Next Generation Search and the Death of Google

Where is search going ? Real-time search ? Mood search? How about "no search" ? Shouldn't the internet learn about me and just anticipate my needs ? There are inklings of cognitive search making noises on the internet fringe. Companies such as Hunch, a collective intelligence app that delivers personalized recommendations based on individual responses to a set of taste-determining questions has taken to social search scene by storm. Google, with the introduction of "Buzz" and "Social Search" has targeted the social network as a tentpole in their bid to extend their dynasty through the Facebook era. For those unfamiliar with Google Social Search, it launched across all of Google in January, after initially rolling out as an experiment last October.When you’re signed in to Google, the service shows matching web pages and other content on the topic of your search that’s created by your friends or others that Google has determined that you’re “connected” to.

Why is social search interesting ? In a word, relevance

A lot of people write about iPhone, so if I do a search for iPhone on Google, my blog post about Facetime probably isn't going to show up on the first page of my results. Probably what I'll find are some well-known and official sites. With Social Search, the idea is you can scrape relevant public content from your friends and contacts.

From links to likes

The heir apparent to social search is without a doubt Facebook. When Facebook launched its Open Graph protocol in April,  it seemed obvious that one of the company's goals was to use the resulting behavioral data to power a social search engines. The company has confirmed that all Web pages that use the network's open graph plug-ins show up in the social network's search results in the same way traditional Facebook pages do...a real buzz-kill if you are Google.

The social network can improve, but it does not solve the question of relevance completely. In the end, relevance is subjective. You can do a search for 'coffee' in Canada and find Tim Horton's website as the most relevant. Makes sense, as that’s the most popular coffee chain in Canada, but for somebody in California, Peete's might be the most relevant result. You can do a search for the ‘49ers’ and be looking for the football team, but a historian may be looking for research material on California.

Social search presents a key evolution point in search, but we need more ! There are three cardinal steps before we realize a truly "invisible web": real-time cognition and authority.

Real-time

The Web's new currency is all about the diminishing half-life of data, and sifting through them for useful information is a challenge for search engines. Its most daunting aspect is not collecting the data. Facebook and Twitter are happy to sell access to their data feeds--or "fire hoses," as they call them--directly to search providers; the information pours straight into Google's computers.

What's really hard about real-time search is figuring out the meaning and value of those fleeting bits of information. The challenge goes beyond filtering out spam, though that's an important part of it. People who search real-time data want the same quality, authority, and relevance that they expect when they perform traditional Web searches. Nobody wants to drink straight from a fire hose.

The benefit of real-time search is a just-in-time composition of tweets and facebook posts relating to a time-sensitive topic, such as which bar is hopping tonight ? Taken to the extreme, real-time search could provide trending data that could allow predictions of where a stock could track.

Cognition

This is the secret sauce of the search world - how do you anticipate user intent ? The notion is familiar to anyone using iTunes Genius or Netflix recommendation tools. Most of these systems revolve around a process called collaborative filtering. Collaborative filtering systems discover new items you might be interested based on your historic preferences such as explicit ratings or implicit click behavior. The correlations that can be drawn can be prescient (all democrats like Star Wars, you are a democrat, so you like Star Wars), or they can be real long shots.   The end goal is to uncover new and and relevant information without you searching for it.

Indeed companies such as Hunch have bet the farm on their ability to generate these correlations using vast data sets of user preferences. Clearly, these techniques presents challenges if you are not Facebook. The key is to be able to get people to provide self-declared data and a list of likes and dislikes...not easy. The traditional method is to get users to volunteer the information through a series of mini-games, and to apply this information by signing up to a search community prior to entering the search string. It would be great if there was a "taste passport" we could port to these communities to exploit the nuanced databases that have been compiled - in fact metastructures such as "open social" is an early portent of such a universe.

Authority

Obvious right ? While social search provides trust, recommendations provide cognition, is the information authentic ? Sure I have friends who know a thing or two about music, but how authoritative are they on the topic of next generation search ? How can we create a peer-reviewed authority index, and how can it be brokered effectively. For instance, if I'm working at Ericsson and I serach for a topic in wireless networking, I want results that my CTO has deemed legitimate - or perhaps I would want a view of the material that all wireless CTOs in my industry has authentticated. A possible approach may be to elect regional, national and global indexes to track peer-reviewed experts divided into categories. IEEE may be a good start, or an analysis of papers that are most referenced (page rank for academia). The incentive would be to a create an authority marketplace or even an acution process where a user who has been deemed an authority could be part of a just-in-time vote to provide his or her "paper of choice" for a given search result (much like Google auctions advertisment) - the reward being accrual of authority credits to be used to "purchase" an authoritative search referral. Ok, maybe that last bit is going overboard.

I'm personally looking forward when the web can figure me out, and start placing that advance order for the next Cure comeback album without me having to go through the drudge of typing it myself, and maybe even cancel it when it figures out that my fellow Cure fanboys decides the album sucks. That is the world I want to live in...

Cellular Design for Applications

I asked the following question the other day at an industry conference, with several carriers in the room: "How will video impact the way you design your cell edge ?" the room went deathly quiet, not the least because it was just after lunch - however, this was my subtle way saying that we have not yet linked our cellular design practices to services other than voice. Yes, cellular data has been deployed, but make no mistake about it, we are still designing our cell edge to correspond to voice throughputs and we are still dimensioning our backhaul networks to correspond to the maximum number of simultaneous voice conversations at a given cell site.

Signal strength affects the coverage as well as the caller capacity. It is relatively easy to prevent signal source hunting and dropped calls - but these values are engineered for voice, which dictates both coverage and capacity. Data introduces new challenges which require both a lower link budget and a higher carrier-to-interference-plus-noise ratio (CINR) the higher the data rate required. In HSPA, these values are based on power and code dynamic (or fixed) allocation, which provides allowance to segregate the resources to avoid starving, say,  voice for data - but the data rewsource analysis is usually no more sophisticated than "what's left over" after voice has been optimized.

There is good reason to oprimize for voice and have only "best effort" data. The economic reality is that, in North America, 70% (or more) of the revenue comes from voice - and this represents only 30% of the operator traffic ! Therefore the revenue per bit of voice is much more precious than for data. This model is not long for the world, given the rising tide of over-the-top solutions driving the "everything is a data application"-future. These trends will force new planning models, and the question of : "what is the data outage criteria ?" will loom large.

Fundamentally, the notion of engineering the network for, say,  "video" versus "voice" will require separation of applications - potentially to separate carriers, or at a minimum some degree of isolation on the RF scheduler. It is simply not pragmatic to engineer for worst case applications uniformly throughout the network (the cost of engineering 2xHD channels at cell edge, when the majority of users are web browsing, doesn't make fiscal sense) - more so, we will need "virtual" cell-edge limits for different application classes - from video, social media and web browsing, each having diffferent handoff thresholds (and associated "virtual layer management"). Each application class will contribute to a new dynamic composite data model which will be needed to evaluate network bottlenecks from social media signalling intensive models to video data path intensive models. The "customer experience" for each application class will also require further definition. For instance, "always on" sessions are important for presence base services - but "internet snacking" may well have looser requirements.

This management has been avoided on the wireline side of the house, as the bandwidth has been ahead of the usage curve - however, the wireless throughput still very much lags the broadband wireline equivalents, and therefore quality of service and traffic management will continue to dominate the conversation in cellular.

Wireless data is really just "coming into its own" - we have spent the past 100 years getting voice figured out, and, while it's not the sexy part, we need to pick up the slack on defining the same for data - data erlang anyone ?

The 100 Dollar Smartphone

Creeping under the belly of the telecom machine, away from the frontlines of smart pipes and cloud computing, a revolution is underway. Commoditization threatens to drain the emergent profit pool of the smartphone duopoly, and may yet wrest the final jewel from the carrier crown.

A key control point for operators are device subsidies. Many carriers in North America wield large marketing "subscriber acquisition" budgets for devices, granting them authority to dictate terms to their handset suppliers. Apple has challenged this model, albeit on only 2% of the market. The Android license-free model has the potential to extend well beyond a tiny market share, and could precipitate retail price erosion to point of rendering subsidies meaningless. Smartphones in particular drive the global mobile ecosystem. According to Gartner, 172 million of the 1.2 billion mobile phones shipped were smartphones (an increase of 23.8% from 2008)

With all handset vendors (apart from Nokia) investing in Android powered handsets, the industry dynamic is already changing. In parallel, chipset vendors led by Mediatek and Qualcomm (who recently announced a cross-licensing deal on 2G, 3G and 4G patents) have created out-of-the-box reference designs integrating hardware, OS and applications (Android Market and well known Google applications). A handset vendor can take Qualcomm's design, pre-integrated with Android, and go-to-market within 9-12 months (down from 16 months for Motorola Cliq, and 24 months for HTC G1). Other major chipset vendors are joining the fray, such as TI OMAP3. ST-Ericsson and Broadcom - they are all ramping chipset designs with native support for Android.

The net result of this paint by numbers handset design will rapidly decrease price. The Mediatek exports will create a new competitive ceiling for chipset vendors, and the silicon power curve will allow integration of smartphone capacbilities at feature-phone price points. When we put Android, Qualcomm and Mediatek together - it is a certainty that we will have a $100 smartphone within 3-years.

The faith will have to be placed in a new class of device to create a demand outside the industry power curve. However, it is more important than ever for the operator community to invest in cloud-based services and other meaningful brand deliverables to cross the coming chasm between the old and new value chain of the telecom industry.

The Mobile VideoChat Faceoff

Ok, full disclosure: I'm a mobile videochat hater. I stress "mobile", because I have no problem with a telepresence, or even a PS3 videoconference. Mobile videochat has always been a useability nightmare: when you interact with a mobile device today, you don’t hold it at head height – you hold it a lot lower than that. Video telephony requires holding the phone higher. Mobile devices don’t provide the experience you get by having a video call in a conference room or from your laptop. Camera positioning is key here: on mobile handsets, the front facing camera forces the user to hold his hand in front of his face in an uncomfortable position – especially taking into consideration that today’s video calls are usually long ones. Add to that the fact that you need to deal with the phone’s speakers or connect a headset, add the noisy surroundings, and you have a recipe for bad experience.

Having said that, I will be following the new Apple Facetime application with great interest - because when you choreograph videochat into the application toolkit, the useability may be trumped by usefulness. In other words, the concept f mobile video telephony as a killer application is wrongheaded, but the ue of it as a building block for various applications is compelling. What is this usefulness you ask ? Let me count the ways:

1. Video Mail: video mail support has been deployed by multiple operators worldwide. It allows people to leave video messages from one to another and retrieve them later. In the same way that voice mail services suffered from the rise of SMS, so does video mail, which was already disadvantaged by the limited use of video calling services.

2. Mobile TV and video on demand (VOD): while there are other options for mobile TV, mobile video telephony provides a solution that is standardized and available across most handsets on the market. Where mobile TV is fragmented between standards, video telephony can come to play.

3. Entertainment TV: Mobile TV is nice, but adding interactivity was thought to be a killer application, especially for sports programs and reality shows. Trials of connecting video calling with sporting events and big brother have been done, but none have caught up.

4. PC-to-mobile: video calling over the desktop is used a lot more than over 3G. That being the case, the ability to bridge the two has been tried by a number of operators around the world.

5. Banking: banks have warmed up to video communications. They use it to enable access to specialists in remote branches or to allow people to contact a bank clerk remotely. They offer some of these services from mobile handsets as well – using mobile video telephony. Another interesting use of video communication in banking is accessing ATM services through video calling instead of voice calling.

6. Visual call centers: this is an easy one. Wherever there is a voice call center, a video one that allows mobile phones to call by video makes sense.

7. Healthcare: Mobile video telephony is used today around the world by doctors to communicate between peers and consult with specialists.

8. Hard of Hearing (VRS): Video Relay Services enable deaf and hard of hearing people to communicate with the world by way of a mediator who communicates with them through the use of sign language using video communications.

What becomes apparent from these use cases is that video is not used as a bidirectional conversation, but rather as a one-way real-time video communication for the consumer who wants to see the person they are talking to.

Where is mobile video telephony used ? While we have no real mass adoption of mobile video telephony, there have been some notable trials that have been going on for the past several years around the world - with several key industry events creating a baseline:

1. 3G-324M (a 3GPP specification), along with a GCF alidation process, ensures that any handset coming to the market with video telephony can interoperate with any other handset out there. Putting the GCF test cases in place was a process of over a year, discussing the various tests that should be included and the creation of the ecosystem around it – mainly test labs and testing tools.

2. Operators have mandated the inclusion of 3G-324M support in all 3G handsets that they sell to their customer base.

3. Roaming agreements between operators in Europe and Asia have been put in place so that you can now dial an international mobile video call to others.

4. Call setup time has reduced from 7-15 seconds to below 1 second using additions to the standard. This was pointed as a barrier for consumer acceptance of the service, and operators have worked to successfully remove the barrier.

So the standards have matured and interoperability is usually a solved problem. But still – consumer adoption is lacking. This isn't because of a lack of demand. Consider that with Skype, 36% of Skype to Skype calls are video calls.

The key issue again is useability (as discussed above), video quality (today it is 64 kilobits per second vs a high definition video channel of 2-4 megabits per second - the iPhone will be the first litmus test for wether video quality can increase demand, given the high resolution of it display) and finally coverage.

I think the combination of Apple's application ecosystem, it's reliance on WiFi for broadband speeds, and the expected iPhone 4 population (to create a videochat user pool) may provide the proof for this pudding - even if this is one dessert I may have to take a pass on....

Small Cells

I've discussed the balanced case for femtocells in prior posts, but it is important to keep in mind that femtocells provide one approach in the battle for broadband coverage and capacity. Femtocells could be a low cost way of providing hotspot capacity in public areas. But in some situations, it will compete with existing techniques. This post will compare and contrast femtocells with DAS.

What is DAS

Distributed Antenna Systems (DAS) are a commonly used technique to provide excellent coverage and capacity throughout large buildings or campus areas. They are commonly deployed in large office buildings, shopping centres and conference centres. Radio power levels are much reduced for both transmitters and mobile devices, because an antenna is usually within line of sight and only a short distance away. This benefits call quality, data rates/response and battery life.

The formal definition of DAS?

Like every good new invention, the technology has its own not-for-profit organisation – the DAS Forum , although this seems to be very US centric and not very active. Their official definition of DAS is “a network of spatially separated antenna nodes connected to a common source via a transport medium that provides wireless service within a geographic area or structure. DAS antenna elevations are generally at or below the clutter level and node installations are compact”.

How does DAS work in Plain English?

What this means in plain English is that the RF signals to and from the mobile operators cellsite basestation are piped through a system of multiple antennas. There are a number of different types of DAS each with their own characteristics :

Passive DAS – where RF signals are combined using passive components such as filters, splitters and couplers. Great for multiple bands and small to medium size locations

Active DAS – RF signals are converted and distributed over fibre. Great for larger installations but more costly and have to be dedicated to bands.

Hybrid – combination of the above techniques

DAS can be employed purely within a large building (In-building DAS) or across a large urban area (Street Level DAS). Street Level DAS can provide a very efficient solution for large urban regeneration projects which require dense coverage. They can also be provided in other busy areas such as Metros, Airports or Railway Stations.

Pros and Cons

The benefits of this approach include:
Operator equipment is located in one place – simpler maintenance and upgrade procedures Supports multiple network operators, allowing sharing of costs and resources
RF coverage can be tailored to meet the needs of specific buildings and use cases
Easily upgraded to handle new frequencies, transmission technologies, capacity
Allows capacity uplift by offloading in-building traffic from Macro layer
Supports rollout of bandwidth hungry, low latency applications

But the downsides include:
High capital investment – only justified for large airports, businesses, campuses, centres
Complexity - Needs specialist RF expertise to design and maintain Large Business Premises

DAS is commonly used today by operators for equipping large business premises. Where a business signs up exclusively with one operator, large amounts of capacity and good coverage are required and the investment to provide this is justified. It’s cheaper than trying to penetrate the building from several high capacity outdoor macrocells, which could have interference issues in the wider area.

Shared Buildings

A separate situation arises for large shared offices, shopping centres, airports and the like where large numbers of mobile users can be expected from all national networks. The business case for the DAS is largely driven by volume, but some building owners now deploy DAS as a ‘hygiene factor’ to encourage more people on site.

Shared Access Approach

One commercial approach is where property owners deal with an intermediary who installs a common DAS system and arranges the commercial and technical aspects with each individual operator. Typical companies in this field include Shared Access, who have both the technical and commercial expertise and an enlightened approach on financing the solution.
In a deployment, each network operator would install their own high capacity cellsites (separate 2G and 3G equipment) in a central location. The RF output from all is combined and distributed throughout the building via the DAS. The system can support a wide variety of technologies and frequencies, including GSM, CDMA, WCDMA, HSPA, LTE, WiFi and even PMR.

DAS versus Metro-Femto

With the recent hype around the Metro-Femto concept, these DAS systems may be more easily upgraded to provide LTE capability than installing separate LTE femtocells from each operator around the buildings. Operators would only need to install one central LTE basestation each, rather than many independent LTE femtos. This depends to some extent on what spectrum is used for LTE – additional spectrum may require additional active components.
Outside these areas, where individual operators need to provide coverage in dense urban areas/hotspots, the metro-femto concept could be more advantageous. Where femtocells are very low cost, and are self-organising, there are likely to be situations where these are more attractive solutions.

DAS versus Domestic/Small Business Femtocell

The business case for femtocells in small and medium businesses remains however. DAS systems are only viable for large premises with 1000’s of employees or users, whereas femtocells can be viable for single employee businesses through to hundreds.
There is some overlap between femtocells and single repeaters in SME's but whilst repeaters installations have been de-skilled to a large degree, operators look at them carefully from a commercial perspective before proceeding. Today, businesses have to be spending several thousands of dollars a month before operators are interested in specially engineered solutions - something which could change dramatically with the lower price point and TCO of femtocells.