IELTS Academic Reading Practice Test 181

Walk into almost any large building and you will meet a moving staircase. It carries passengers between floors at a gentle angle, folds itself flat at each end and disappears beneath a metal plate, and it does so with so little drama that few people give it a thought. The escalator is nonetheless a machine of some subtlety, and its history involves two rival inventors, an amusement park, a trade name that became an ordinary word and a fire that changed the safety rules of an entire underground railway.

The first practical version was the work of Jesse Reno, an American engineer. In 1896 he installed an inclined elevator at Coney Island, the amusement district of Brooklyn, where it carried paying visitors up a slope at an angle of about twenty-five degrees. It was not a staircase at all but a continuous belt fitted with cleats, rather like a moving ramp, and passengers stepped onto a flat plate at the bottom and were carried up. Reno later built similar machines for department stores in several American cities, and the belt, rather than the step, remained the standard American design for some years.

A different approach came from Charles Seeberger, who wanted to move people up and along at the same time. His design used a chain of flat steps that stayed horizontal as they climbed, an arrangement that felt to passengers like an ordinary staircase and was much easier to board. He coined the name by which the machine is known, joining the Latin word for steps to part of the word elevator. The moving staircase was exhibited at the Paris Exposition of 1900, where it won a prize, and visitors queued to ride it.

Consolidation followed. The Otis Elevator Company, already the dominant maker of lifts, acquired the rights to Seeberger's design and within a few years had bought Reno's company as well, combining the horizontal-step mechanism with Reno's understanding of how such machines were installed and sold. The merged business promoted the new product vigorously, and before long the moving staircase had crossed the Atlantic and found its first serious use in public transport.

London was where that happened. In 1911 the Underground installed an escalator at Earl's Court station, built by Otis, to spare passengers the climb from deep platforms. The public was wary of the device, so the company employed a man named William Harris, who had lost a leg, to ride up and down all day and show that it was safe; he became a familiar figure known to the staff as Bumper. The experiment worked, and escalators were soon being fitted at stations across the network.

By the middle of the twentieth century the machine was a standard fitting in department stores, railway stations and airports. The name, meanwhile, had escaped its owner. Seeberger's coined term was registered as a trademark by Otis, but in 1950 the United States Patent Office ruled that it had become a generic term which any manufacturer was entitled to use.

An escalator is, in essence, a pair of endless chains running on either side of a steel frame called a truss. A motor and gearbox at the top drive the chains, and each step is fixed to both of them by wheels that run on two separate tracks. Because the two tracks sit at different heights, every step stays level throughout the climb, and at either end the steps slide together into a flat platform and pass out of sight beneath the floor.

Two details make the machine safe to use. At each end the steps pass beneath a comb plate, whose teeth mesh with the grooves cut in every tread, so that a shoelace, a coin or a scrap of paper is pushed out rather than dragged into the works. If a step is forced out of line, or a handrail begins to slip, a switch trips and the machine stops within a short distance.

Safety was learnt at a cost. On 18 November 1987 a fire broke out beneath a wooden escalator at King's Cross St Pancras station in London and spread rapidly up the shaft, killing 31 people. The treads and risers of the older machines were made of wood, and years of accumulated grease and rubbish beneath them fed the flames; a discarded match is thought to have started it. The fire led to the replacement of wooden escalators throughout the network and to much stricter rules on cleaning and on fire-resistant materials.

The modern machine has changed little in principle but a good deal in detail. Because a busy escalator runs empty for much of the day, manufacturers fit variable-speed drives that let it slow to a crawl, or stop altogether, when no one approaches the entry plate, and accelerate gently when a passenger steps on. Regenerative drives go further, turning the energy released by a heavily loaded escalator travelling downwards into electricity that is fed back into the building's supply instead of being burnt off as heat.

Designers once measured an escalator by the number of people it could move in an hour. Some now measure it by how little power it draws when nobody is riding it — an unglamorous advance, but one repeated thousands of times a day.

Cloud computing is the supply of computing — processing power, storage, software — over a network, so that the organisation using it does not own or maintain the machines. The definition most often quoted, from the National Institute of Standards and Technology in the United States, lists five essential features: on-demand self-service, so that a customer can obtain capacity without negotiating with a person; broad network access; resource pooling, in which many customers share the same hardware; rapid elasticity, the ability to grow or shrink capacity quickly; and measured service, with usage metered and billed. Services are usually divided into three layers — infrastructure, platform and software — according to how much the customer manages.

The word itself was already familiar in the 1990s, when engineers drew the internet as a cloud in diagrams in order to avoid showing its internal workings. The phrase 'cloud computing' began to appear in the middle of the following decade and was popularised by firms that had built very large data centres. In 2006 Amazon, until then an online bookseller, began renting out both storage and raw processing capacity to other businesses, an event generally taken as the start of the commercial cloud. Competing services from Microsoft, Google and others soon followed.

Behind the apparent simplicity lies virtualisation. Software called a hypervisor runs on a physical server and divides its processor time, memory and disk space among several virtual machines, each behaving like a complete computer with its own operating system. Because most servers are idle much of the time, a provider can sell the same physical capacity to many customers and reallocate it from minute to minute; this statistical sharing, known as oversubscription, is what makes cloud computing cheap. Containers take the process further, packaging an application with its dependencies but sharing the host's operating system instead of emulating a whole machine, so that they start in seconds and carry less overhead.

The hardware sits in data centres: windowless buildings filled with rows of racks. A large modern facility may draw as much electricity as a small town, and almost all of it ends up as heat. Cooling is therefore a major cost. Operators express efficiency as a ratio called power usage effectiveness, the total power consumed divided by the power delivered to the computing equipment, so that a perfect score is 1.0. The best-run hyperscale sites report figures close to 1.1, while the average across the industry has been nearer 1.6. Some operators cool their halls with outside air in cold climates; others rely on evaporating water, which raises questions about consumption in dry regions.

Distance matters because information cannot travel faster than light. In an optical fibre a signal moves at roughly two-thirds of the speed of light in a vacuum, so a round trip between London and Singapore takes well over a tenth of a second before any equipment has responded. Software that must react instantly — a trading system, a robot arm, the braking control of a vehicle — cannot afford that delay. Providers therefore build smaller data centres near their customers, and researchers are exploring computing at the edge of the network: on the device itself, or in a cabinet at the foot of a mobile mast.

Providers promise availability of well over 99.9 per cent, and they obtain it through redundancy. Data are copied across machines in separate buildings, called availability zones, so that the loss of one does not destroy the information. Instead of keeping three complete copies of everything, large systems increasingly use erasure coding, a mathematical method that splits data into fragments and adds parity information, so that the original can be rebuilt from a subset of the pieces. Few organisations could afford such protection on their own equipment, and this is one reason they give for moving to the cloud.

There are real drawbacks. Because tenants share hardware, a flaw in the isolation software may expose one customer's data to another. Customers must also trust the provider's staff and internal controls, since the data are physically out of their hands. Legislation adds a complication: some countries insist that certain records be held inside their borders, and a provider storing copies in several jurisdictions may struggle to satisfy every rule at once. Relying on a supplier's specialised services makes it expensive to leave, a difficulty known as lock-in. Concentration carries its own risk too, for when a major provider suffers an outage, thousands of unrelated organisations fail at the same moment.

Research is aimed at these weaknesses. Confidential computing keeps data encrypted in memory while a program is using them, decrypting only inside a protected region of the processor, so that even the operator of the machine cannot read what it holds. Serverless computing, in which the customer writes a small function and the platform supplies everything else, is being refined so that programs start with almost no delay. Heat is being tackled by immersing servers in tanks of non-conducting liquid, which removes it far more effectively than air. None of these methods is yet routine, but each addresses a limitation that has been obvious for years.

Providers offer their services in different forms. A public cloud is open to anyone able to pay; a private cloud is reserved for a single organisation, either on its own premises or in space rented from a supplier; and a hybrid cloud combines the two, keeping sensitive records in house while using public capacity for work that fluctuates, such as a retail website in the weeks before a major sale.

Anyone who has renewed a driving licence at midnight, from a kitchen table, in less than ten minutes has felt the benefit of electronic government. Anyone who has queued for two hours at an office to hand over the same information has felt the absence of it. Over the past two decades most governments have moved a large part of their dealings with the public onto the internet, and the argument about whether they should has largely been won. What remains disputed, and matters far more, is the terms on which the change is made.

Electronic government began in the late 1990s as little more than a set of websites publishing information: opening hours, forms, legislation. It then became transactional, so that a citizen could file a tax return or apply for a permit without printing anything. The United Nations publishes an index every two years comparing countries on the scope and quality of their online services, and the leaders are consistently small, wealthy states in northern Europe, together with a few others that have invested heavily. The index measures only what is offered online, however, not whether the people who need a service most are able to use it.

Estonia is the example invariably cited. After independence the country built a data-exchange layer, known as X-Road, which allows separate databases held by different agencies to be queried securely rather than pooled into one; a citizen's records are not stored in a single place. A compulsory electronic identity card, introduced in 2002, let people sign documents digitally with the same legal weight as ink. Since 2005 Estonians have been able to cast a vote online, and the practice was extended to parliamentary elections in 2007. In 2014 the state began issuing 'e-residency' to foreigners, who may use Estonian digital services but are not residents for tax purposes.

The financial case is genuine. Handling a transaction online costs a fraction of what the same transaction costs by telephone, and a tiny fraction of the cost at a counter. Britain created a Government Digital Service in 2011 and consolidated hundreds of departmental websites into a single domain, GOV.UK, which opened in 2012; the savings came less from closing websites than from replacing long, bespoke contracts with common platforms built once and reused. Where such work has been done patiently, waiting times and error rates have both fallen.

The failures have been just as instructive, and they were rarely failures of technology. In England, a programme intended to install a single system of electronic health records across the National Health Service was begun in 2002 and abandoned in 2011, having consumed billions of pounds, because its scope was vast and its contracts were negotiated with a handful of large suppliers over years rather than delivered in stages. The lesson drawn by most specialists is that public software should be built in small pieces, shown to users early, and altered as it goes, and that the freedom to experiment is worth more than a perfect plan.

Yet the most serious objection to electronic government is not cost but exclusion. Roughly a tenth of adults in wealthy countries rarely or never use the internet, and the proportion is much higher among the elderly, the poor, the disabled and those in remote areas. India's biometric identity system, Aadhaar, which has enrolled more than a billion people, is used to authenticate claims for subsidised food and welfare payments. Researchers and journalists have documented cases in which the fingerprint of an elderly or injured claimant would not register, and a payment was withheld until the claim could be proved by other means. A system that is faster for most and impossible for a few is not an improvement.

Governments have answered in different ways. Denmark requires citizens to receive post from the state in a digital mailbox, though it grants exemptions and keeps alternatives for those who cannot manage. Others have made online the default and paper the exception. The principle behind both is reasonable, since the state cannot run two parallel systems indefinitely; but the principle is only defensible if the exception is easy to find and quick to use, not buried three menus deep.

What would a defensible policy look like? First, digital identity should be treated as public infrastructure, issued once, usable everywhere and governed by law, so that citizens are not asked to prove who they are to a dozen agencies and to a private company besides. Second, services should be developed by small teams inside government with the authority to change what they build, rather than specified in advance by people who will never use them. Third, every service should have an assisted route that is properly staffed, and a minister, rather than a supplier, must remain accountable when it fails. And fourth, the data collected should be published where it can be, because openness is the cheapest audit there is.

The question, then, is not whether the state should move online. It should, and the convenience and the savings are real. The question is whether digitisation is treated as an opportunity to redesign a service around the person who needs it, or as a way of saving money by making the service someone else's problem. A government that exists only on a screen is cheap to run and easy to ignore, and the people it fails are, by definition, the ones least able to complain.