Data centers are nerve centers of enterprise IT. Businesses therefore pay a lot of attention to investments in enterprise application software, high-performance and highly redundant computing, storage, communications, security (physical as well as network), power and cooling infrastructure, and skilled and specialized maintenance teams to support data center uptime and performance.
A smaller cost of the overall data center investment is LAN cabling systems – the essential foundation for successful application delivery within the data center. They can have a profound impact on the network throughput, packet losses and delays on the network if adequate thought is not paid to the choice of media and architecture while designing the data center.
Fundamental and profound changes are happening in the way applications are being deployed.
- Servers and storage are becoming increasingly virtualized, and that is creating newer traffic end points on the network. This is leading to increased traffic within the modern data center between servers (East-West), as well as against traffic between servers and clients (North-South) in conventional data centers.
- Cloud computing applications are creating dynamic and multi tenanted workloads.
- Ethernet is fast emerging as the common convergent protocol to support application and storage traffic. Traditional network architectures – core, distribution and access – cannot optimally service these newer traffic patterns. Data center switch fabric architectures (such as spine-leaf) are being proposed to flatten the network for lower latencies and enable connectivity within the data center.
LAN cabling connects the end and intermediate points of the traffic patterns. It connects the servers, storage, network switches and traffic access points (TAPs). They need to be adequately sized to ensure faster speeds and feeds that may be required at these end points. While arriving at cabling system design, adequate understanding of how the server, storage, switch, TAPs and rack architectures are aligned to meet successful application delivery objectives is a must.
Cabling standards have been evolving at a fast pace recently and offer insights into how network speeds are likely to evolve in the future. Ethernet speeds on fiber have been standardized on optical fiber up to 100G, and work in standards is underway to support 400G. OM5 (WideBand multimode fiber) fibers are on the horizon, intended to be backward compatible to OM4 fibers today and yet support emerging SWDM technologies and offer capacity expansion once SWDM switch interfaces are available. On copper, Category 8 cabling system standards have been ratified and are intended to serve 40G – primarily in data centers up to 30m.
Data center cabling design is an important piece of the network and should align with rack, network switching and TAPs, storage and application architectures. A top-down approach to cabling design brings a lot of clarity to data center network design and ensures that applications are best supported today and in the future.