Calculating costs of data center cabling infrastructure

Part one of our series on data center cabling infrastructure decisions outlined the available options. In part two we are going to develop an equation to determine the total cost of ownership (TCO) on a data center cabling project.

Putting this to dollars and cents

If we contrast the end to end system costs for copper versus fiber channels including the cost of electronics, we can use the following formula:

(Day one costs + day two costs) / number of years of useful life = annualized costs

In the equation above, day one costs include equipment and channels, plus installation including facilities. If there is exiting equipment and/or abandoned cabling that must be removed to start the installation these will be included. Include any time and costs to add additional cards to equipment that do not come preinstalled (for instance, adding a fiber card to a server). Any downtime needed to complete the transition will also be added to the day one sum. By performing these calculations based on channel type and electronics type you will get an overall calculation that will make infrastructure decisions easier. For example:

Server ABC + fiber cards + 4 channels + installation + facilities + 4 switch ports = Day One Costs

Yearly day two costs include additional maintenance costs, power consumption, cooling and upgrades. Add the scheduled downtime to all upgrade costs. This can be a very critical factor between one system that requires upgrades and patches each quarter to one that requires them once per month or even biweekly.

Maintenance + power + cooling + (upgrade cost * number required) + downtime costs = Day two costs

In a new data center, with a clean slate, your manufacturers will be your best source of information. It is always a good idea to discuss a similar installation with someone that has already done one, preferably a year ago, so that you can get an idea of the day two costs for similar equipment. One fact remains in new, consolidation, relocation or upgrade environments. It is in your best interest to try to install all channels accommodating growth so that these spaces do not have to be revisited as stipulated in TIA-942. This provides ease of use in adding new equipment as well as a protection for existing channels that could be altered with moves, additions and changes. It also allows for a clean under the floor or overhead environment without any abandoned cables that contribute to fuel load in the event of a fire, which can also keep air handlers from performing at optimal levels.

When consolidating or moving, it is a rather simple matter to count all ports that exist and are in use and add a growth factor to that number. A general rule of thumb used in many data centers is to provide 48 ports of copper and 12 strands of multimode fiber to each cabinet as a start. However, switch cabinets and cabinets with rack-mounted servers may require significantly more. But counting what is in place will certainly provide a starting point. If there is no existing equipment or none that will move, a bit of foresight is required as in a new installation. Consider the channels, the switches, routers, servers, PBX or IP PBX systems and any network attached storage or storage area networks. KVM switches and any other monitoring will require connectivity. Take into consideration both in-band and out-of-band management.

Data center cabling cost calculator

For new equipment and even when considering retrofitting older equipment with replacement fiber cards or other capabilities, an end to end planning guide will help. To put this into practical dollars and cents, consider the following worksheet which utilizes a per port/per server model. In each row the costs for any and all active components required to connect a server are considered. If you have 4 switch ports necessary than 4 channels will be used for the calculations. Use these calculations for any channel in question (copper versus fiber) and the equipment that will run over the channel including switches for network and storage connections. Some may be a mix of copper and fiber where fiber attached storage is in use.

Options can be compared based on all media types. Remember that your cabling channels, either copper or fiber will remain in place ideally for 10 to 15 years supporting 3-4 iterations of active electronics and 2 or 3 versions of applications.

In column one, list the installed cost of all cabling channels that you are considering for any piece(s) of equipment. If you are using trunking cables, the installation costs will be lower. If you are removing abandoned cable, include those costs. If you will need to modify pathways and spaces, include these amounts, as well. All connections to your bonding grid will be added, as these grids are generally required whether you install a shielded system or need them only for electronics, pathway and floor connections. If you are adding connections for a shielded system, divide the cost for that connection by the number of ports sharing a common ground bar. With the new shielded connectors that self terminate the shield and share a common connection, this will only require one connection per 24 ports, so this cost may be negligible enough not to be included.

In column two, determine the port costs for your equipment including installation, configuration and testing. If you are planning a chassis based switch or blade servers, evenly divide the chassis costs by the number of supported ports and add that cost to the per port costs. This can be significant over each port. For instance, on average from several manufacturers' prices for a 48 port switch, a copper 1 gigabit port is roughly $197.92 per port in a non chassis unit. The blade only for a chassis unit averages about $208.33 per gigabit port on the card.

The cost of the chassis, and all management modules, redundant power supplies and any other shared portion should be added and divided by the number of ports for each port that will touch a channel. This sum should be added to the port cost. In a chassis based environment per blade costs may be more expensive, but not having to increase real estate and provide more connections per port may be more attractive. You can assign a weighting factor for these decisions later.

The same ports on fiber average around $875.00 each. On the fiber side, however, make sure to consider the density of the blades. Some do not provide the same density and may require additional rack space or real estate to provide the same number of connections. This will come later in the weight factor column.

((Chassis cost + management module + redundant supplies) / total ports) + per port blade cost = total active equipment port cost

Day two maintenance costs include your maintenance and optionally your power consumption. Remembering that maintenance costs may be based on original purchase pricing this can be a hefty factor in the decisions. If you have a factor to address cooling costs you will also want to add that factor. Also add expected costs for upgrades to software. For upgrades and patches, if you have 24 rack-mounted units compared to four chassis based units, these costs will likely be less for the chassis based unit. Consider the time to change from one machine to the other, etc. While this is a projected number, consulting with current help desk reports for the data center or other data center managers can give you enough to use for your estimates.

Add these three columns together.

Channel costs + day one costs + (maintenance costs * number of years of service) = total per port server cost*

*This is going to be your total cost for equipment and channels in column four.

Assuming that you want a 10-20 year lifecycle, determine what would have to change in that period to support your current hardware and the next generation or two you want to install. For instance, if you have category 5e cabling, it will have to be replaced. If you have the older 62.5 micron fiber installed, it may have to be replaced. There are electronics that work to support older installed systems, but they are generally more expensive both initially and for maintenance.

Projected downtime in column eight will be the hardest to determine. Costs for that downtime may be a bit of a guessing game. One way to figure downtime is to divide the revenue by the number of users and the number of hours worked per year. This allows you to determine potential revenue impact. Then divide the average salary that has been weighted to include taxes and overhead by that same number of hours worked. This tells you how much lost salary expenses are being incurred while someone is being paid but are not productive. These two sums added together will give you downtime per user per hour. Multiply this by the number of users that use a server and you will have a basis for figuring downtime. Multiply this hourly number by any time needed to apply patches to an operating system, apply firmware upgrades, etc. If you will have downtime for cabling and for electronics factor the hourly figure accordingly. If you are in compliance related industry or have shared spaces between various unions and must pay additional personnel to accompany or supervise these workers, add these to your downtime calculations.

See the example below, based on equal replacement of equipment, the larger factors being replacement of channels and increased maintenance costs have a significant impact on a data center over the 20 years used in the example. Augmented category 6 and category 7/class F channels with equipment are the most cost effective factors.

An optional weighting factor can be included in the last column to allow you to weigh the different options based on your specific criteria such as floor space required, additional rack space, or even the order of importance. If you add the annualized costs, plus yearly costs for downtime, and multiply by your risk factor and weighting factor, you can determine what makes the most sense for each part of the data center. This small amount of planning, or large depending on the size of your data center can help you to make the right decision.

Trunking cables can bring additional benefit due to their construction. Multiple channels are pulled at one time as they are grouped together. They are factory terminated and are just popped into a patch panel in the cabinet or connected via fiber plug and play modules. This lowers your installation costs, and assures that all terminations are factory quality. Wire management for the channels lengths is easier due to their bulk construction. For upgrades and removal of abandoned cable, these cables are easy to identify as new runs so that older runs are distinguishable.

Running additional channels or changing out channels can be much more expensive than having them installed at the start of the project. The increased risk of downtime and harming what is already installed can also be serious. As you complete this exercise, you will see that in many cases, higher performing systems can mean the difference between penny wise and dollar foolish. For an additional tool to assist in your decisions, contact your Siemon sales representative for assistance with the interactive whitepaper that will help populate many of these columns based on your situation, number of drops and local labor rates for installation and downtime calculations based on your criteria.

About the author: As the Global Network Applications Market at The Siemon Company, Carrie supports the end-user and electronics communities. She has won the "Communication News" Editor's Choice Award for the last two years. See Carrie's Ask The Experts page for more of her articles.