Intel experts: Analysis and Challenges of cloud era large data

Intel anthropologist Genevieve Bell shared her research on big data, thinking deeply about possible changes to the future from big data, and exploring the implications of these big data phenomena in many different contexts.

&http://www.aliyun.com/zixun/aggregation/37954.html ">nbsp;

"For me, in this job, the data itself is not what interests me the most, because everything produces data, and the real interesting thing is how to develop algorithms that are better able to use the data," she explains.

Today, the way we use data is not all visually revealing the meaning of the data itself,

Bell's official title at Intel is the director of interactive and experiential research at Intel Labs, who works in a different sense from Intel's other scientists and engineers, covering all the recent data tsunamis (tsunami), cloud computing, and device proliferation that have sparked heated discussions in the industry ( Device proliferation), and more powerful servers. "Big Data" is Bell's next major research project.

"The question is where to start?" For the topic we want to discuss, we are still at the stage of trying to visualize the outline and size of the subject, "Bell said at a conference on the Xeon E5 processor in London earlier this time. "Today, all the databases that have reached full edge are just a start." ”

Next, Bell says, the world has to think, how do we filter that data? What changes will this bring to the traditional online database? Will the existing data world begin to become disorderly or even face collapse and then build?

And in all this change, Bell says, at least we are beginning to understand that everything is going in the direction of building a data-tracking law, including simulations and digital data, and that we have to be able to manage all the data.

"For me, in this job, the data itself is not what interests me the most, because everything produces data, and the real interesting thing is how to develop algorithms that are better able to use the data," she explains.

Today, the way we use data is not all visually revealing the meaning of the data itself, Bell said, especially the algorithm--to make all the data good to use, it is necessary to understand the correctness of the data itself, which is closely related to the original data-filling person.

"All the ideas we try to make the data more rational will be placed in the first place at the beginning," she said. In fact, she also points out that when we begin to look for fixed patterns, causal relationships, and associations between data in a single data model, we are already in the inherent bias.

"For me, the most fascinating aspect of big data analysis is not just that it brings together a wide range of data from various fields, but more importantly, what you can do with them," Bell says, "We are increasingly reliant on data analysis, which, in a way, represents cultural scripting (barely script). Is entering a new level.

The idea of such data analysis also drives the data towards "personification" (anthropomorphizing). Bell points out that all this data will be necessary at the beginning of the discussion of the "Secret Life of Data" (The secret). These discussions will further expand applications such as large data and cloud computing.

Bell says that all data actually has its own characteristics, which is what makes it different from all other data, so it's best to clarify the characteristics of all the data before analyzing it.

The physical characteristics of cloud computing and the concept of "wild" data

Bell believes that the data itself is capable of reproducing (ferality) or reproducing (reproductive). "The data will not always be in the same order as the algorithm or the input, and after cloud computing and other control mechanisms, it's important that the data change, because we're all trying to control the data that is" wild "in itself," Bell said. "Feral

In fact, Bell, for example, puts the same type of data all in one category, like putting rabbits or other highly fertile animals together, perhaps making an initial effective classification. She also points out that since data can easily be transferred between categories, or completely transferred to another category, context (contexts) can also be easily transferred, which may allow people to try to manage data or reduce the multiplication of these data. "In the data world, it's a pretty interesting thing to think about these strategies," she said thoughtfully.

Another feature of the data is that, while most of the data has a clear concept (conceptually), it is elusive. For example, the "cloud" actually still requires a lot of physical construction. "Cloud computing will eventually have a physical room. Cloud computing Data Center must be a physical building with servers in it, "Bell stressed the importance of physical construction."

Bell also stressed that cloud computing is not a special concept. "We're not just thinking about a single cloud system, we're thinking about a lot of clouds and thousands of physical devices, including data centers for data Analysis (farm), and for me all of these physical devices are a key part of the process of studying large data," she says.

This also raises a lot of questions about the study of large data, including whether the final data will be everywhere. Where is the datacenter located? Where should these data be governed by what law? What kind of network operation is used? And how to provide and what services will be provided?

At the same time, it also makes people think about whether everything or everyone will produce data, or, as William Gibson's comment on Cyberspace (cyberspace): "The future is coming, but it is not well known." "(future is already here but unevenly distributed)

Bell believes that the data will appear in various places in different ways, but not all the data is useful. "You can easily talk to Apple's Siri and it looks like The voice butler seems smart, but basically I just think it's a bunch of meaningless talk," she said.

What's more, how to deal with old devices and old data, which may not be digitized, how to deal with them and present in a new state can be a big problem.

"A growing amount of data is constantly being produced, but that data does not necessarily come from the human hand," she notes, noting that even though static data may be processed into dynamic information, the same result does not necessarily occur in the opposite case.

Around the world, the Internet of Things (things, IoT) mode of application is fermenting, from traffic lights to cars, refrigerators, lawn mowers, to fixed telephones, tablets, pens and televisions, all devices produce data, but in the future these data may not be suitable for every individual user, Said Bell.

Some families, for example, share telephones, while some family members use their own tablets while working during the day, but are offered to their children at night or on weekends. "How do we sort data to know how many users share a device?" "This makes sorting data a harder task," says Bell.

The characteristics of human data--uncertainty

However, the nightmare of development algorithms is not limited to ordering data from so many individual users and individual devices, Bell says, and more importantly, you have to keep in mind that some data is not real at all.

In a study conducted in the United States, she pointed out that almost 100% of the respondents to the study lied about dating sites in their personal data. So how do you sort these fictional data and give the results a very humane view?

"So at the moment people and machines are building data, but we need algorithms to help build more useful data," she says.

"The data is basically just a string of 0 and 1, and it's easy to imagine, and all you have to do is find the right tool and try to play that data." But if we stop thinking, the data will become a bunch of unimaginable digital symbols. So we're going to start imagining the contours of the data and imagine what it will look like when it's processed, "she said. "We're trying to make recommendations based on these datasets, but we can't assume that all the data are 100% true." ”

In addition, Bell notes that as the number of data users increases, their expectations for data are increasing, and these people will want every story, every moment, and every data generated to be processed accordingly.

Another question, then, is whether the data needs to be handled well and has a strong correlation. Or can it be presented in a scattered form?

With the rapid growth of devices, services, and applications, our experience with data is rapidly accumulating, and as cloud computing systems continue to grow, the concept of big data is growing, Bell said, which means that while the volume of data continues to grow, we will face a new set of challenges.

But are all these questions answered properly? Or are we still unable to reach the core of the problem? The big data represents "incredible business opportunities", Bell says, and it's not just a huge number, it's going to have a profound impact on today's system architects, engineers, device manufacturers, and users.

(Responsible editor: Liu Fen)