WHAT’S HOT wor ld news
3. Hyperscale and cloud drive efficiency and innovation.
Hyperscale operators (Google, Amazon, eBay, etc.) will continue to drive datacentre productivity by innovating relentlessly, exploiting economies of scale, improvements in processing and storage, piloting new designs, and pushing up virtualization and utilization. This will drive efficiency because competing operators and even enterprise datacentres will be under pressure to match cloud efficiency – and to better understand their own costs.
The extent to which these hyperscale operators will enjoy a sustained cost advantage is a matter for debate – there will be several types of datacentres optimized for their situations and workload, and innovations don’t always favour scale. Third- party operators of all kinds will increasingly apply metrics and invest to get more insight into true costs.
4. Business goals will drive operations in real time. Until now, day-to-day decision-making about the datacentre, from the IT down to the power supply, has largely been the domain of the technical management. While many datacentres are run as businesses, the metrics used rarely get down to the level of moving workloads or switching equipment off and on. Aided by advances in software and by real-time information systems (i.e., energy prices), and spurred by intense competition and commercial business goals (in a big sector of the datacentre market), operational technical decisions will increasingly be aligned to business goals or initiated because of business goals.
Decisions such as whether to switch on generators, or whether to put servers into hibernation will be driven as much by business parameters, such as cost management and service levels, as by technical issues.
5. The costs of new builds will fall. After a sustained period of rising prices, new enterprise-class datacentres in 2014 and beyond will be built for less than in the past decade, falling from highs of $15-20 per watt to $8 per watt, sometimes much less – even those with resilience and redundant architectures (this is a wide band, dependent on many local factors). And they
Summer 2014 I
www.dcseurope.info 9
will increasingly be built to a step-and-repeat model or using prefabricated modules, which will limit or eliminate investments in capacity before it is needed. Operational tools will be used to ensure capacity is tightly and efficiently managed.
6. Energy efficiency up – and energy trading?
PUE ratios, the main measurement of facility efficiency, are falling across the datacentre industry. From 2014, it is likely that no new datacentres will be built with a design PUE above 1.2-1.5 (depending on geography and resiliency). Using monitoring and facility optimization to reduce power waste and unnecessary cooling, the most efficient datacentres will operate with very low PUEs at most points of utilization.
At the same time, datacentres will further slash their energy bills (currently about one-third to one-half of datacentre operating costs) by buying more energy- efficient IT, by using software for workload management and by using IT power management. Some net ‘savings’ will also come in the form of ‘transactive energy management’ (managing the supply and demand of electricity in the datacentre, including transacting with the energy utility in real time or switching to on-site power sources at certain times).
The opportunity for the latter will become more apparent as the power grid itself becomes more volatile, with more use of renewable energy and smart- grid technologies. The opportunity to sell as well as buy power may swing the economics toward greater use of renewable generation.
7. Cloud-level resiliency.
Redundant topologies have served the datacentre extremely well, and will continue to do so, but more datacentres will be built with reduced resilience and redundancy, and without compromising on availability. As new capacity is brought online, IT and facilities designers will work together more to find more effective ways of meeting availability requirements without necessarily constructing full continuous-operations datacentres. Many applications and services are being run on infrastructure that is not justified by business demands, adding a layer of costs.
Among the strategies being deployed are: the use of multi-tier, modular-build datacentres, where some modules meet the higher levels of Uptime’s Tier models, while other modules operate at lower Tier II levels; the use of management software to move workloads when there are actual or expected disruptions; greater coupling of datacentres for synchronous replication, using fibre networks (including, at some point, 1Tb Ethernet); and the use of actual or reserved cloud capacity for backup and recovery, or reducing peak demand (‘cloud bursting’). Within the datacentre, high-speed photonics may encourage reduced use of power redundancy and more use of ‘distributed clustering’ for high availability.
8. The cloud is the computer: supernodes to microdatac enters.
The ability to move workloads between datacentres lies at the heart of the Internet, and was the foundation for the many ‘grid’ computing initiatives that preceded cloud computing. As cloud orchestration becomes more advanced, and increasingly more datacentres are linked by high-bandwidth fibre, the opportunity for work to move between datacentres and compute nodes increases. More datacentres will therefore be able to share work, to specialize in processing the most suitable tasks, and workloads will be able to seek out the ‘best execution venue.’ The datacentres themselves will vary – some will be
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56