Inside the numbers: 10-GbE cost comparisons

by Patrick McLaughlin

This vendor roundtable discussion provides perspective on the pricing issues associated with different 10-Gigabit Ethernet media choices.

As a technical undertaking, 10GBase-T is an accomplishment of significant proportions. The engineering efforts required to create a protocol for transmitting 10 billion bits of data per second on eight copper wires are worthy of celebration and marvel.

For professionals in the cabling industry, however, it is doubtful that those contemplating 10-Gigabit Ethernet deployment have used the word “marvelous” to describe the deliberations and considerations that have gone into selecting which flavor (10GBase-T or 10GBase-SX) to deploy. And some users who have settled on 10GBase-T have found the option of using unshielded or shielded twisted-pair cabling a source of consternation.

In the 10-GbE ecosystem, technical issues raise cost questions, which give reason for deeper consideration of the financial investments involved. Logical questions arising from cost analysis involve the long-term technical benefits of deploying 10GbE; the cycle completes itself and starts over again.

Cabling-system vendors provide full assurance that their twisted-pair solutions accommodate 10GBase-T transmission-even though the Telecommunications Industry Association’s (TIA; www.tiaonline.org) full set of Category 6A specifications is not yet finalized. With the first 10GBase-T-compliant networking products hitting the market, these cabling vendors are getting their first opportunities to prove those claims.

Meanwhile, fiber-based 10-GbE systems have been available and deployed for some time. The truism we have all heard is: even with the use of vertical-cavity surface-emitting lasers, the cost of optoelectronics equipment (i.e., fiber-based networking equipment) so far exceeds that of copper-based electronics, holding out for the introduction of copper-based systems is the most cost-effective move a user can make. And this author, for one, typically takes that statement at face value.

Just how much of a value is 10GBase-T over 10GBase-SX? And even if 10GBase-T proves to be the financially advantageous option, how should a user make the decision about whether to operate the protocol on shielded or unshielded cabling?

With those questions lingering, or put another way, fueling the 10-GbE ecosystem, Cabling Installation & Maintenance sought out three cabling-system providers, each of which offers multimode fiber, unshielded twisted-pair, and shielded twisted-pair solutions. With some vendors rather adamantly advocating one medium over the others for 10-GbE transmission, we invited ADC (www.adc.com); CommScope (www.commscope.com), which offers the Uniprise and Systimax brands; and Tyco Electronics/AMP Netconnect (www.ampnetconnect.com) to participate in a Q&A-style discussion about the technical and, significantly, the financial considerations that users face when 10GbE is in their future.

Speaking for ADC is John Schmidt, senior product manager for structured cabling. Representing CommScope is Matt Brown, global data center solutions manager. Brian Davis, global product manager, represents Tyco Electronics.

(Note that in providing these individuals’ answers to the questions posed, they had the option of answering or not answering any question. Also, when more than one individual answered a question the same or essentially the same, Cabling Installation & Maintenance has chosen to provide a single response.)

Q:On a “1x/1.5x/2x”-type basis, generally what are the installed-cost figures for 100-meter, 4-connector channel Category 6A UTP, Category 6A foiled/unshielded twisted-pair (F/UTP), and laser-optimized 50-µm multimode fiber cabling systems?

Davis: Category 6A F/UTP and multimode fiber are 1x; Category 6A UTP is 1.1x.

Brown: Fiber-cable prices have remained stable during the past few years, while copper raw material prices have driven the cost of UTP and F/UTP cabling up. Based on UTP as 1x, F/UTP installed links will be 1.15x, and laser-optimized fiber links will be 1.75x.

Q:Can you provide similar information about the cost of 1-Gig network equipment (network interface cards [NICs], LAN cards), comparing 1000Base-T with 1000Base-SX?

Schmidt: According to Intel pricing, 1000Base-T NICs are $145 while 1000Base-SX NICs are $510. And Cisco prices 1000Base-T small-form-pluggables (SFPs) at $315 while 1000Base-SX SFPs are $380. On the NIC side, 1000Base-SX will be approximately 3.5 more expensive than 1000Base-T. Keep in mind that 1000Base-T NIC ports have astronomically higher volume than 1000Base-SX. For switch ports, which have higher 1000Base-SX volume, the cost is much closer, with 1000Base-SX having only a 20% premium over 1000Base-T ports.

Brown: 1000Base-T and 1000Base-SX NICs have both fallen dramatically over the past two years. They are between 20% and 35% of their 2005 price. The price gap between -T and -SX has narrowed from 4x in 2005 to 3x today.

Q: Do you have any insight into what the market might expect, in terms of a cost difference between -T and -SX, when a full complement of 10GBase-T networking equipment is available?

Brown: Estimates at IEEE have historically predicted 10GBase-T prices will be 40% of 10GBase-SX prices. We believe the biggest hurdle for 10GBase-T is the electronics power requirement. The most efficient NICs on the market have power consumptions of <6 watts, while fiber-based 10G NICs are readily available between 2 and 4 watts.

Schmidt: Two NIC vendors, Tehuti Networks and Chelsio Communications, have announced product. Chelsio’s 10GBase-T NIC is priced at $1,995. Neterion’s 10GBase-CX4 NIC is $1,095, while its 10GBase-SR NIC is $1,995 and its 10GBase-LR NIC is $2,895. 10GBase-T NICs are currently about the same price as 10GBase-SR, as they have just been introduced and have low volume. It is expected that by 2009 the relative cost between 10GBase-SR and 10GBase-T will be 4x, with 10GBase-T NICs sub-$200 and 10GBase-SR around $800.

Q:Given the economics of optical networking, users may be tempted to dismiss it out of hand for channels of 100-meter distances or less. But are there circumstances under which fiber can be the best overall choice in systems ≤100 meters?

Davis: There are several reasons to deploy fiber in horizontal links less than 100 meters, including the following seven scenarios:

1. Cabling runs through or is exposed to high radio-frequency interference/electromagnetic interference (RFI/EMI) noise areas;
2. Cabling runs through or is exposed to high-voltage areas-there are no induced voltages on all-dielectric fiber cable;
3. Restricted pathway space, including but not limited to small conduits, small bulkhead openings, and fire barriers;
4. Cabling runs through or is exposed to high-temperature areas;
5. 10-GbE-capable runs are needed today;
6. Secure areas with concerns about transmitted/radiated signals;
7. Limited termination space for highest-density connectivity.

Brown: Fiber has its strongest play when density is a major concern, as in the storage area network (SAN) environment. This relatively short-distance application is dominated by fiber cabling due to density concerns as well as the potential for fiber to upgrade to 100-gigabit and beyond. Fiber makes sense below 100 meters when density and a clear upgrade path to next-generation speeds are major concerns.

Schmidt: Absolutely. In particular, laser-optimized multimode fiber is an ideal medium for shorter-distance transmission for the following situations:

1. Fibre Channel transmission in SANs;
2. Highly secure network connections, due to fiber’s high degree of difficulty to tap or monitor;
3. Very high density network connections that require small cable diameters;
4. Sub-100-meter lengths that are expected to be 100-Gigabit Ethernet in the future.

4. Sub-100-meter lengths that are expected to be 100-Gigabit Ethernet in the future.

It is expected that 100-Gigabit Ethernet will have a 100-meter distance limitation on laser-optimized multimode fiber.

Q: When planning 10GBase-T-capable twisted-pair systems, end users face complicated decisions about cable types and those types’ characteristics. Assuming we are talking about a 4-connector channel in a “friendly” environment (no excessive external noise sources or extreme bends along the pathway), does your organization recommend one medium over another?

Schmidt: Unshielded twisted-pair Augmented Category 6, for the following reasons:

1. UTP Category 6A is available in the smallest diameter that will support 10GBase-T-0.275 in. versus 0.285 in. in F/UTP cables-which will improve conduit-fill ratios by allowing more cables to fit in a given-size conduit;
2. UTP Category 6A is the lowest-cost medium that will support 10GBase-T, accounting for all factors including component cost, installation cost, troubleshooting, and testing;
3. UTP Category 6A is designed to meet and exceed all the electrical requirements set forth by the IEEE, ISO, and TIA to support 10GBase-T without resorting to shielding;
4. In 2006, more than 6.2 billion feet of category-rated cable was installed in the United States; of that, less than 1.5% was F/UTP, according to market-research firm BSRIA (www.bsria.co.uk). As a result, there is very little installation, troubleshooting, and maintenance experience with F/UTP cabling in the United States-precisely why the vast majority will continue to use UTP cabling.

F/UTP and S/FTP cabling certainly has its place within the network. It will provide superior external-noise suppression for installations near high noise sources, such as radio or microwave transmitters; however, standard noise from common sources, and certainly alien crosstalk, can be more than adequately eliminated by UTP Category 6A. So, for 98.5% of installations in the United States, UTP cabling is going to be a more economical choice that does not sacrifice performance.

Brown: We recommend TIA Category 6A, ISO Class EA-compliant UTP cable. F/UTP solutions have several drawbacks beyond their 15% price premium. When using UTP cabling, the installer does not need to implement the additional bonding and grounding steps that are required with F/UTP cables. There is additional time and cost associated with terminating the shield and drain wire on F/UTP cabling. Due to the sensitivity of this operation to installer technique, between 5% and 20% of terminations need to be re-worked to properly ground them. The possibility of the foil tape folding or kinking can give F/UTP cables a larger bend radius than UTP cables. F/UTP cables typically have 2-inch bend radii (similar to coaxial cables), compared to typical UTP bend radii of one inch. Ease of cable routing and dressing are critical in today’s dense connectivity environments, such as the data center. While F/UTP cables are smaller than some UTP cables, the density advantages of F/UTP are small and do not justify the increased cost and difficulty of installation. If density is a critical concern, then fiber is preferred.

PATRICK McLAUGHLIN is chief editor of Cabling Installation & Maintenance.

---

Cost model lets you make your own calculations

The TIA’s Fiber Optics LAN Section (FOLS; www.fols.org) provides on its Web site a cost model that lets users compare the costs associated with deploying the cabling infrastructure and network hardware for protocols, including Gigabit and 10-Gigabit Ethernet.

Cost-model users must register before beginning, but the registration and use of the model is free. FOLS has updated the pricing data in its cost model periodically, and the model lets users input their own numbers as well. The participants in this article-ADC, CommScope, and Tyco Electronics-are all FOLS members.

“FOLS continues to be committed to keeping our cost model current,” said Andrew Oliviero, FOLS chair, when the model was most recently updated in February. “For this update, we made changes that will make it easier for users to make accurate comparisons among architectures, essentially updating the data without making significant changes to the structure of the model.” The newest updates include new media choices, aggregate pricing reflecting current market conditions, the ability to customize port-utilization calculations, and a graphic comparison of different architectures.