How smart cabling choices can future-proof infrastructure investments

We’re already seeing the increasingly rapid rise of 10GBASE-T and 10 Gigabit will be the most frequently used Ethernet protocol within the next few years. The most important standards I see emerging today are IEEE 802.3bm*, 802.3bq** and 802.3bj***. Furthermore, the 400 Gb/s Ethernet Study Group are doing some important work, and there’s also the ISO/IEC 11801-99-x Guidance for balanced cabling in support of 40 Gb/s data transmission and higher.

The EN 50600 series of data centre classification standards represent another vital development. These detail everything from terminology, parameters and reference models to addressing issues in relation to business risk and operating cost analysis. And of course, there’s the ISO/IEC14763-2 standard, which specifies requirements for planning, installing and operating cabling and related infrastructures.

Five practical suggestions

There are a number of key steps which DC operators and designers should take, and several items they need to consider, in order to be future-proof.

1: You need to evaluate network connectivity as a key component of your physical infrastructure. Compliance with industry standards needs to be closely examined and you must determine whether the network is capable of supporting your current needs and future business initiatives. The goal must be to understand where you are, where you want to go and what it takes to get there.

2: In the area of fibre, choosing OM4 is the wise thing to do. These multimode fibres are created with optimised gradient index profiles and provide the highest bandwidth-distance-product. This allows transmission of data at a higher speed over a given distance, or at the same speed over a longer distance. Thus, it decreases dispersion and BER (bit error rate).

3: Do you need to save space beneath a raised floor? Consolidating LC trunk cables with MTP trunks is a practical, cost-effective option. This will also provide you with a solid foundation for a 40G or 100G rollout.

4: Although 24-fibre MTPs might seem like an attractive proposition, the upcoming 100G standard will be realised with 4x25G and therefore 12-fibre MTPs suffice.

5: In the field of copper, there has been quite some hype surrounding 40GBASE-T Cat 8.2. However, it is still too early to provide any specific details, as engineers of NIC manufacturers are still solving technical challenges. The old office cabling wisdom “Install now, you will need the capacity later” no longer applies to data centres!”

A closer look at High Performance Network Connectivity

In a data centre project, cabling infrastructure accounts for just 2 to 4% of the total investment. However, cabling is of crucial importance for the overall performance of networks. Cabling, in practice, isn’t something which should be replaced each time a new generation of active equipment comes along. Instead, it has to support several consecutive generations of equipment.

Between 40 and 50% of all breakdowns can be traced to low quality connectivity, and the resulting poor performance of physical networks. However, the fact that poor cabling might not only lead to the complete breakdown of a link, but can’t also significantly reduce network performance by causing high packet loss, is often forgotten.

Data packets consist of frames which have fixed sizes, as well as actual data (payload), which comes in a variety of sizes. In order to maximise the performance of a network, the payload size must be maximised and the overhead minimised.

However, bit errors may result in a loss of packets, which then have to be re-transmitted. Links with a high bit-error rate are, therefore, often set to small packet sizes in order to minimise the impact of lost packets. But as a higher number of small packets need to be switched and latency time is increased, an additional burden is placed on the network equipment. When using Fibre Channel, we’re not just looking at packet loss, but at frame damage, too.

Cutting losses

There are more dangers related to skimping on cable, connector and installation expenditure. Poor quality of components or work can lead to higher insertion and return loss, as well as increased modal and reflection noise. Choosing higher quality means reliable, uninterrupted operation.

BER is kept to a minimum which means lower packet loss rates (PLR) for your network. That, in turn, means that fewer data packets have to be re-sent and you obtain valuable bandwidth for your network.

Retransmission means delays, which can mean costly competitive disadvantages in many customer segments and applications, such as trading business at banks, running simulations at research data centre or ‘Software as a Service’ (SaaS) applications in a cloud (e.g.Citrix or SAP).

A more efficient network with bigger data throughput and smaller latency also results in higher availability and significantly better overall network performance. And ensuring your network lives up to these standards could simply mean no more than investing 1% extra on the total budget.

* Next Generation 40 Gb/s and 100 Gb/s Optical Ethernet

** IEEE Standard for Ethernet Amendment: Physical Layer and Management Parameters for 40 Gb/s Operation, Type 40GBASE-T

*** 100 Gb/s Backplane and Copper Cable