Warning: Data centre reality distortion field
Data centres are hot. As business turns to computing and infrastructure services, almost anyone with a spare server and half-empty computer room is promoting data centre services.New Zealand’s data centre land grab is well underway. Like all periods of growth and expectation, marketing involvement loosens promotion and breathless claims about the ‘biggest’ or ‘most secure’ trump international standards for assessing data centre availability, such as the TIA-942 Telecommunications Infrastructure Standards for Data Centres, and the Uptime Institute. Even this hallowed ground isn’t safe, with tier names twisted into invented terms like ‘Tier 3-plus’ and flagrant overstatements, like ‘Tier 4’ data centre services (until New Zealand has a dual electricity network Tier 4 data centres are an impossibility here). While data centre tiers are a useful benchmark, they’re not the only consideration, and factors including operating practices, multiple inter-data centre capability and service platform innovation and lifecycle can make a lower-tier facility in practical terms as reliable as a facility with a higher-tier rating.Tier timeFirst, let’s clear the air on data centre tiers. The first point is to never trust a tier rating that has a data centre vendor’s name in front of it. The recognised standard for availability and a good discussion start point for data centre reliability is the TIA-942 Telecommunications Infrastructure Standards for Data Centres. Introduced in 2005, it is the first standard specifically addressing data centre infrastructure, and covers site-space layout, cabling infrastructure, tiered reliability and environmental considerations. Visit the TIA website to get all the details, but to summarise: Tier 1 (Basic, 99.671% availability):Susceptible to disruptions from both planned and unplanned activity; single path for power and cooling (no redundant components); may or may not include generator or UPS; complete shut down for preventative maintenance. This is mostly the domain of ISPs who almost overnight have changed their signs to say data centre services. Tier 2 (Redundant components, 99.741% availability): Less susceptible to disruption than Tier 1; single path for power and cooling (includes N+1 redundant components); power path and other infrastructure maintenance requires processing shutdown. The majority of New Zealand data centres are Tier 2.Tier 3 (Concurrently maintainable, 99.982% availability):Planned activity doesn’t disrupt hardware operation; multiple power and cooling distribution pathways; N+1 redundant components. There are a handful of Tier 3 data centres in Auckland, one in Wellington and none in the South Island. Tier 4 (Fault tolerant, 99.995% availability):Not available in New Zealand.Other determinantsTiers only get you so far, and practicalities demand consideration of other factors. For example, a Tier 3 data centre is inherently risky when located on Wellington’s waterfront wash zone, on a flood plain, a flight path, an earthquake high-shake zone or in regions known for extreme weather events.Data centre checklists flag fundamental user considerations like these, however, checklist fundamentals are just that – table stakes (you can’t play the game if you don’t have them) – and two other key determinants of data centre performance and reliability include remote data centre synchronisation (a data centre network) and granular control. Many hands make safer workThe greater the number of data centres managing systems and data, the smaller the risk. You don’t want all your eggs in one basket – it’s just simple commonsense.Equally, locating DR capability within an hour’s drive of production systems is still too close for proper geographic risk mitigation. If something in a single monolithic data centre breaks, or a regional disaster prevents access, re-establishing client systems is a huge task, because the lion’s share of data centre capacity has gone. A data centre network containing three or more nodes gives you options. If one data centre is disabled or inaccessible then capacity isn’t substantially diminished. Combined with smart networking, you’re able to manage remote data centres as a single entity, providing instant options.So, how does the New Zealand data centre scene stack up? In my mind, New Zealand is geographically top-heavy with most data centre investment sunk into monolithic facilities in the North. What happens when a cyclone or weather event of Bola or, more recently, Queensland, proportions hits the region? And what about earthquakes? We’re not called the Shaky Isles for nothing. The only data centre strategy that truly mitigates risks is one that includes multiple geographically remote nodes.Monolithic limitationsIt’s also worth pointing out the limitations of monolithic data centres.Computers are getting smaller, so why the acres of space? Regardless of tier rating, monolithic data centres are wasteful when not full. Blades today pack five times more punch than blades we were buying two years ago. You can fit more and more into less space. So what is there to worry about? The grunt of commercial computing is being drawn to the continuous supply and power and air, which is mostly an availability issue. Yes, it involves data centre buildings, but rather than huge floor space and high-walled facilities, the new availability rules dictate modularity, granular control and standardisation, so businesses draw only what they need. This is an exercise in thinking small, physically, at least. It’s smarter, safer, and more standardised.