Continuing my assessment of analysts I haven’t heard before: here at the Green IT Expo is Andy Lawrence of 451 Group, talking “Green Datacentres to Green Clouds”. Andy looks after data centre disruptive technologies, and eco-efficient IT, for 451. It’s the first time, again, that I’ve heard 451 directly.
He promises an overview including a European Union Framework project called Optimis. I’ve been involved in EU research in the past, didn’t know about this one: that’ll be interesting.
Here’s a sort-of hierarchy of energy efficiency for the data centre. At the base, five years’ work on reducing the power use of datacentre infrastructure: PUE, best practices, EU Code of Conduct. One tier up: work on lower power chips, efficient drives, virtualisation, power management etc. Above that again, the ability to look holistically at an application or service: for example, what’s the eco-impact of choosing 1 second response rather than 1.5 seconds?
So: Cloud. Cloud should be more eco-efficient and it’s often asserted to be so. But is it? 451 believes the assumptions are largely unproven. [Private] cloud and virtualisation, as a matter of observation, seems – so Lawrence says – to show under-utilisation so some of the eco gains are not realised..
We’re about to see another measurement framework. Here, there are four axes: economic, compliance, CSR (corporate social responsibility), operational effectiveness. Again, take a holistic look: e.g. what’s the energy cost of insisting backups are permanently online rather than powered-down (on tape, for example).
How do you measure resource efficiency? There are some proxy metrics; there are direct measures (i.e. actual measurements, not estimates: how much carbon now); and metrics (e.g. PUE). They are “good; but be careful: unreliable for business decisions”. We’re promised a tour of some cross-industry initiatives, and also a few highlights from individual companies.
The EU’s Optimis project provides a list for assessment: trust, risk, eco-efficiency, cost (TREC). The aim is to create an architectural framework that looks at all of these, and a development toolkit. Lawrence asserts the need for multiple metrics: “a lone metric never works”. The hard stuff is the effort to associate carbon with a cloud service, especially where the actual data are locked up in the provider’s data centre and they may well have no interest in providing the detailed data to feed into the models. It is, at the least, a hard problem.
Lawrence outlines an alternative proxy approach. It still relies on cloud providers doing the sums; but they may well already be gathering the data, and may well be more willing to deliver a category-based per-hour or per-VM footprint (kWh and carbon per VM hour, perhaps). Its accuracy needs to be similar to that of billing, neither much more nor much less.
This presentation has given me an incentive to revisit what I know of 451 Group: perhaps the most encouraging aspect was Andy Lawrence’s willingness to identify, and review, academic/industrial research projects which are easily overlooked by an insight market which tends to look only at vendors’ own development pipelines. It admits that development of real, workable methodologies is some time away: Optimis, like all EU Framework projects, is pre-competitive research. But while the project itself may not deliver the ultimate solution, the ideas it generates will certainly inform future metrics and tools.
• The 451 Group and The Uptime Institute
• Optimis EU project: Optimized Infrastructure Services
• EU GAMES: Green Active Management of Energy in IT Service centres (similar, for high performance computing)
• (These projects are within the EU’s 7th Framework Project; the CORDIS database holds information on these and all projects)