Designing a data center has always been a challenge. These days the technology that we use to power and cool the data center is much better than previous generations. But because the compute and storage networks are evolving so quickly, it’s difficult to predict the network infrastructure required to sustain the rapid growth in speed/capacity that will be needed over the next few years. And when things change, you need the assurance that everything will work smoothly.
TIA 942 A standard offers guidance on the implementation of structured cabling systems that provide for scale, agility and availability of data center networks. The reasons to avoid point-to-point network topologies are well-known to infrastructure designers – network engineers aren’t always aware of the disadvantages with that standardization and structure. Certainly the operations team can point to the ongoing cost and risk that quickly mounts when these design rules are not followed. But the network wants what the network wants – speed! The budget demands restraint, and the operations team wants availability.
To answer the question, “What is the best network infrastructure to deploy?” there are some tools that will help. The objective of the physical network design is simply to support the best network alternative both now and in the future. Understanding the alternatives that are available – multimode and singlemode optics – duplex or parallel fiber paths – will certainly help ensure that the speed is available at an optimal cost. Before you move forward, you need to ask these questions:
Will the network links be long enough to support the size of the data center?
- Will the links also support the structured cabling design?
CLICK TO TWEET: It's technical, but it's assurance. See what your data center needs.
Engineering optic links consider various signal impairments and insertion and provide a specific optic power budget and physical link to support the optic application. The design can then be expressed as a practical topology (number of connections) and total link length for a given optic application. Focusing on all parts of the power budget provides a complete guaranteed channel design. Accounting for the entire loss budget including inter-symbol interference (ISI) and channel insertion loss provides solid guaranteed application performance.
ISI is caused by the interaction of the optical signal with the fiber media. ISI is a measure of signal impairment or, in other words, an increase in the noise level in the signal. The minimum received power requirement must be increased to maintain the required signal to noise ratio that will meet the target bit error rate. Figure 1 10GbE shows this imparement as 3.02db. Other optic applications will have different optical power budgets that are adjusted to compensate for these signal impairments as they change in value due to increased speeds for example.
Operating with multiple wavelengths increases data center capacity
OM5 fiber supports higher bandwidth at longer wavelengths. This additional bandwidth is being used to carry multiple optical signals/channels on the same pair of fibers. When this capability is combined with new optical modules the result is a higher capacity link that maintains the lower cost of a duplex fiber physical infrastructure.
Standardized OM5 cable enables optic module manufacturers to guarantee the reach of these wavelength division multiplexed technologies (WDM). For instance SWDM4 100G extends the supported reach least 50 percent to 150m (see graph above).
Combining SWDM4 or BiDi technology with duplex cabling can provide a significant cost saving when migrating to higher speed, switch to switch links. As a result they are fast becoming the common choice for these applications.
There’s a lot to think about. In part two, I’ll explore how to determine lowest link cost and how the CommScope Application Assurance combines all new technologies into an engineered guaranteed link design.