Wednesday, 26 June 2013

IT as a Utility

This is an old post from another blog, which seems to fit here as well. These are my reflections after the Digital Economy IT as a Utility network+ meeting in January 2012. At the end of the meeting reflections on outcomes and ways forward were invited – below is my contribution.



A Utility

What does it mean for IT to be a utility?
It wasn’t clear to me that we were all referring to the same end point, much less the same issues — with some people focused on narrower or overlapping definitions. I obviously arrived at the day with some preconceptions about this, but the discussions were useful in shaping these. I shall start by considering how I currently characterise existing utilities:
Generally, they are widely to universally available.
In some cases, such as sewers or septic tanks for waste water and gas or oil tanks for heating, the service is distinctly different for some users; but in most cases the gas, electricity and water are largely the same across the country and in homes, business and public services.
The quantity available scales very well.
Within some pre-arranged limits a user may draw varying quantities of the utility and expect the service to be consistent. For larger users the basic connection and contracts will be different to smaller users — the pipes, cables etc will be different at least.
In the UK these services are generally paid for.
The payment is generally for quantity, although some users may have serivce level agreements with a financial response. At a domestic level although there is a charge utilities are actually quite hard to disconnect people from for non-payment.
Much of recent memory characterises utilities as being provided by large organisations through a fixed infrastructure. However, small scale generation can contribute to a sustainable delivery provided that an agreed model of sharing exists. In both cases this assumes that the utility involves the use of a consumable resource.
Can “IT” be delivered as a utiltiy in the sense I’ve just outlined? The closest parts to this status at present are data communications and grid computing.

Where We Are Now

Data communications are widely available, although the variation in quality in the UK and elsewhere is much greater than most utilities exhibit. Although there are rural pockets with very limited service there is also widesperad availability away from the fixed infrastructure we usually associate with utilities. The available quantity scales well within pre-arranged limits, although in the case of broadband local demand is not as well balanced as most other utilities. The service is paid for, but generally according to nominal capacity rather than actual service delivered or consumed — some large users notwithstanding. At present data communications are mostly the domain of large providers, with very limited sharing of infrastructure — there isn’t a consumed good that can be separated from the infrastructure.
Grid computing was the basic model in many of the discussions at the meeting. It is generally available, provided an internet connection is available. For domestic customers the service is generally experienced as an application, while for business users it may be applications and computing services used by custom applications. The range of applications does differentiate the experienced quality of the service provided. This contrasts with electricity where the same provision can be used in many ways by the consumer. The primary selling point of grid computing is scalability — additional CPU cycles available on demand. The service is generally metered on use, so in many respects is rather like a utility. The sophisticated understanding the service provider has of each customers’ usage also represents an element of the exchange beyond a basic financial transaction. There have been some efforts to provide computational power on the basis of using spare capacity in small scale resources as well as through large scale computation centres. However these often take the form of applications delivered for free.
Communications services have some utiltiy aspects — where the utility is enabling human communication and storing data and links, to supply email, social networking etc. They are widely available, in the case of email fairly homogenous in terms of function but for social networks each provider is different to the next. The provision scales well for different levels of use and protocols and published APIs mean that some basic interoperability exists. Payment generally happens through advertising and/or by the provider making use of the data provided by users. However, is email a communications utility or an application running over storage, computation and communications services?

IT As Utilities

First, I suggest that IT is not a single utility.
Neither is it applications — these are consumption of a service.
The services which have a widespread use are internet access, computation, data storage (possibly with some computation attached), and accessing input and output devices. The first three are already present in some form, through broadband providers and grid and cloud services. The last enables much of pervasive computing and the internet of things and might push us to explore a wider domain than at present.
However, a utility needs to be more generic than what we have in IT at the moment. I do not think of my water, gas or electricity as being distinctive because of their brand — and indeed the competition here is somewhat artificial. My communications are somewhat differentiated by coverage and quality, that is assocaited with the provider — but I have an expectation that most other providers would be somewhat similar. Cloud services have a strong tie between use of the utility and the provider.

Research Questions

An inclusive agenda is hard to form, but the following topics came up during the day and interested me — from a computer science centric viewpoint. As a research community it is important to engage with industry, but as this market already exists there will be many aspects which are too close to market for academia to be useful. My suggestions are therefore focused on areas where business imperitive doesn’t necessarily drive development.
If we have distributed computation how do we trust the providers of service to do what they say they will and not look at what use we make of the utiltiy too closely? I suspect that we can look to operating systems to give us tools for providers to ensure that we stay within agreed limits?
If we have a very scalable resource how do we program to make use of this scalability — managing our use according to need for- and availability of- resource. I suspect that combining small programs is part of the key – much as I have several lightbulbs in my living room or use multiple heat sources to cook a meal. Similarly, models of use that enable us to model and predict resource use while addressing sustainability challenges will be needed.
If we have a common grid, where one provider’s computing is much like anothers’, how do we interface to this and between resources? There is a standardisation question here, to enable units of computation which are sufficiently general as to be usable on a range of platforms and for many purposes.
What are the economic models which make a computing utility work? Is there a meaningful market?  From my previous observations one might imagine that the market is a much more plausible outcome than for most present-day utilities — if there is an incentive to create a standard provision. What technology to ensure customer mobility is required for this market to function? Are there social and policy drivers that modify use of resource as well as economic ones?

Networks

I have a couple of observations on networks:
The value in a network with regard to knowledge comes where we meet those beyond our normal sub-domain. This may give rise to new collaborations and questions; this may reveal that work has already been done elsewhere.
The participation in a network can be motivated in various ways, including: by a view of an appealing challenge; by the availability of funding; and by the chance to network and play a part. My career benefitted from UK-UbiNet and the Grand Challenge activities — which subsidised my engagement (and that of various others) while still a junior researcher and able to put in the time. I would hope that this network might also engage beyond faculty.

Summary

I see three significant activities to take forward:
First, the definition of what the utilities are and the technology which makes them work as utility provision.
Second, the exploration of what economic and social models enable a utility rather than a commodity.
Third, the development of the technology which enables the effective consumption of the utility – separate from the provision.
These three can be thought of as reflecting the business of energy transmission, water storage and pipes etc; the business of resource management, pricing, metering, billing and regulation; and finally that of design of goods which deliver something that the customer needs from the utility: heat, light, a bath etc. While we don’t need to account for every possible application of IT, we ought to consider a range of applications — and investigae what the proprerties of an application which drives the need to turn IT into a utility are, rather as lightbulbs drove connecting houses to the electrical grid.

No comments:

Post a comment