The old adage that history repeats itself is very true. If we don't learn from history, we are doomed to repeat it. Many data centers today are victims of historical point-to-point cabling practices.

Direct connections - "Point-to-Point" (i.e. from switches to servers, servers to storage, servers to other servers, etc.) are problematic and costly for a variety of reasons. In the best of data center ecosystems, a standards-based structured cabling system will provide functionality and scalability with the maximum available options for current and future equipment. While Top of Rack (ToR) and End of Row (EoR) equipment mounting options are now available, these should supplement, not replace, a structured cabling system. ToR and EoR equipment placement both rely heavily on point to point cables, typically fiber jumpers and either twinax copper assemblies or stranded patch cords to connect the network or storage equipment ports to servers.

Data centers are evolving in a rather cyclical manner. When data centers (the original computer rooms) were first built, computing services were provided via a mainframe (virtualized) environment. End users' dumb terminals were connected via point to point with coax or bus cabling using twinax. Enter the PC and Intel based server platforms, and new connections were needed. We have gone through several generations of possible cabling choices: coax (thicknet, thin net), category 3, 4, 5, 5e, 6. Now, the recommended 10 Gigabit capable copper choices for a data center are category 6A, 7 and 7A channels, OM3 grade fiber for multimode capable electronics and single mode fiber for longer range electronics. In some data centers, samples of each of these systems can still be found under the raised floor or in overhead pathways, many of which originally were point-to-point. Today however, the "from" point and "to" point are a mystery, making cable abatement (removal of abandoned cable) problematic at best. Compounding this problem was a lack of naming conventions. If the cables were labeled at both ends, the labeling may not make sense anymore. For instance, a cable may be labeled "Unix Row, Cabinet 1."

Years later, the Unix row may have been replaced and new personnel may not know where the Unix row was. There are two standards for structured cabling systems in a data center: TIA 942 and draft ISO 24764, the latter of which is slated to publish in September, 2009.

These standards were created out of need. Both data center standards have language stating that cabling should be installed to accommodate growth over the life of the data center. Moves, adds and changes for a single or a few runs are expensive compared to the same channels run as part of an overall multi-channel installation project. For the larger projects, the end user realizes benefits from project pricing, economies of scale, and lower labor rates per channel. Single channels are typically more expensive, as it is more expensive to send personnel to run one channel. The risk of downtime increases with continual moves, adds and changes. Pathways and spaces can be properly planned and sized up front, but can become unruly and overfilled with additional channels being added on a regular basis.

Data centers that have issues with cable plant pathways typically suffer from poor planning. Growth and new channels were added out of need without regard to pathways. In some cases, pathways do not accommodate growth or maximum capacity over the life of the data center. Overfilled pathways cause problems with airflow, and in some cases cabling becomes deformed due to the weight load, which can adversely affect transmission properties of the channel. This is particularly true in point-to-point systems that have grown into spaghetti-like conditions over time. Likewise, data centers that have not practiced cable abatement or removal of old cabling as newer, higher performing systems are installed experience the same disheveled pathways.

Figure1. Depicts a ToR patching scenario between switch ports and servers without a structured cabling system. Rack 2 to Rack 3 connections are indicative of point-to-point serverto-switch connections, also without a structured system. While proponents of these systems tout a decrease in cabling as a cost offset, further examination may negate such savings.

If a central KVM switch is used, the centralized structured cabling system would need to co-exist anyway, albeit with less channels day one. Newer electronics may have different channel minimum/maximum lengths resulting in the need for new channels. As electronics progress, the structured system may need to be added back to the data center to support future equipment choices, completely negating the savings.

It will cost more to add the structured system later as pathways, spaces, and channels were not planned for and must be installed in a live environment increasing labor costs and the likelihood of downtime. When adding pathways and spaces, fire suppression systems and lighting may need to be moved to accommodate added overhead pathway systems. Floor voids may need to be increased and cabinets may need to be moved to allow new pathways to be routed in a non-obstructive manner for proper airflow.

Further examination highlights other disadvantages of ToR and Point-to-Point methodologies beyond the limitations outlined previously. In either the Rack 1 or Rack 2 -> Rack 3 scenario above, switch ports are dedicated to servers within a particular cabinet. This can lead to an oversubscription of ports. Suppose rack/cabinet 1 had the need for only 26 server connections for the entire rack. If a 48 port switch (ToR switching) or 48 port blade (point-to-point server to switch) is dedicated to the cabinet, this means that 22 additional ports are purchased and maintenance is being paid on those unused ports.

A greater problem occurs when the full 48 ports are used. Adding even one new server will require the purchase of another 48 port switch. In this case, assuming two network connections for the new server, an oversubscription of 46 ports will be added to the cabinet. Even in an idle state, these excess ports consume power. Two power supplies are added to the cabinet. Active maintenance and warranty costs are also associated with the additional switch and ports.

Many of these ToR technologies have limitations for cabling length. Maximum lengths range from 2-15m and are more expensive than a structured cabling channel. Short channel lengths may limit locations of equipment within the shorter cable range. With a structured cabling system, 10GBASE-T can be supported up to 100 meters of category 6A, 7 and 7A cabling and allows more options for equipment placement within the data center.

Any-to-All Structured Cabling System

The concept behind any-to-all is quite simple. Copper and fiber panels are installed in each cabinet which correspond to copper patch panels installed in a central patching area. All fiber is run to one section of cabinets/racks in that same central patching area. This allows any equipment to be installed and connected to any other piece of equipment via either a copper patch cord or a fiber jumper. The fixed portion of the channel remains unchanged. Pathways and spaces are planned up front to properly accommodate the cabling. While tthis method may require more cabling up front, it has significant advantages over the life of the data center. These channels are passive and carry no reoccurring maintenance costs as realized with the addition of active electronics. If planned properly, structured cabling systems will last at least 10 years,supporting 2 or 3 generations of active electronics. The additional equipment needed for a point-to-point system will require replacement/upgrade multiple times before the structured cabling system needs to be replaced. The equipment replacement costs, not including ongoing maintenance fees, will negate any up front savings from using less cabling in a point-to-point system.

The red lines (fiber connections) all arrive in the central patching area in one location. This allows any piece of equipment requiring a fiber connection to be connected to any other fiber equipment port. For instance, if a cabinet has a switch that requires a fiber connection for a SAN on day one, but needs to be changed to fiber switch connection at a later date, all that is required to connect the two ports is a fiber jumper change in the central patching area. The same is true for copper, although some data centers zone copper connections into smaller zones by function, or based on copper length and pathway requirements. As with the fiber, any copper port can be connected to any other copper port in the central patching area or within the zone.

Cabling standards are written to support 2-3 generations of active electronics. An "any-to-all" configuration assures that the fixed portion of the channels is run once and remains highly unchanged if higher performing fiber and copper cabling plants are used. As a result, there will be less contractor visits to the site for MAC work as the channels already exist. Faster deployment times for equipment will be realized as no new cabling channels have to be run. They are simply connected via a patch cord. Predefined pathways and spaces will not impact cooling airflow or become overfilled as they can be properly sized for the cabling installed. Bearing in mind that the standards recommend installation of cabling accommodating growth, not only will day-one connectivity needs be supported, but also anticipated future connectivity growth needs are already accounted for.

With central patching, switch ports are not dedicated to cabinets that may not require them; therefore, active ports can be fully utilized as any port can be connected to any other port in the central patching area. Administration and documentation are enhanced as the patch panels are labeled (according to the standards) with the location at the opposite end of the channel. Patch cords and jumpers are easy to manage in cabinets rendering a more aesthetically pleasing appearance as cabinets will be tidier. In contrast, with point-to-point cabling, labeling is limited to a label attached to the end of a cable assembly.

With a structured high performing copper and fiber cabling infrastructure, recycling of cabling is minimized as several generations of electronics can utilize the same channels. Being able to utilize all switch ports lowers the number of switches and power supplies. All of these help contribute to green factors for a data center.

To further explain the power supply and switch port impact, contrasting the point-to -point, ToR scenario in section 1, in an "any-to-all" scenario, the 48 ports that would normally be dedicated to a single cabinet (ToR) can now be divided up, on demand, to any of several cabinets via the central patching area. Where autonomous LAN segments are required, VLANs or address segmentation can be used to block visibility to other segments.