As data centers continue to expand and internal and external customers continue to demand more capacity, better efficiencies, and higher speeds, data center managers are installing new, faster capacity throughout their network infrastructures.
“One of the things that we are starting to see is that data centers are building for what is coming down the road,” says Mark A. Gilmore, president of Wired Integrations (www.wiredint.com). Here is a look at some of the network infrastructure trends that SMEs are looking at to help increase their networking capacity.
Fiber vs. Ethernet
Some data centers are ensuring top speeds by using blazing fast fiber connections, but most are still using CAT 6 Ethernet connections and a high-speed backbone to all data center devices, Gilmore says. “Fiber is going to cost them twice as much, so most of them are fairly resistant to [installing fiber] right now,” Gilmore says. “As the technology continues to advance, the throughput will continue to grow.”
• Data centers are adding faster components throughout their infrastructures to meet the demand for ever higher throughput.
• Although data centers are trying to increase speed and capacity, they are also trying to simplify the infrastructure wherever possible.
• Security is an essential component of any data center and can add unneeded costs and latency if not done properly.
Fiber is most useful for any long connections needed in the data center, according to Tara Van Unen, senior manager of market development for Ixia (www.ixiacom.com). “We’re seeing more of a push for service-oriented platforms and an increasing amount of on-demand services, so data centers need applications that are accessible at any time. To do this, they need to have large enough pipes.”
Van Unen adds that Ethernet connections, while less costly than fiber, suffer from noise and signal degradation over longer distances. Additional power is needed to boost the Ethernet signal to attempt to compensate for these problems.
To combat these problems, many data centers are using a combination of Fibre Channel and Ethernet connections by moving to FCoE (Fibre Channel over Ethernet), which enables them to combine the two types of connections in a single Ethernet network. Combining the connections lowers the cost of running the two connections over their own dedicated networks, which would require multiple interfaces.
“The main goal is to provide consolidation for the first hop to the core switch,” Van Unen says. “FCoE simplifies the data center architecture while reducing costs and increasing bandwidth capacity.”
Newer, Faster Switches
Another way that throughput is growing is through 100Gb switches replacing 40Gb switches for the core of the network, says Ray Suarez, director of product management for Enterasys (www.enterasys.com).
Though there are still many data centers with 10Gb switches in the core, Suarez recommends that any new switches for the middle or end of the row be 40Gb or 100Gb and have modular designs that make for easy upgrades. Upgrading other switches throughout the network can also help provide better speed. However, Suarez says that although the more robust switches provide benefits of higher speeds, there isn’t enough ROI to undertake it unless it is part of an expansion or server refresh project, which most data centers undertake every three years or less.
Some vendors are building more proprietary elements into their routers, switches, and other devices, meaning they are attempting to lock buyers into that vendor’s set of products rather than being able to use a best-of-breed approach. Suarez recommends that managers use infrastructure devices and components that enable the data center to operate in a heterogeneous environment.
Although data centers are expanding, there are some areas where the infrastructures are becoming less complex. Many of the newest technology trends are designed to simplify the management of the data center network infrastructure.
A few years ago, it was commonplace for three to four cable drops to accompany each workspace, according to Gilmore. But with the widespread adoption of VoIP connections, one or two cable drops are more common because the additional ones are no longer needed.
Similarly, many of today’s server racks have three sets of switches, which data center managers can consolidate into two, says Asas Somekh, vice president of marketing for Voltaire (www.voltaire.com).
By using larger-capacity switches, data centers can eliminate one set of switches throughout the data center, simplify design, and minimize points of failure, Somekh says.
Additionally, using switches with large enough capacity and with enough open ports eliminates bottlenecks at the switches, which Somekh cites as a major problem in some data centers today.
Just as data centers have handled additional demands for processing power by adding servers, most have met the need for additional network security by adding different devices to handle increasing security threats, says Howard Ting, director of marketing for Palo Alto Networks (www.paloaltonetworks.com).
However, just as the complexity of the network infrastructure grew from additional switches, the same is true of the additional security devices, Ting says. “As threats to the network evolved, rather than solving the problem by adding threat protection to a device they already had, vendors have solved the problem by selling new point solutions,” he continues. “Some data centers were running as many as six to 12 additional devices, meaning a lot of additional costs.”
In addition to the cost of the devices themselves, there are also additional power and cooling costs for each. A better idea, and one that is more cost-effective for data centers, is to offer all of the necessary firewalling and other security solutions in a single device, according to Ting. But whether on one device or on many, the security solutions have to work in such a way that they don’t slow down the speed of the network, he adds.
Ting also recommends working with a security vendor that will keep the data center abreast of the newest and latest security threats.
by Phil Britt
Top Technology: Network Management Software |
As data centers continue to grow, it is increasingly imperative to be able to quickly diagnose and solve problems with switches, connections, or other “physical fabric,” says Asas Somekh, vice president of marketing for Voltaire (www.voltaire.com). “People need to be able to manage the whole fabric of the data center, not just individual switches.”
Software to manage and troubleshoot such an infrastructure must be able to work in a multivendor environment, Somekh adds, because as data centers expand, they will invariably have legacy and newer equipment, so they are likely to have a combination of products from multiple vendors. So the application must be able to find problems such as switches or connections that are not working properly and identify actual or potential bottlenecks in the data center.
“For example, if there is latency between two servers, you need to know that so that you locate them close together [to keep latency to a minimum], rather than having them on opposite sides of the data center,” Somekh says.