DCN April 2016 - Page 25

cabling the ‘next generation’ of data centre infrastructure, into many. The ‘next generation’ label Before we go any further, I want to take issue with the apparently idle use of the term ‘next generation’. One of the many skills data centre professionals have had to hone over the last 5-10 years is separating the instructive from the nonsensical whenever technology is talked about. As this gets harder I’m sure I’m not the only one who gets fatigued by the use of ‘next generation’ to describe new breakthroughs. In the Star Trek TV series, ‘The Next Generation’ is literally just that – a completely new ship, crew and capabilities working to much the same objectives as the last. If someone were to offer you the next generation of personal transport, you’d expect a driverless car or a jetpack, not the same familiar family saloon with some slightly remodelled bodywork. To my mind something that’s truly ‘next generation’ should immediately compel you to buy it. In its present usage, it’s a little too clear cut to describe the rapidly changing landscape of solutions and technologies that enable you to evolve your data centre evolution at your own, often very fast, pace. Standards, standards – my kingdom for some standards Whether you’d call their efforts ‘next generation’ or not, the transceiver manufacturers have certainly hit a rich seam of innovation of late. Each has plunged headlong into a race to achieve the fastest possible speeds, pursued by the insatiable performance demands created by the hunger for faster, higher scale, cloud delivered services. The idea of 10GE as some kind of data centre bandwidth panacea is now a distant memory. Analysts at Dell’Oro predict 20 per cent of all server port shipments will be 25GE by 2018, with 40GE and 50GE ports accounting for a further 18 per cent. We’ve already seen the first 100GE shipments over the last 12 months. Higher server-switch links pass their burden on to aggregated switchswitch links. So when you look at 40GE switch ports sales, you can see these grew at twice the rate of 10GE in 2015 – by Dell’Oro’s estimate – and 100GE faster still. By 2018, with 100GE ports overtaking 40GE, the next milestone will be in the industry’s sights: 400GE. You want choice, well now you’ve got it. Take your pick from 10, 25, 40, 50, 100 and 400GE speeds – all are here or soon to be on their way. But for the poor unfortunates who must integrate this technology inside the data centre, the lack of commonality is quite alarming. Standards haven’t been able to keep pace with the white heat of innovation. The most logical migration is anyone’s guess. On what basis should data centre operators make their choice of vendor? safe in the knowledge that their networking infrastructure won’t let them down. Achieving that boils down to deploying an infrastructure that covers all eventualities, without having any costly, time consuming or disruptive upgrades. Flexibility has never been more important in data centre cabling. Why 8 is the magic number A new kind of cabling flexibility There’s so much choice for data centre operators in terms of applications or which technology to adopt, it’s becoming less and less clear which infrastructure will support this. Flexibility has never been more important in data centre cabling. Not so much flexibility of the ‘bend-it-round-your-finger’ variety, but flexibility in terms of enabling data centre operators to determine their transceiver strategy, The overwhelming majority of data centres have relied upon 2-fibre serial standards (notably Base2) – to aggregate multiple fibres to gain higher speeds. Applying this technology makes use of LC duplex connections, so that no infrastructure upgrades are necessary. A lot of innovation has focused on continuing this pursuit of 2-fibre solutions to achieve incrementally better performance, reusing installed bases of LC terminated fibre cabling and connectors. 25