Whether interconnecting facilities in support of high-frequency trading, high-performance computing, business continuity, disaster recovery, migration of virtual machines among physical servers or some other demanding application, there are certain networking requirements that tend to be shared among enterprise data-center managers. Optical networking is not a commodity technology space, and data-center managers must investigate at least seven primary areas of technological differentiation in doing their due diligence to deploy the right solution for their given organization’s specific interconnection requirements:
- Latency—Critically important in electronic trading and a growing range of industries, network latency varies substantially from fiber path to fiber path and multiplexing platform to multiplexing platform. Reducing the fiber path by one physical kilometer of fiber translates into five microseconds of one-way latency that is eliminated, and Time Division Multiplexing (TDM) introduces significantly more packet delay than does Wavelength Division Multiplexing (WDM). Furthermore, significant differences in latency are generated by the way a given WDM system performs regeneration, amplification and color conversion. For example, some WDM systems use a technique for offsetting chromatic dispersion that adds additional latencies to the signal. Other systems are “carrier” based in their design and add extra overhead like Forward Error Correction (FEC) and other non-data-center-friendly transmission technologies. This is especially true for users in the metro (under 200km) that want to run “native” channels over their WDM infrastructure.
- Power efficiency—Some WDM systems consume double the energy of others, which can add up to the price paid for hardware after just three years. Furthermore, the ability to support the highest native data rates (such as 5G and 10G InfiniBand, or 8Gbit/s and soon 16G Fibre Channel, or 10G QFabrics from Juniper) enables this end-to-end power efficiency.
- Rack-space efficiency—Data-center real estate is expensive, and some WDM systems demand twice the equipment rack space of others. Search for a WDM system with a data-center Design.
- Simplicity—Some WDM systems are up and running within one day; whereas, others typically require three days or more. This adds significant OPEX cost to the solution. If it’s hard to install, suspect that it is hard to maintain.
- Investment protection—Some WDM systems available in the marketplace today still fail to support 8G Fibre Channel, and the protocol has been available since 2008. Similarly, tried-and-true, legacy protocols that continue to be relied upon across the enterprise cannot be simply abandoned. Data centers are likely to be heterogeneous protocol environments for the foreseeable future, so wide protocol support is a must for a WDM solution to shoulder interconnection.
- Security—Fiber can be tapped more inexpensively and easily than ever, so a comprehensive, modular security strategy is necessary to ensure protection for a company’s valuable information assets. Some WDM systems offer a cost-efficient capability to fully encrypt certain channels of traffic and monitor for intrusions at the physical layer. The result is a sophisticated security solution—without threatening degradation of the superior, low-latency performance for an organization’s most demanding local and storage area network (LAN and SAN) applications.
- Qualification—Consider the example that the largest storage system vendors have never accepted support calls if the WDM platform that a data center employs is not qualified specifically with that system vendor. Qualifications with all leading SAN vendors like IBM, EMC, Brocade and HDS is a must for interconnecting data centers.