Although 40 and 100 Gigabit Ethernet represent the industry's cutting edge, the vast majority of enterprises are...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
still in the midst of a transition from Gigabit Ethernet to 10 Gigabit Ethernet (GbE) in their networks. While 10 GbE has been broadly available for years, it is still new to most network engineers. With that in mind, we are busting some of the most common myths surrounding the technology in this 10 GbE edition of myth versus reality.
Myth: Deploying 10 Gigabit Ethernet is too expensive.
Reality: While the price per port of 10 GbE is still higher than Gigabit Ethernet, the gap between the two is narrowing. In high-traffic areas of the network, it might actually be more cost effective to use a single 10 GbE port rather than ten Gigabit links to achieve the same level of bandwidth and performance.
What may really work: Mike Spanbauer, principal analyst for enterprise networking and data center technology at Current Analysis, suggests that enterprises evaluate the entire cost of an access port, including power consumption and port density. In data centers where power, cooling and physical space are a premium, 10 GbE may actually be the more cost-effective option.
Myth: 10 Gigabit Ethernet only works on expensive optical cables and transceivers.
Reality: Engineers have a range of cabling options for 10 GbE, with more on the way. While optical is the primary choice for long-range links between switches in the data center, enterprises can choose lower-cost optical cabling or twinax copper cabling for shorter range or to interconnect within the server rack. Also, many server and switch vendors are preparing to launch products with 10GBase-T, which will offer the convenience of the more common twisted pair cabling with RJ45 connectors for distances up to 100 meters.
What may really work: Cabling data center networks can be an expensive, time-consuming proposition, so data center managers should look into all of the available media choices to maximize the life span of the physical network. Likewise, many data center managers prefer to standardize on a single, small form-factor pluggable (SPF+) transceiver module and cabling type, according to Spanbauer. So while a mix and match approach to cabling may offer short-term cost savings, installing fiber optic might offer greater investment protection to support 10 GbE and beyond.
Myth: Enterprises will have to completely rip and replace switches or do forklift upgrades to support 10 Gigabit Ethernet.
Reality: Many enterprises will discover that the modular switches running Gigabit Ethernet in their networks are capable of supporting 10 GbE line cards. Also, many new 10 GbE fixed configuration switches fit into the same footprint as existing, racked Gigabit Ethernet switches.
What may really work: The need for the additional bandwidth that comes with 10 GbE may quickly spread beyond the data center. A number of 802.11n wireless access points could fill a Gigabit Ethernet uplink, necessitating 10 GbE links to aggregate the wireless LAN traffic, Spanbauer said. A network manager should consider 10 Gigabit options throughout the entire enterprise network.
Myth: 10 Gigabit Ethernet is a requirement for building a data center fabric.
Reality: Many networking vendors offer data center fabric switches in both Gigabit and 10 GbE port options. These are offered as a low-cost option, and data center managers can take advantage of the any-to-any connectivity and low-latency functionality that a data center offers to existing servers that lack 10 GbE adapters.
What may really work: While data center fabrics offer support for Gigabit Ethernet, most network administrators will need 10 GbE to support a virtualized data center, Spanbauer said. The effect of running multiple virtual servers on a single physical machine is an increase in network utilization on every network port. Without sufficient bandwidth, the network might become the bottleneck in a highly optimized data center.
Myth: The only reason to move to 10 Gigabit Ethernet is as part of a network and storage convergence project.
Reality: Prior to the rise of server virtualization and data center fabrics, many networking vendors made a significant marketing push for network convergence. Since the physical link is shared by IP and storage networks must have high bandwidth capacity, deploying technologies such as Fibre Channel over Ethernet (FCoE) or iSCSI will demand 10 Gigabit Ethernet.
What may really work: “Network convergence is one of many drivers for enterprises to make the case for 10 Gigabit Ethernet,” said Spanbauer. Network convergence, virtualization and even high-capacity WAN links are all demanding more bandwidth than Gigabit Ethernet can provide.
Myth: Isn’t 40 Gigabit or 100 Gigabit coming soon? Why would anyone bother with 10 Gigabit Ethernet?
Reality: Switches that support 40 and 100 Gigabit Ethernet are just beginning to come to market. However, many of these products are proprietary or based on draft standards. While network switch vendors are pledging to provide an upgrade path from their current offerings and offer future support for the ratified standard, those vendors that do have 40 GbE on the market use it exclusively for interlinking between their own switches. Also, the price per port for 40 and 100 Gigabit Ethernet will remain extremely high until the technology is more broadly adopted.
What may really work: Network administrators must continually weigh the cost of bleeding-edge technologies against the benefits of the technology. Very high-end data centers may require the bandwidth that 40 or 100 Gigabit Ethernet provides, but like 10 GbE a few years ago, the majority of enterprise networks will wait some time before the price and standards barriers of these technologies are low enough to adopt.