models [25]. A language model is a probability distribution that characterizes the likelihood of observing a particular sequence of words, estimated from a large corpus of texts. They are useful in a variety of applications, such as speech recognition (to determine what the speaker is more likely to have said) and machine translation (to determine which of possible translations is the most fluent, as we will discuss in Section 6.4). Since there are infinitely many possible strings, and probabilities must be assigned to all of them, language modeling is a more challenging task than simply keeping track of which strings were seen how many times: some number of likely strings will never be encountered, even with lots and lots of training data! Most modern language models make the Markov assumption: in a n-gram language model, the conditional probability of a word is given by the n − 1 previous words. Thus, by the chain rule, the probability of a sequence of words can be decomposed into the product of n-gram probabilities. Nevertheless, an enormous number of parameters must still be estimated from a training corpus: potentially V n parameters, where V is the number of words in the vocabulary. Even if we treat every word on the web as the training corpus from which to estimate the n-gram probabilities, most n-grams—in any language, even English—will never have been seen. To cope with this sparseness, researchers have developed a number of smoothing techniques [35; 79; 102], which all share the basic idea of moving probability mass from observed to unseen events in a principled manner. Smoothing approaches vary in effectiveness, both in terms of intrinsic and application-specific metrics. In 2007, Brants et al. [25] described language models trained on up to two trillion words.4 Their experiments compared a state-of-the-art approach known as Kneser-Ney smoothing [35] with another technique the authors affectionately referred to as “stupid backoff ”.5 Not surprisingly, stupid backoff didn’t work as well as Kneser-Ney smoothing on smaller corpora. However, it was simpler and could be trained on more data, which ultimately yielded better language models. That is, a simpler technique on more data beat a more sophisticated technique on less data.
Recently, three Google researchers summarized this data-driven philosophy in an essay titled The Unreasonable Effectiveness of Data [65].6 Why is this so? It boils down to the fact that language in the wild, just like human behavior in general, is messy. Unlike, say, the interaction of subatomic particles, human use of language is not constrained by succinct, universal “laws of grammar”. There are of course rules that govern the formation of words and sentences—for example, that verbs appear before objects in English, and that subjects and verbs must agree in number in many languages—but real-world language is affected by a multitude of other factors as well: people invent new words and phrases all the time, authors occasionally make mistakes, groups of individuals write within a shared context, etc. The Argentine writer Jorge Luis Borges wrote a famous allegorical one-paragraph story about a fictional society in which the art of cartography had gotten so advanced that their maps were as big as the lands they were describing.7 The world, he would say, is the best description of itself. In the same way, the more observations we gather about language use, the more accurate a description we have of language itself. This, in turn, translates into more effective algorithms and systems.
So, in summary, why large data? In some ways, the first answer is similar to the reason people climb mountains: because they’re there. But the second answer is even more compelling. Data represent the rising tide that lifts all boats—more data lead to better algorithms and systems for solving real-world problems. Now that we’ve addressed the why, let’s tackle the how. Let’s start with the obvious observation: data-intensive processing is beyond the capability of any individual machine and requires clusters—which means that large-data problems are fundamentally about organizing computations on dozens, hundreds, or even thousands of machines. This is exactly what MapReduce does, and the rest of this book is about the how.
1.1 COMPUTING IN THE CLOUDS
For better or for worse, it is often difficult to untangle MapReduce and large-data processing from the broader discourse on cloud computing. True, there is substantial promise in this new paradigm of computing, but unwarranted hype by the media and popular sources threatens its credibility in the long run. In some ways, cloud computing is simply brilliant marketing. Before clouds, there were grids, 8 and before grids, there were vector supercomputers, each having claimed to be the best thing since sliced bread.
So what exactly is cloud computing? This is one of those questions where 10 experts will give 11 different answers; in fact, countless papers have been written simply to attempt to define the term (e.g., [9; 31; 149], just to name a few examples). Here we offer up our own thoughts and attempt to explain how cloud computing relates to MapReduce and data-intensive processing.
At the most superficial level, everything that used to be called web applications has been rebranded to become “cloud applications”, which includes what we have previously called “Web 2.0” sites. In fact, anything running inside a browser that gathers and stores user-generated content now qualifies as an example of cloud computing. This includes social-networking services such as Facebook, video-sharing sites such as YouTube, web-based email services such as Gmail, and applications such as Google Docs. In this context, the cloud simply refers to the servers that power these sites, and user data is said to reside “in the cloud”. The accumulation of vast quantities of user data creates large-data problems, many of which are suitable for MapReduce. To give two concrete examples: a social-networking site analyzes connections in the enormous globe-spanning graph of friendships to recommend new connections. An online email service analyzes messages and user behavior to optimize ad selection and placement. These are all large-data problems that have been tackled with MapReduce.9
Another important facet of cloud computing is what’s more precisely known as utility computing [31; 129]. As the name implies, the idea behind utility computing is to treat computing resource as a metered service, like electricity or natural gas. The idea harkens back to the days of time-sharing machines, and in truth isn’t very different from this antiquated form of computing. Under this model, a “cloud user” can dynamically provision any amount of computing resources from a “cloud provider” on demand and only pay for what is consumed. In practical terms, the user is paying for access to virtual machine instances that run a standard operating system such as Linux. Virtualization technology (e.g., [15]) is used by the cloud provider to allocate available physical resources and enforce isolation between multiple users who may be sharing the same hardware. Once one or more virtual machine instances have been provisioned, the user has full control over the resources and can use them for arbitrary computation. Virtual machines that are no longer needed are destroyed, thereby freeing up physical resources that can be redirected to other users. Resource consumption is measured in some equivalent of machine-hours and users are charged in increments thereof.
Both users and providers benefit in the utility computing model. Users are freed from upfront capital investments necessary to build datacenters and substantial reoccurring costs in maintaining them. They also gain the important property of elasticity—as demand for computing resources grow, for example, from an unpredicted spike in customers, more resources can be seamlessly allocated from the cloud without an interruption in service. As demand falls, provisioned resources can be released. Prior to the advent of utility computing, coping with unexpected spikes in demand was fraught with challenges: under-provision and run the risk of service interruptions, or over-provision and tie up precious capital in idle machines that are depreciating.
From the utility provider point of view, this business also makes sense because large datacenters benefit from economies of scale and can be run more efficiently than smaller datacenters. In the same way that insurance works by aggregating risk and redistributing it, utility providers aggregate the computing demands for a large number of users. Although demand may fluctuate significantly for each user, overall trends in aggregate demand should be smooth and predictable, which allows the cloud provider to adjust capacity over time with less risk of either offering too much (resulting in inefficient use of capital) or too little (resulting in unsatisfied customers). In the world of utility computing, Amazon Web Services currently leads the way and remains the dominant player, but a number of other cloud providers populate a market that is becoming increasingly crowded. Most systems are based on