High on Fibre

We tested three Fibre Channel switches to determine how compatible they really are.

December 9, 2002

15 Min Read
Network Computing logo

For our tests of midrange switches, three vendors sent us products that met our specifications. QLogic Corp.'s SANbox 2 and Brocade Communications Sytems' SilkWorm 3800 have 16 ports each; McData Corp.'s Sphereon 4500 is a 24-port unit.

Full Mesh Throughputclick to enlarge

We also invited Cisco Systems, Gadzoox and Vixel to participate in our tests. But Cisco's switches are currently in beta and therefore didn't qualify. Vixel declined, commenting that our Spirent SmartBits Fibre Channel test equipment is incompatible with its switch. The Spirent equipment worked fine with the other switches, so we wonder what's up with Vixel's switch. Gadzoox, currently under Chapter 11 protection from bankruptcy, claimed a shortage of available people and equipment to support the review, as the company moves its headquarters to a less pricey facility.

Performance testing, a key part of our review because of the importance of the traffic the switches carry, gave us a wild ride. Nervous vendors ushered their switches into Network Computing's Green Bay, Wis., Real-World Labs® like parents taking their children to kindergarten for the first time. We conducted full-mesh testing, in which every port sends frames to every other port; frame-loss testing; and several latency tests, including full-mesh, standard industry latency and our own latency-under-load test. Only the frame-loss tests came out exactly as expected: None of the switches dropped a single frame, even when we accidentally misconfigured a test.

None of these products can be characterized as a bad performer; however, results did vary. QLogic's SANbox 2 showed consistent numbers across the board. But Brocade SilkWorm 3800 gave us some squirrelly test results. At 100 percent load on the full-mesh latency test, for example, the Brocade switch's latency ranged from seven to 85 times that of the second-place competitor. Brocade attributed this anomaly to the SilkWorm 3800's extra buffering algorithms, which threw the results off because our test measures buffer latency only. (Brocade also does not have a single integrated circuit, which the McData and QLogic products have, nor is its buffer RAM integrated on the IC.)

McData's Sphereon 4500 showed some good latency numbers, especially in the full-mesh latency tests. Curiously, that switch also showed the worst numbers in the industry-standard port-to-port tests but very good results in latency-under-load tests and full-mesh latency tests.Working Together

Full Mesh Latencyclick to enlarge

Switch interoperability has finally arrived in the Fibre Channel market. After years of reluctance to embrace this practice (a common problem in the storage industry), at least some vendors are giving it a try, as Ethernet and IP have found room to compete in the storage market. McData and Brocade have been among the worst offenders. Brocade owns more than 90 percent of the midrange market, while McData dominates in the high-end director-class switch arena, yet neither has officially supported interoperability between their respective switch brands.

Recently, however, McData got religion. The company has transformed itself into an active force for interoperability--much like QLogic has always been--and is truly trying to make its products work with those of its competitors (those that will cooperate, at least). The company understands that having 90 percent market share doesn't do much good if the whole market gets killed by Ethernet and iSCSI.

Furthermore, Cisco has decided to enter the Fibre Channel market with the spin-in of Andiamo Systems. This market intrusion is the largest event to hit the Fibre Channel switch space in years, and Cisco's considerable marketing muscle and money will likely represent a huge challenge to the existing Fibre Channel switch companies once its products get off the ground.

Features Chartclick to enlarge

Brocade still hasn't had an interoperability epiphany, regardless of competing market forces. The vendor's habit of adding useful but incompatible features to its products damages not only Brocade but Fibre Channel as a whole. When confronted with the interoperability issue, Brocade representatives said it is not in the customers' best interest for Brocade to "dumb down" its Fibre Channel fabric to work with other switches. That sentiment might be true if customers wished to purchase only from Brocade; however, we like choices and believe that purchasers of Fibre Channel switches do too.QLogic Takes the Prize

We gave our Editor's Choice award to QLogic for its SANbox 2, a midrange switch with outstanding latency and throughput performance, good pricing, and excellent design. The McData Sphereon 4500 came in second with good latency numbers, outstanding price, but with the worst throughput numbers of the three. And though the Brocade SilkWorm 3800's performance was quirky but fine, that unit is simply too expensive compared to its competitors. Brocade's stand on interoperability didn't help its cause, but price was our main concern about that unit.QLogic Corp. SANbox 2 | McData Corp. Sphereon 4500 | Brocade Communication Systems SilkWorm 3800

QLogic Corp. SANbox 2

Great features, reasonable price, and excellent performance, especially on our latency and throughput tests: That's what we call a hat trick. Add ease of use to this mix, and the QLogic SANbox 2 is the obvious Editor's Choice.

Built for easy access inside and out, the SANbox 2 is a 16-port, 1U, 2-gigabit Fibre Channel switch with dual power supplies set up in a redundant configuration. Power cords plug into the front of the switch, and the power supplies are hot-swappable. Forget the trip to dust-bunny land after a power failure; if a power supply fails, pull it right out of the front. The ports are laid out in two rows of eight, with both GBICs (gigabit interface converters) backing each other so the release for the optical cables is on the bottom. This layout makes it easy to remove cables. The front also features a link light, fan and temperature lights and a "heartbeat" that indicates the switch is working properly.

We liked the SANbox 2's design and construction, which reminded us of QLogic's SANblade 2300 Fibre Channel HBAs (host bus adapters). The insides are simple, with a single integrated chip for all ports.

Average Port-to-Port Latencyclick to enlarge

This switch bested the competition in the majority of our tests; its performance on our latency and throughput tests really impressed us. In all but the full-mesh latency tests (which McData mostly won), the SANbox 2 had the best numbers. Latency, particularly on our latency-under-load tests, was evenly distributed. Only the SANbox 2 put out such even scores, a tribute to the switch's superior design. Furthermore, in the more challenging latency-under-load test, the SANbox2 behaved as consistently and predictably as it did in the standard latency tests.The SANbox 2 also proved the clear throughput leader. QLogic's switch handled the smallest frame size better than the other switches. This result bodes well for the SANbox 2, as much of a Fibre Channel SAN's message traffic tends to be on the small side.

QLogic's SANbox 2 comes with SANmanager management software, an independent Java-based application. Although we rated the software's interface and simplicity last among the three products' management tools, it was still easy to use. Both McData and Brocade use browser-based implementations, a smarter approach than SANmanager, which has to be installed on each machine you wish to use as a manager.

QLogic's switch is respectably priced, right in the middle of our pack, with a $17,995 list price and $17,935 street price. The QLogic SANbox 2 comes with a one-year warranty. Extended warranty and service is available. This is a solid product from a solid company that is committed to the storage industry in all aspects.

SANbox2 16 Port Fibre Channel Switch, $17,995. QLogic Corp., (877) 975-6442, (949) 389-6000. www.qlogic.com

McData Corp. Sphereon 4500

The McData Sphereon 4500 replaces the Sphereon 3216 in McData's midrange lineup. This new switch has the best price, with an intriguing pay-as-you-go plan; however, our tests showed that McData needs some improvement in both throughput and latency.

Unlike its two competitors, McData equips its 2-gigabit Fibre Channel switch with 24 fixed ports, rather than 16. However, the company sells the box for use in eight-port, 16-port, or 24-port configurations. A license key activates the unit in those eight-port increments. Such a port scheme provides good investment protection if you're installing a nascent SAN and are unsure of its growth potential.

The Sphereon 4500 has two hot-swappable power supplies/cooling assemblies. Unlike the SANbox 2, however, these units are removable from the rear, rather than the front, of the machine. An Ethernet management port resides on the unit's front, to the left of the Fibre Channel ports. The Sphereon 4500 is also one of two single-chip designs in our test. With its single 24-port chip, McData has designed the insides of this box cleanly.

Average Latency Under Loadclick to enlarge

The outside, however, is another story. The Sphereon 4500 is numbered right to left in a double row, with the zero port on the bottom right, rather than the more conventional left to right. McData tells us that the numbering is an engineering legacy; that's how all its past switches were numbered. We understand that, but nothing else with ports is numbered like that--and with good reason: It's a stupid idea, and needs to be changed on future switches.Although the Sphereon 4500 posted some good full-mesh latency test results, it came in last on the standard port-to-port latency and latency-under-load tests. We find the throughput results particularly disappointing, since we tested only 16 ports on this switch, rather than the full 24-port complement. While this result might indicate an early firmware problem, it's still a concern. The Sphereon 4500 should perform better than it does. On the latency front, the Sphereon 4500 seemed to perform well only while under load. Industry-standard latency performance and latency-under-load tests were not exactly what we expected either. However, the Sphereon 4500 performed very well on the full-mesh latency tests. We can only surmise that because McData's single chip design is newest, the speed of the RAM used for buffer credits must be superior to that of QLogic or Brocade. Superior RAM speed would have a positive effect on latency scores.

The Sphereon 4500 comes with SANpilot, a browser-based configuration- and fabric-management utility. SANpilot is the easiest configuration utility to use, though it has fewer features than Brocade's browser-based software. Unlike the others, this switch features nondisruptive code load, called HotCAT, a feature that's normally reserved for director-class switches. This handy feature lets new code be loaded to the switch and activated without disrupting traffic passing through the switch.

McData sells the Sphereon 4500 through its channel partners and offers a 13-month warranty on the switch. We had to ask: Why 13 months? McData said it was a customer courtesy, as most people end up having the switch sitting for a month before implementation. Nice. Almost nice enough to make up for the stupid port-numbering trick.

Because this article was written so close to the release date for the Sphereon 4500, we were unable to secure street prices for the switch. We estimate street prices to be 10 percent to 20 percent less than list. The list price, $15,000 for the 16-port version, includes optics, and the list price for the 24-port activated version of the Sphereon 2500 is a couple of thousand less than the street price of the 16-port Brocade SilkWorm 3800. McData offers an incentive to purchase the Sphereon 4500 at the 24-port activated price of $21,695 (list). It's far cheaper to purchase the ports up front, as each eight-port upgrade sells for $9,959.

Sphereon 4500 Fabric Switch, $21,695 fully populated including optics, McData Corp., (800) 545-5773, (303) 460-9200. www.mcdata.comBrocade Communication Systems SilkWorm 3800

Brocade dominates the midrange Fibre Channel switch market, and the SilkWorm 3800 has many features and a strong GUI. It's a bigger unit than the competitors' switches, and its ports are set in a single row, sensibly numbered from left to right. Perhaps these facts alone justify the SilkWorm 3800's $25,000 street price, because the unit's performance, while mostly fine, doesn't bear out the premium over QLogic's and McData's prices.

Our SilkWorm 3800 turned in some incongruous test results. At very small frame sizes, Brocade had both latency and throughput problems. Furthermore, when we ran our full-mesh latency test, the numbers were appalling. The latency at maximum frame size was seven to eight times that of the Sphereon 4500 and SANbox 2, respectively. At the smallest frame size, latency was 85 times that of the Sphereon 4500 and 211 times that of the SANbox 2.

Brocade's representative said the numbers we got at 100 percent load measured buffer latency. He added that the switch has extra buffer credits that are dynamically allocated, and as such, the sending device does not need to throttle in response to a lack of buffer credits. Instead, the originator can use the extra credits to keep sending data, and the latency will occur at the switch. Brocade characterizes this arrangement as a benefit, and claims that SANs with 300 or more nodes can see a performance increase from this dynamic buffering. That explanation sounded plausible, but when we tested further, on uncongested ports with the rest of the switch under 100 percent load, our results didn't improve. Nothing in our testing indicated Brocade's dynamic buffering is a measurable benefit. That leaves us with an unsubstantiated vendor claim.At larger frame sizes, the SilkWorm's throughput was close to that of the QLogic SANbox2 and beat the McData switch. It ran into problems with small frame sizes, which is a concern, though not an overriding one.

We really liked Brocade's browser/Java-based configuration and management tool. It had a series of easy-to-use screens that we found intuitive enough to get simple tasks done, but also with a deeper level of complexity that let us handle even the most complex tasks in switch setup and maintenance. It was a fast, stable interface that other companies should look at.

This Brocade switch enjoys a wide range of certifications on many different pieces of SAN hardware. Brocade offers a one year warranty on its switch and on-site service and support packages to ensure data reliability. However, with a list-price range of up to $30,000, the SilkWorm falls short on bang for the buck.

Brocade SilkWorm 3800 Enterprise Fabric Switch, $18,000 to $30,000 depending on final configuration, Brocade Communication Systems, (408) 487-8000. www.brocade.com

Steven J. Schuchart Jr. covers storage and servers for Network Computing. Previously he worked as a network architect for a general retail firm, a PC and electronics technician, a computer retail store manager, and a free-lance disc jockey. Write to him at [email protected]. If anything can scare a vendor into playing nice with its competitors, it's the threat of a different technology driving the vendor into obsolescence. That's the state of the Fibre Channel market today, as companies such as McData and Brocade fend off threats from Ethernet and iSCSI with products that fairly boast of their interoperability.We tested three midrange Fibre Channel switches--2-GB units with 16 to 24 ports, selling for $25,000 or less: Brocade's SilkWorm 3800, McData's Sphereon 4500 and QLogic's SANbox2. Our tests taxed these devices by operating in full-mesh mode (every port connected to every other port), and by measuring performance under loads ranging from 25 percent to 100 percent, with sizes ranging from 60 to 2,048 frames. Although the machines all performed respectably or better, QLogic stood out with its combination of excellent throughput, reasonable latency, great features and a good price.We tested midrange Fibre Channel switches' throughput and latency, under varying loads (25 percent, 50 percent, 75 percent and 100 percent) and three different frame sizes (60-frame, 1,076-frame and 2,148-frame messages). Our test bed included a Spirent Communications SmartBits 6000B, and SmartBits 600 with SmartFabric software and Spirent's 2-gigabit dual-port Fibre Channel test cards. The SmartBits units were connected directly to the switches under test. We collected test results on a Dell Precision Workstation 410 with 512 MB of RAM and Microsoft Windows Server 2000.

We performed two full-mesh tests--that is, tests in which every port was connected to every other port. The full-mesh throughput test is a killer challenge that measures throughput, high-end latency, frame loss, latency distribution, and gives a latency snapshot. The throughput and frame-loss tests are most important. None of the switches we tested dropped a single frame under any circumstances.

The full-mesh latency test works the same as the full-mesh throughput test, but measures latency. However, because this test really measures buffer latency, the Brocade SilkWorm's extra buffering credits sharply skewed the unit's test results. The Brocade results show that we are clearly measuring frames in the buffer, as the frames have aged in the buffer.

In our port-to-port latency test, we sent 50 frames from one port to another, in pairs only. This tests basic switch latency without any significant load on the switch. Port 0 sends 50 frames to Port 1 and average latency of those frames is measured.

Finally, we created a new test, to measure latency under load and to try to verify Brocade's claims of improved switch performance. We used our SmartBits 600 to measure the port-to-port latency from Port 0 to Port 4 and from Port 8 to Port 12, while our SmartBits 6000B ran a full-mesh test on all the other ports. The result is an uncongested latency measurement while the switch is under load. This is what latency will look like on uncongested links while other links are loaded.


Fibre Channel Switches

your browser
is not Java

Welcome to

NETWORK COMPUTING's Interactive Report Card, v2. To launch it, click on the Interactive Report Card ® icon

above. The program components take a few moments to load.

Once launched, enter your own product feature weights and click the Recalc button. The Interactive Report Card ® will re-sort (and re-grade!) the products based on the new category weights you entered.

Click here for more information about our Interactive Report Card ®.

Stay informed! Sign up to get expert advice and insight delivered direct to your inbox

You May Also Like

More Insights