called 'cross-point' switching. A cross-point switch begins to send the packet out before it even receives the entire thing. The cross-point switch used to be the choice for expensive, enterprise switching due to it's lower latency (doesn't wait for the whole packet) and overhead (doesn't analyze packet size, etc). Today, however, store-and-forward switching technology has evolved to the point where it is so fast that the benefits of entirely taking in and analyzing the packet before sending actually outweigh the detriments. That is, the dropping of packets determined to be bad by the analysis reduce traffic enough to more than offset the minor gain in speed of using a cross-point switch.

Now that you know all you ever wanted to know about switch technology, let's take a look at the switches themselves. First off, we should go over the things these products have in common, in order to get a baseline from which to start the comparison. One of the most important things to note is that all the products in this round-up come with unlimited, free technical support. Although switches are extremely simple to set up and get running, the free tech support could become very important if you do encounter any issues. All the switches here are able to auto-sense and auto-adjust for speed (10 or 100 Mbps) and duplexing.

The mention of duplexing brings up an interesting point: you will see notices by some of the manufacturers, most notably the huge ad on the Linksys box, that the switch is capable of "Blazing Speeds" of 200 Mbps per port. This is a bit of a marketing spin, in that the 200 Mbps is really 100 Mbps each direction (as in full-duplex). Don't be fooled into thinking that you will be able to transfer files at 200 Mbps…it's not gonna happen.

Each switch also carries a MAC address table for storing addresses for the purposes mentioned above in the tech brief. While the space for possible entries varies, from 1000 to an impressive 12288 in the case of the Linksys, but all have plenty of space considering the intended use of the switches. If you have more than 1000 MAC addresses to deal with, the table limits will be the least of your worries J. Finally, all of these switches are able to daisy-chain to other switches for expansion via an "uplink" port. The uplink ports are, without exception, shared with one of the regular ports, so that you lose the functionality of one port on each switch when expanding them. That is to say that two of the 8-port switches together will give you 14 available ports rather than 16.

D-Link DS-8+

The first product is D-Link's DS-8+. This is the smallest 8-port switch we tested. At only about 6 ¾ inches wide by 3 7/8 inches deep and 1 inch high, it is almost exactly the same size as the Netgear 5-port product. The diminutive metal case holds an LED display which is very well laid-out. It also contains a very small, and very quiet, cooling fan on the left side, which keeps the DS-8+ from getting the least bit warm, even when left on 7x24. The ports are on the back of the switch, allowing the LED display to take up the entire front panel which, along with the brightness of the LEDs themselves, makes the switch's display very easy to read.

The tiny DS8+ gets high marks right away for being so small (8-port, 100tx switching that literally fits in a jacket pocket or a ladies' purse), and for having a street price of about $119.99.

Features

• Store and Forward switching scheme ensures data integrity
• Auto-polarity feature corrects reversed polarity on the transmit and receive twisted-pairs for each port
• 100% full wire speed data forwarding for 100Mbps Fast Ethernet (148,880 pps) and 10 Mbps Ethernet (14880 pps) on all ports
• 1.6 gigabit per second aggregate bandwidth supporting full-duplex Fast Ethernet connections on every port.
• Data filtering eliminates all bad packets (CRC Align errors, runts, fragments, etc.) at 100% wire speed for all ports
• 1K active MAC address entries with self-learning and table aging
• 1 MB memory with dynamic port buffering reduces lost packets
• Number of Ports: 8
• 10Mb or 100Mb NWAY auto-negotiation on all 8 ports
• All ports RJ-45
• 10Mb connections support Cat. 3, 4, 5 UTP or STP cabling
• 100Mb connections support Cat. 5 UTP or STP cabling
• IEEE 802.3 10Base-T Ethernet
• IEEE 802.3u 100Base-TX Class II Fast Ethernet repeater
• IEEE 802.3.1d
• Duplex: Half or Full per port
• Protocol: CSMA/CD
• Partitioning: Automatic for each port
• Uplink Port: MDI-II RJ-45 shared with port 1
• LED's Per Port: Link/Rx, auto-partition and port speed (10/100Mbps)
• LED's Per Device: Power, Collision (10/100Mbps)
• Operating Temperature: -10 degrees to 55 degrees C (14 degrees to 131 degrees F)
• Humidity: 5% to 95% non-condensing
• Classification: FCC Class A, CE Mark, VCCI Class A, CSA 950, UL 1950, C-Tick, TUV/GS

SMC EZ Switch 2 6308tx

Next up is the 6308tx. SMC has put together a really nice looking product here. As with the D-Link, the ports are in the back, so there is plenty of room for the LED display. SMC has made good use of this space, too, because the display features large LEDs that are very easy to see and to read. The only failing here is the absence of a collision indicator. While you shouldn't expect many collisions from a switch when all is well, a collision light can help you quickly diagnose a problem, such as a chattering NIC (the culprit in a broadcast storm). Other than that, though, the SMC has arguably the best display of the bunch. The light gray case is a reasonably-sized 8.66x5.24x1.46 inches. It's rounded shape, while attractive, means that it is difficult to stack any switch other than another SMC 6300 series on top of it. Not just another pretty face, the practically priced SMC 6308tx comes in at a $134.99 street price.

Features/Specs

Ports 5 or 8 RJ-45 ports, auto-negotiation Ports 1-4/7: Fixed crossover Port 5/8: Alternate crossover or straight-through Network Interface 10BASE-T RJ-45 UTP Categories 3, 4, 5 100BASE-TX RJ-45 UTP Category 5 Forwarding Mode Store-and-forward MAC Address Table 2K entries per system Filtering/Forwarding/Learning Rates Full line rate Memory Buffer 244 Kbytes per system System Status LEDs Power Link/Activity FDX (full duplex) Size 8.66in. x 5.24in. x 1.46 in. Weight SMC6305TX - 1.22 lbs. SMC6308TX - 1.301 lbs. Temperature Operating 32º to 104ºF Storage -40º to 158ºF

Humidity 5% to 95% Power Supply 9 to 12 VDC, 1.5A Power Consumption SMC6305TX - 5.6 W SMC6308TX - 9.0 W Heat Dissipation SMC6305TX - 19 BTU/hr SMC6308TX - 31 BTU/hr Maximum Current SMC6305TX - 0.62A SMC6308TX - 1.00A Standards IEEE 802.3u Fast Ethernet IEEE 802.3 Ethernet ISO/IEC 8802-3 EMC/Safety Compliances CE Mark Emissions FCC Class B VCCI Class B Industry Canada Class B EN55022 (CISPR 22) Class B C-Tick - AS/NZS 3548 (1995) Class B Immunity IEC 1000-4-2/3/4/6 Safety CSA/NRTL (CSA 22.2 No. 950 & UL 1950) EN60950 (TÜV/GS) IEC60950 (CB Report) Warranty Limited lifetime

Linksys EZXS88W

Aside from the long and cryptic model number, the first thing you'll notice about the Linksys EZXS88W is the size. It is easily the largest switch we tested at 6 inches long, 2 inches high, and fully 11 inches wide. This is still not a huge, hulking beast mind you, and it is very light. However, for all it's size, the 88W has the smallest LED display of all the switches. It's 32 separate LEDs are neatly and tightly packed into a 1 ½ by ½ inch space, making it very difficult to read at any distance over a couple of feet. The ports are in the front, which is part of the space problem, but there is still plenty of unused real estate on the face of the switch that could hold LEDs. Even if the small layout if necessary for some reason, a better system of indicators could be used, such as D-Link's use of two-colored LEDs. The case itself is computer-case-beige and almost perfectly rectangular, so it should fit in with the rest of the equipment in your rack or shelf. Price-wise, the Linksys is the lowest of all the contenders, at an amazing $108.95 street price.

Features/Specs

• 8-Port 10/100 Dual-Speed Per-Port
• Perfect for Optimizing 10BaseT and 100BaseTX Hardware on the Same Network
• Reach Speeds of Up To 200Mbps in Full Duplex Operation
• Eliminate Bandwidth Constraints and Clear Up Bottlenecks
• NWAY Technology Detects Cabling Type, Speed, and Duplex Operation
• Advanced Store-and-Forward Packet Switching
• Auto-Partitioning Protects PCs from Downed Network Lines
• Preamble Regeneration and Incoming Frame Retiming
• VLSI Components for Reliability
• Free Technical Support on the Phone and on the Web
• Five Year Limited Warranty
• Model No.: EZXS88W
• Standard: IEEE 802.3 and 802.3u
• Topology: Star
• Protocol: CSMA/CD
• Speed: 100Mbps & 10Mbps
• Ports: 8 ports and a shared Uplink port
• LEDs: Power & Collision, Link/Activity, Full Duplex, and 100 Per Port
• Power: External AC Adapter
• Input: 100-240VAC 50-60Hz
• Output: 5VDC, 3-6A
• Certification: FCC Class A & CE
• Cabling: Category 5 UTP/STP
• Dimension: 11"x6.25"x2.1"
• Weight: 1lb 15oz

Netgear FS-105, FS-108

The FS-105 and FS-108 are basically the same switch, except that the 105 has 5 ports and the 108 has, you guessed it, 8 ports. As with the Linksys, the ports are on the front of the switch. Back or front is a choice determined by your intended use of the product, but all things being equal, having the ports on the back of the switch makes for a much cleaner, slicker piece of equipment. Not because the ports themselves look bad, but the cables patched into the front, especially if ratty-looking or different colors (or both) can make the whole setup look pretty cobbled-together. Anyway, for having the ports on the front, the Netgear switches look pretty good and have kept a pretty small profile. They are only 4.1 inches deep and 1.1 inches high. The 8-port is 9.3 inches wide, while the 5-port is 6.2 inches wide. The ports and LEDs are widely spaced from each other, because the LEDs for each port are set on the ports themselves , rather than being grouped together separately for the ports. This makes it easy to distinguish which lights go with which port, at least when the switch isn't loaded up. In high traffic situations it seemed a little blurry, but maybe it was just our eyes J The cases are an attractive blue metal, and look and feel like the high quality units you'd expect from Netgear. They are also competitively priced at $99.99 for the FS-105 and $119.99 for the FS-108.

Features

	FS105	FS108
Network Ports	5 auto speed sensing UTP ports	8 auto speed sensing UTP ports
Forwarding Mode	Store-and-Forward	Store-and-Forward
Performance Forward Rate (10 Mbps port) Forward Rate (100Mbps port) Latency (10 to 100 Mbps) MAC addresses 10/100 buffer memory	14,800 packets/sec 148,000 packets/sec 75 usec max 1000 1 MB per port	14,800 packets/sec 148,000 packets/sec 75 usec max 8000 128 KB per port
AC Power	7.5 W	12.5 W
Dimensions	W 157 mm (6.2”) D 103 mm (4.1”) H 27 mm (1.1”)	W 235 mm (9.3”) D 103 mm (4.1”) H 27 mm (1.1”)
Weight	0.6 kg (1.25 lb)	0.74 kg (1.7 lb)
Environmental Specifications	Operating temperature: 0ºC to 40ºC Operating humidity: 10% to 90% non-condensing	Operating temperature: 0ºC to 40ºC Operating humidity: 10% to 90% non-condensing
Safety Agency Approvals	UL (UL 1950), CUL,T-mark	UL (UL 1950), CUL,T-mark
Status LEDs	Power Collision/Rx/Tx and speed indicators for each port. Link and duplex indicators built-in to each vista RJ45 port	Power Collision/Rx/Tx and speed indicators for each port. Link and duplex indicators built-in to each vista RJ45 port
Electromagnetic Compliance	>CE mark, commercial FCC Part 15 Class A EN55 022 (CISPR 22), Class A VCCI Class A C-Tick	CE mark, commercial FCC Part 15 Class A EN55 022 (CISPR 22), Class A VCCI Class A C-Tick
Standards Compliance	IEEE 802.3i 10BASE-T Ethernet IEEE 802.3u 100BASE-TX Fast Ethernet Windows, MacOS, Netware, TCP/IP	IEEE 802.3i 10BASE-T Ethernet IEEE 802.3u 100BASE-TX Fast Ethernet Windows, MacOS, Netware, TCP/IP
Warranty	Switch - 5 years Power adapter - 1 year	Switch - 5 years Power adapter - 1 year

The Tests
We engineered the tests to show the true throughput of the switches by avoiding any dependence on other hardware in the test machines. The unloaded test were done by hooking only the two test machines to the switch and running our favorite throughput testing tool, NetCPS. The beauty of NetCPS is that is can be configured to pass any amount of data from point to point without ever using the hard drive. It's small enough to sit completely in memory, and the data it passes is made up on-the-fly rather than in a file sitting on the test machine. We ran three passes of a 500 MB data transfer and three passes of a 30 MB data transfer for each of the switches. The NICs in each machine were Intel 10/100s, and both test machines were Celeron 500s with 128MBs of RAM. The loaded tests were performed with the same two test machines and three machines for creating network traffic. We performed exactly the same data sizes and number of passes as with the unloaded test, of course. To demonstrate the contrast between switches and hubs, we also ran a high quality 10/100 Ethernet hub through the loaded and unloaded 500 MB transfer tests and included the results in the comparison table below. The loading setup is shown in the diagram below:

Here's how it works:

1)	The loading machines, 1 and 2, download a single 1 GB binary file from the FTP Server (red lines)
2)	At the same time, machines 1 and 2 also perform a 1 GB NetCPS transfer between them (green lines) Note: Machines 1 and 2 each have only one physical connection to the switch. The lines are separated to indicate the two separate, silmultaneous processes.
3)	The test machines, A and B, perform the NetCPS transfers recorded in the comparison table below.

The Results
Below are the results of our tests, followed by a comparison table showing the average speeds for the transfers in Mbps:

Load	Product	Pass 1 (Mbps)	Pass 2 (Mbps)	Pass 3 (Mbps)	MB Transfered
Unloaded	D-Link DSS-8+	10.09	8.49	8.48	30
Loaded	D-Link DSS-8+	8.67	8.84	8.66	30
Unloaded	D-Link DSS-8+	10.27	10.02	10.10	500
Loaded	D-Link DSS-8+	10.58	10.23	10.48	500
Unloaded	Linksys EZXS88W	9.70	7.96	8.84	30
Loaded	Linksys EZXS88W	8.73	8.80	8.55	30
Unloaded	Linksys EZXS88W	10.14	10.33	10.17	500
Loaded	Linksys EZXS88W	10.03	10.06	10.42	500
Unloaded	Netgear FS-108	9.99	8.03	8.91	30
Loaded	Netgear FS-108	8.52	8.67	8.71	30
Unloaded	Netgear FS-108	10.44	9.47	10.49	500
Loaded	Netgear FS-108	10.42	10.17	9.85	500
Unloaded	Netgear FS-105	8.53	8.75	8.89	30
Loaded	Netgear FS-105	8.88	8.45	8.32	30
Unloaded	Netgear FS-105	10.33	10.02	9.60	500
Loaded	Netgear FS-105	10.23	9.76	10.29	500
Unloaded	SMC EZ Switch 2 6308tx	8.77	9.02	8.66	30
Loaded	SMC EZ Switch 2 6308tx	8.53	8.62	8.62	30
Unloaded	SMC EZ Switch 2 6308tx+	9.79	9.94	9.90	500
Loaded	SMC EZ Switch 2 6308tx	10.57	10.40	9.67	500

Unloaded Summary

Switch	500 MB transfer. Avg. time in Mbps	30 MB transfer. Avg. time in Mbps	Unloaded avg
Linksys EZXS88W	81.68	70.64	76.16
Dlink DSS-8+	81.04	72.16	76.60
Smc 6308tx	81.44	70.56	76.00
Netgear FS-105	79.84	70.56	75.20
Netgear FS-108	80.24	71.84	76.04
Hub			76

Loaded Summary

Switch	500 MB transfer. Avg. time in Mbps	30 MB transfer. Avg. time in Mbps	Loaded avg
Linksys EZXS88W	81.36	69.60	75.48
Dlink DSS-8+	83.44	69.76	76.60
Smc 6308tx	81.68	68.72	75.20
Netgear FS-105	80.72	68.40	74.56
Netgear FS-108	81.20	69.04	75.12
Hub			28.88

Before we get to how the results boil down, take a look at the hub statistics. As might be expected, the hub's throughput of 76 MBps is about 94% as fast as the average of the times posted by the switches in the same test. Now look at the hub under load. It's speed of 28.88 Mbps is 35.38% of the average speed of the switches under load. This illustrates what we said earlier about the differences between the two devices. The increased traffic caused by collisions and broadcasts drastically affects the performance of the hub under load. The hub that we used sells on buy.com for $134.99, exactly the price of the most expensive switch we tested. There are less expensive hubs out there, but you won't find any that are so inexpensive that they justify the loss in performance. For our money, we'll take the switch every time.

Now on to the switch summaries:

Category	Best	Worst	Margin (%)
Fastest (loaded/unloaded avg)	D-Link DSS-8+	Netgear FS-105	2.3
Performance drop under load	D-Link DSS-8+	Netgear FS-108	1.3
Bang for the Buck ($ per Mbps)	Linksys EZXS88W SMC 6308tx	SMC 6308tx	19.5

Conclusion
So, which switch is the best? Well, the fastest is separated from the slowest by about 1.72 Mbps, which means it would be about 2.4 seconds slower to copy a Gigabyte of data across the network. In other words: no significant difference. OK, let's try it another way. The switch that took the worst beating in performance while under load dropped in speed by 0.92 Mbps. That means it would take just under 1.3 seconds more time to copy a Gigabyte of data when loaded. Once again, no significant difference. How about price? Nope, foiled again. These switches all sit within a $35.00 price range. You could choose one by which has the best warranty: The D-Link and SMC have lifetime warranties, while the switches from Linksys and Netgear only have 5 years (only 5 years…yeah, right). The simple fact is, these products are all so close together that you might better pick one by which color you like best or whether the ports are in the back or the front.

At this point you might be thinking "Gee, there's no clear winner here." But there is: You, the consumer. The price to performance ratio of any of these switches means that for about 3 times the price of a 10 MB hub, you are going to get anywhere from 10 to 30 times the speed. The happy truth is that no matter which of these switches you buy, you are going to get awesome performance for your money.