Posted by Ido Safruti on Fri, Dec 30, 2011 @ 05:59 PM
I’m happy to report that 2011 was a great year for Web and mobile performance, for acceleration, and for Cotendo. As we begin 2012, I want to share some thoughts on the acceleration industry and the environment we work in. Specifically, I’ll focus on trends in 5 key areas and add some predictions of my own.
1. Mobile
Overall, 2011 was a great year for mobile adoption. We saw the emergence of HTML5, responsive Web design, and many frameworks and libraries that pushed the envelope and broke down the barriers between "the mobile Web" and “the Web.” In 2012, this trend will continue, with mobile adoption moving from thought leaders such as Google and Facebook to the main market. Increasingly, content providers will see mobile as part of their standard offering and develop it in-house, rather than outsourcing it to third parties and mobilization companies.
The exciting part is that designers and developers are figuring out what defines the mobile experience and what customers are looking to do via mobile with their brand. They are also finding new needs for acceleration and performance improvement such as handling location-based, personal, and social data, and handling the limitations and different characteristics of the devices.
In 2011, we introduced our Mobile Acceleration Suite. And, in 2012, there will be great opportunities for tools that address mobile content and application acceleration needs.
2. Enterprise Cloud Acceleration
The way I see it, 2012 is a break-out year for cloud acceleration services for the enterprise.
In 2011, many enterprises began to integrate and migrate their applications and data centers to different cloud solutions. They built consolidated large datacenters, used public infrastructures for their services, outsourced solutions such as SaaS, and often employed different combinations of the above.
All these services have one thing in common: users are no longer a few rooms away from the servers that their applications run on; in fact, they’re often not even in the same facility. In addition, these people may be mobile users, working from a home office or on the road, or accessing the service through their mobile device. All this distance, of course, leads to data latency and a huge demand for acceleration of the kinds of services that weren't originally created in such a far-flung environment. The way I see it, without proper acceleration, the transition to cloud services can’t be successful.
There are many challenges to overcome in the enterprise segment: security, public Internet vs. VPN, on-prem/off-prem, and more. But there are also enormous opportunities, not the least of which is content and application acceleration. In many cases, in fact, acceleration will act as an enabler of the cloud migration process. Tapping into the challenges and the special needs of the enterprise market will be a key to cracking this market. Cotendo’s partnership with Citrix in 2011, for example, gives us a good start with the launch of the CloudConnector Service.
3. SPDY and SPDY Based Services
Being a strong believer and supporter of SPDY, I regret to say that I don’t believe that in 2012 SPDY will become abundant and served from every rooftop.
I am sure, however, that we will get a whole lot of supporting data from large-scale implementations and experiments with SPDY (or its concepts) to add to the published data so far provided only by Google.
It is important to emphasize that I am not necessarily speaking of SPDY in its current form but more about how it evolves, and the concepts is stands for.
SPDY had a great year in 2011. It was enabled by default on Chrome and Google’s SSL services, which helped to troubleshoot and address many of its “birth pains” as well as to stabilize and mature the protocol, let alone providing a working proof that it can work on large scale. Aside from Google, other players started supporting SPDY, both on the client side (Firefox and Android planned support, and Amazon’s Silk) as well as on the server side, where we at Cotendo were the first to offer SPDY as a service.
In 2012, SPDY will be tested on a larger scale and will have an opportunity to deliver on its promise. This is very exciting. Aside from the immediate improvement over standard HTTP/HTTPS, and the potential benefit for mobile, I believe we will see many more technologies and services that will take SPDY further, offering new class of acceleration services. Amazon's Silk is a good example. The potential is great, and SPDY can become a game-changing technology when addressing the challenges and pains of mobile networks and devices with limited resources. The foundation has been laid, and 2012 must be the year to deliver.
Don't get me wrong – there are many more steps to go through. But I am sure there will also be many interesting announcements in 2012 and that they will be backed by hard evidence that all of this effort is well worth the investment.
4. Front-End-Optimization (FEO)
Last year showed great progress for FEO and the application optimization market. Awareness of the space grew as more tools and services appeared, and more optimization resources became available.
To my view, the key challenges in FEO that are yet to be proven are automation, service, and scale. Automation is the ability both to easily set up and automatically apply optimizations and to automatically detect changes or problems, to ensure reliable delivery. Providing FEO as a service and ensuring that such a service economically and efficiently scales is critical for it in order to have a meaningful impact on the market.
I believe that 2012 will clearly mark which vendor/technology figured these well, and will come up as the winner.
Another critical challenge for the success of FEO services is getting solid data and robust measurement tools, which are even more challenging in the mobile world.
This last point brings me to the next – and last – item.
5. Measuring Mobile Performance
The first step in optimization and performance improvement is getting the data right, and the mobile acceleration industry does not currently have relevant and standardized performance measurements. The traditional Gomez/Keynote synthetic tests are reaching their limits in measuring mobile traffic as both devices and network conditions change. These tools don’t take into account or track browser specific optimizations, repeat visits, and local storage, to name a few. Tools such as webpagetest and its mobile addition mobitest, and resources such as httparchive, are all great steps forward, but we aren't there yet.
I don't expect that in 2012 the industry will reach a concrete gold standard for measuring mobile performance (like Gomez/Keynote provided with CDN tests in the last few years). But I believe we will see progress and the formation of new standards. So get vocal if you care about this issue. Now is the time to make an impact. Defining what we track will also define what we optimize for.
• • •
As we begin 2012, we have much to look forward to. But, along with some marvelous opportunities, we have some big challenges – we also have much to do. At Cotendo, we are excited about addressing the challenges and hopefully help some of these wishes come true.
Posted by Shlomi Gian on Thu, Dec 08, 2011 @ 02:25 PM
We have been arguing that latency is the number one performance killer for mobile. And today, we are more fervent about this than ever.
Many factors contribute to latency’s deadly power. To begin with, mobile delivery time is much longer than broadband. In addition, mobile is sensitive to RF signals, network congestion, cell switching, and even weather conditions. And, depending on all these variables, page loads can take anywhere from five to 20 times longer than broadband.
But I have long suspected that there’s more to it than this, and I asked myself: Just how strong is the correlation between latency and slow load times? Our friends at Blaze IO provided some great insights, which you can see in the graph below.
To help answer this question in more detail, I created a chart based on performance data I collected under different latency conditions from major mobile news sites such as CNN, Fox, and The Washington Post.
I found the results very intriguing. When latency was just 30 milliseconds, for example, the CNN mobile site took just 1.4 seconds to load. But, with 300 milliseconds of latency, load time took roughly 6.9 seconds. So, a latency increase of 270 milliseconds delayed CNN’s page load start time by 5.5 seconds. In the world of a mobile on-the-go user, that’s forever!
The key learning here is that latency doesn’t have a linear impact but an exponential one. Every millisecond counts, and it counts for a whole lot than most of us assume.
Inside every cloud, the saying goes, there’s a silver lining. And there’s one here, too—if we think about things a little differently. Imagine that, if we could reduce the latency by just a few hundred milliseconds, we can dramatically reduce the page load times for our customers.
So when we designed our Mobile Acceleration Suite, a major goal for us was to shorten that latency on mobile content delivery. In the process, we learned that what affects it most is not network congestion or weather conditions but rather the distance between the end user and the server.
The distance between the server and the user dramatically affects latency and the speed of mobile content delivery.
With these thoughts in mind—and knowing that mobile carriers have the ability to modify data routing—we looked for a partner who could work well with us to achieve this goal. And we found that partner in Equinix, a company that shares our philosophy of a mobile “echo system” and allows us to move servers within its infrastructure to remain as close to possible to its GGSNs. By doing this, we ensure our customers of consistently low latency levels and their users of much faster page loads, higher levels of engagement, and, yes, conversions. Isn’t that ultimately what it’s all about?
Today, we discussed mobile performance issues in our webinar with Equinix “Structuring Mobile Content and Applications for Performance and Revenue Gains.” If you could not attend the live event, we urge you to watch it on demand. It will be available for the next couple of weeks.
And remember, latency does not have to kill you. You can kill latency.
Posted by Corinna Krueger on Thu, Oct 27, 2011 @ 11:44 PM
We love the thrill of speed when we can instantly share updates on Facebook, check out he latest deals on Groupon or purchase movie tickets via Fandango when heading out for the evening to take in the latest blockbuster hit.
The same need for speed holds true in the Enterprise world as users access their BI, CRM and other applications. But agonizingly poor performance is the last thing you want these users to deal with. The challenge Enterprise IT departments have been facing is that many applications have moved away from traditional Enterprise data centers secured behind corporate firewalls into the cloud or distant consolidated data centers. Coupled with a continuing trend toward supporting employees working remotely or accessing applications on an ever-growing variety of mobile devices, Enterprises have had to come to terms with performance and security issues that are the result of this changing environment.
At Cotendo we believe the time has come to address these transformations and extend the benefits of optimized application delivery including TCP optimization, compression, data de-duplication and SSL offload beyond the confines of the enterprise data center, ultimately delivering LAN-like performance regardless of where the application resides or the location of the device being used to access it.
From Silo to Fusion
Surprisingly the two common solutions often deployed to help enhance application performance, namely Application Delivery Controllers (ADC) and Content Delivery Networks have existed in separate, siloed worlds. They have helped to improve the performance of applications but the rise of user mobility and more dynamic and web applications requires a new approach in order to deliver predictable and consistent application performance with millisecond to microsecond latencies.
Gartner analyst Lydia Leong has discussed combining network optimization, edge caching and other advanced acceleration technologies like Front End Optimization (FEO) to leverage their synergies in an effort to achieve maximum performance.
Interestingly this is what we have been doing here at Cotendo. You may recall that we implemented Google mod_pagespeed last year to automatically optimize the code of HTML pages as they enter the CDN to provide maximum acceleration benefits to web application delivery.
Putting the Pedal to the Metal
Yesterday another major synergistic innovation was announced (and published by the New York Times), this time with our strategic partner, Citrix Systems, at Citrix’s Synergy 2011 in Barcelona, where we unveiled a unique hybrid networking solution. Known as CloudConnector, the joint offering comes to market just four months after Citrix formed a partnership with and made a strategic investment in Cotendo. Our new hybrid solution achieves performance improvements of 2-3x for web applications and up to 20x for mobile applications. How’s that for speed?
CloudConnector consists of Citrix NetScaler appliances residing within both Cotendo’s global point of presence network as well as Enterprise data centers, creating a secure, high-speed encrypted delivery tunnel that optimizes content for end users. The fusion of these once siloed solutions enables the acceleration of enterprise applications all the way from the end user to the data center. This approach ensures organizations have a streamlined and simplified method of deploying and supporting services as well as managing the network while reducing operating and capital expenditures and extending the security of application delivery.
Ultimately this effort illustrates how 1+1 = 3. Together Cotendo’s cloud-based acceleration service and Citrix NetScaler on-premise application delivery combine to form a more powerful solution than each would on its own.
Enterprises and content providers alike can now speed up and secure their B2C, B2B and B2E applications regardless of network, location or device while reducing CAPEX and data center bandwidth costs by up to 90 percent
Our new hybrid cloud offering is an exciting step for Cotendo and Citrix but it is only the first in what will be a series of solutions which will further advance the state of the industry allowing enterprises to work and operate more productively and efficiently across the globe. So hold on tight. This will be a thrilling and fast-moving ride.
Posted by Boaz Avigad on Tue, Oct 11, 2011 @ 05:09 PM
The use of cookies for storing user-specific or session-specific information is ubiquitous and can be found on all types of websites and for different purposes. Content providers rely on cookies to authenticate users, maintain user preferences, and apply other types of logic in order to provide a customized and persistent experience. Cookies are used in a way in which the information stored in the cookie is sent with each request to the content provider’s service. The service parses the information, applies the defined logic and carries out the required functionality. The delivery of content to mobile users poses new challenges for content providers, stemming from the need to identify which mobile device is used and deliver a differentiated experience for different devices or device types. Cookies can be used to provide effective solutions to these challenges in a number of ways as described below.
Mobile Device Detection
Detecting that a request has been sent from a mobile device requires the content provider to parse the user-agent string in the request header and look it up in some kind of device repository. Depending on how the service is implemented, it may be necessary to identify the type of device, e.g. feature phone, smartphone, tablet, etc. It may also be necessary to identify device-specific capabilities such as screen size (for site content and 3rd party ad delivery), touch-screen support, content format support (e.g. Flash, AJAX, etc.), wireless technology support (e.g. UMTS, LTE, etc.) and other characteristics. This is typically an expensive process in terms of both processing power and added latency and may be required once or more each time a user browses to the service.
Using a cookie to store the information needed by a service can ensure that the resources and time entailed in obtaining this information are required only once for the duration of the cookie’s lifespan, which can be a very long time. This is a strong contributor to reducing operational expenses and service latency.
User Viewing-Mode preference
Content providers create one or more mobile versions of their webpages in the hope of providing their mobile users with a less cluttered, faster loading and more usable browsing experience. But mobile users are not always satisfied with the leaner and limited version of the website and want to have control over how they consume the service. Content providers often solve this by adding “View Mobile Site” or “View original Full Site” links to their webpages. Maintaining a user’s preference throughout a session can be sometimes difficult, while maintaining it for following sessions can be nearly impossible without using a cookie. Placing a cookie with a user’s preference can provide a more satisfying user experience as the user doesn’t need to request for her or his preferred viewing mode each session, and the service doesn’t need to spend resources and time on establishing the default viewing mode and perform unnecessary redirects.
The viewing mode is just one example of the type of user preferences which can be effectively stored using a cookie. Inter-session preference persistency has many other applications.
Caching of Mobile Content
Content providers often use a distributed caching service, such as a CDN, in order to reduce the load on their core servers and to accelerate content delivery by delivering it from a location which is close to the users. Due to the differentiated nature of mobile content it can be difficult to effectively store and retrieve content in/from the cache. The decision of which version of the content should be delivered to the requesting device is typically carried out by the service itself at the location where it is hosted. But having a request reach the service’s core hosts defeats the purpose of using a distributed cache. The cache can be used effectively only when it is possible to establish the required content version at the cache’s location, close to the user. This can be achieved by placing a cookie with information identifying the required content version and by employing a distributed cache scheme which is able to store content with a secondary key (the first being the object’s URL) which can be produced from the information stored in the cookie. This type of caching isn’t natively supported in HTTP1.1 and requires the cache solution to implement the logic described above while still respecting the object’s defined caching directives.
Session stickiness
Web services often require session persistence in which a session’s traffic will always be handled by the same origin server. This can be achieved in different ways by using different types of logic at different locations along the request flow. The stage in the process where the request is analyzed and the required server is established can be very resource intensive and time consuming. By placing a cookie which explicitly indicates which server should handle the session’s requests; this process can be greatly optimized. This will of course require an entity within the flow to be able to parse the request header, retrieve the cookie value and forward the request to the designated server’s address.
Taking your cookies to the edge
The use-cases describe above are examples of how cookies can be leveraged to provide a more cost-effective and responsive web service. In order to enable these functionalities, the necessary logic of creating, storing, parsing and processing cookies needs to be implemented as part of the web service. By utilizing a service such as a content delivery network which is capable of creating and placing cookies at an edge location, close to the users, and implementing the required logic based on them, a content provider can benefit from the offloading of this process as well as the ability to react much quicker when caching or routing decisions need to be made.
At Cotendo, we have identified the need for providing a smarter edge delivery solution and implemented the capabilities described above as part of our Mobile Acceleration Suite (MAS) service. By utilizing Cotendo’s Cloudlet Platform technology, each of Cotendo’s Points of Presence (PoPs) is capable of processing requests, analyzing cookie information and making routing and caching decisions based on this data. Cotendo’s Cloudlet Modules are able to identify mobile device and mobile network traffic, and make decisions in real-time according to the content provider’s requested behavior. This gives us the flexibility to handle requests more efficiently and provides the content provider with ability to accurately control how our service handles each request.
We took our solution even a step further and provided Cloudlet Modules that are capable of generating and setting cookies on the edge themselves. This enables content providers to offload this functionality and even further reduce operational costs and improve performance.
The following example shows how the process of placing and utilizing a cookie for a distributed cache solution is implemented in Cotendo’s Mobile Acceleration Suite service. The flow of the first request, for which the object hasn’t been cached yet, can be seen in the first diagram below:
The browser sends a request to the nearest PoP (1) and is forwarded to the origin server (2). The origin server establishes the requesting device’s characteristics by looking up the user agent string, ‘iPhone’ in this example, in a device repository. The origin server generates the appropriate version of the content and sends it back to the user, along with placing a cookie based on which the device would be easily identified in future requests (3). The PoP stores the device-specific version of the content along with a key which is based on the value set in the cookie (4). The PoP then sends the replied object, along with the cookie, to the user (5). It is also possible for the PoP to perform the device lookup and to generate and add the cookie to the reply sent to the browser.
In following requests, as can be seen in the diagram below, the browser sends a request which now contains the cookie information (1). The PoP performs a cache query with the object’s URL and the key found in cookie (2). The iPhone version of the object is retrieved from the cache (3) and sent to the browser (4).
Utilizing a Cache Service which can execute the process described above can greatly contribute to accelerating a web service while dramatically reducing the traffic and computational loads on the content providers’ origin service.
Posted by Shlomi Gian on Sat, Oct 01, 2011 @ 07:19 AM
Similarly to previous years, GigaOM Mobilize conference in San Francisco attracted the right audience. Attendees included a healthy mix of technical and commercial folks from new startups and mature telcos. We had lots of great meetings and learned about what people are looking for in mobile content delivery. As a highlight, I participated in a panel discussion with Appcelerator and Urban Airship, two new mobile cloud platforms. The topic was “Profitable Gaps - The New Mobile Infrastructure Players”.
On the panel we discussed the huge opportunities present in the gap between demand for rich, dynamic, interactive mobile content and the capabilities of publishers, social network operators, game developers, and ecommerce sites to supply this content quickly and easily without inconveniencing end users. It’s a topic we like to call the “Anywhere Experience.” People consume content anywhere and they expect it to be rich and personalized. Static is boring and old. Rather, people prefer content to be rich and personalized (think Facebook Wall, Netflix queues, Path photo sharing apps, and mobile Shopping Carts).
How rich exactly content is these days? Check out Mobile Http Archive to learn how pages have gotten 20% larger in size in the last nine months. Over half of this increase comes from images. Delivering this sort of exploding content footprint to wireless devices requires an entirely new layer of infrastructure and technology. Because, quite frankly, the old world of CDNs and even content acceleration tools and systems built five or more years ago are not designed for this massive growth in rich dynamic content transmission and consumption.
That’s the opportunity we play in and its a huge opportunity. How huge? Based on Compuware (Gomez) latest survey, less than 20% of all mobile sites load in less than 5 seconds. Most mobile sites load in 10 seconds or more and sites that load that slowly have a 49% abandonment rate, according to Keynote Mobile Retail Index. That rate increases significantly for every additional second of latency. According to Keynote, some of the biggest retailers in the world actually have very slow-loading mobile sites. This gap between users need for speed and what mobile sites are actually delivering today is our primary business driver in the mobile space. And we at Cotendo are excited to be participating in building the new infrastructure to address this gap. In this trend we see big and small companies delivering localized, regularly updated, highly relevant content to thousands of different places and devices at the same time.
The new infrastructure around dynamic mobile content delivery will allow any size of company to deliver its content and its products globally and compete on fairly even footing with the giants of the space. This infrastructure will allow developers to quickly build apps, democratizing media creation in the same way blogs did a decade ago. The developers will be able to use entirely new classes of services to accelerate their apps and ensure that the sites of content providers, optimized for mobile or form-factored in iOS or Android apps, will show up on a handset or a tablet fast enough to satisfy the increasing impatient readers and consumers.
These services will include in-app advertising, in-app purchasing, extremely smart mobile acceleration services with embedded business logic on the edge of the delivery network, and new ways to streamline network traffic to reduce the latency that has remained stubbornly high in mobile data use cases. Most important, the companies building these services will partner with customers and other ecosystem providers. Because the value of every piece of the ecosystem will increase as the mobile ecosystem itself continues to grow and evolve.
So what is stopping this new “Anywhere Experience” from taking shape more quickly? Patents are a giant cloud overhanging anyone trying new models in this space, something made abundantly clear by the ongoing bidding wars over patent portfolios. The ecosystem is still quite fragmented with many pieces unable to talk to each other - cloud computing providers don’t work with tool companies don’t work with content acceleration companies. But more than anything, it’s the lack of market education, which is understandable. We’re still very early in this game.
So back to the panel. One thing I emphasized is the primacy of latency as the biggest problem for companies trying to build around the “Anywhere Experience”. Latency is the reason why mobile apps aren’t even more popular and is a huge break on innovation. With a faster delivery mechanism, undoubtedly developers would try many new things and the next killer application for mobile would arrive. I’d add, too, that latency will remain a killer despite the arrival of much fatter wireless data pipes (4G), bigger tower networks, and more backhaul bandwidth.
What’s more, the core issues causing latency lie deep with the protocols of the Internet (http, tcp). While we and others are making progress moving towards a more modern Internet, in the interim Cotendo is working extremely hard to deliver solutions that optimize the present Internet in ways to ensure that content gets to a handset anywhere in the world, over any network, delivering any applications faster - much faster.
Posted by Corinna Krueger on Thu, Sep 08, 2011 @ 07:17 AM
We spend a lot of time talking about our own customers. So its nice to put the shoe on the other foot sometimes and play the role of the customer. Cotendo was featured in today’s announcement “Cotendo Quadruples Customer Base and Goes Mobile” put out by Equinix, a large provider of connectivity and co-location services that has been a critical partner in Cotendo’s phenomenal growth. We have used their top-notch global facilities to deliver on the promise of our innovative technologies (and specifically on the promise of our mobile application suite). They have supported us beyond our expectations through our remarkable 400% customer growth curve over the past two years. Equinix, too, has helped us expand our global footprint.
Having such strong partners that can hold up their end without fail is critically important because customer expectations of performance are changing, too. As we are pushing deeper and deeper into the market for accelerating dynamic, real-time mobile applications, the margin for error anywhere in our network has grown smaller. That’s because reducing high latency on connected mobile applications is a far more complex problem than reducing high latency on standard broadband connections. Ido discussed the real reasons for mobile latency (hint: it’s not bandwidth) and we have discussed many of these same themes ourselves in other articles and videos. In a nutshell, weaknesses in standard Internet protocols, wide variations in network conditions and speeds, and wide variations in hardware capabilities conspire to make mobile content delivery less predictable and harder to guarantee.
At Cotendo, these types of hard-to-solve problems are what we thrive on and they also represent a real opportunity. In these problem areas, that is where the market for content delivery and site acceleration is growing and will continue to grow. Think mobile ecommerce, gaming, social networks, advertising – that’s where all the growth is and that’s where we are putting our R&D resources and pushing the technology envelope. Dan Rayburn’s post “Here's Why The Future Of The CDN Business Is In Mobile Content Acceleration discusses some of the trends and outlines why coming up with creative solutions to these hard problems facing anyone that delivers dynamic mobile content is essential to keep up with expectations and customer demands (both enterprise and consumer customers, that is).
And in the mobile area, we have come up with some very unique solutions from improving core Internet protocols (such as swapping SPDY for HTTP) to far more intelligent object caching which is done with information about what the device should optimally receive (we call it mCaching). We have moved decision-making closer to end users by applying business logic far out in the network away from origin servers. All of these steps have one thing in common - they reduce the frequency and density of data calls to the origin server. This speeds up mobile content delivery because its much easier to travel back and forth to a tower or to a nearby PoP than to go all the way back to the core Web server(s) that are the original and official content home. And this is why partners like Equinix are so important. They take care of our data center and co-location needs so we can focus on the good stuff, the technologies that drive real value to our business and drive real acceleration solutions to our customers - most lately, in the mobile realm.
But don’t take our word for it. Check out this upcoming webinar on Sept 14th with Redfin’s CTO Michael Young, the guy who put this real estate company’s mobile strategy on the map and turned it into one of the most popular mobile real estate clients around. And do let us know what you think. We’re happy to discuss ways we can eliminate your mobile slowdown pain and create better, more scalable systems for your mobile content.
Posted by Boaz Avigad on Tue, Aug 09, 2011 @ 11:10 AM
Web analytics tools are very popular with content providers. They help gain insight into different aspects of traffic arriving at their service and enable them to make better business and IT decisions. It is therefore extremely important to make sure the data collected and presented is accurate. For mobile data, at least, this may not be the case.
Some Background
The latest version of our mobile acceleration suite includes a set of reports, one of which is a report showing the distribution of mobile operating systems. Our reports are based on the logs we generate for traffic processed by our service. The following is a pie-chart representation of a mobile OS report for one of our customers for June 2011.
As can be seen, the leading operating systems are iOS, Symbian, Android and BlackBerry. The category named “Other” includes mobile operating systems which did not fit into any of the other categories. This includes mostly proprietary operating systems by Nokia, Samsung, LG, Sony Ericsson, Motorola and others.
The Case
A couple of weeks ago one of our customers contacted us to report a bug in our mobile OS report. The customer’s reason for claiming this was that our report was way off what his Google Analytics reports were showing. The customer sent us the Google Analytics reports and we could see there were significant differences between them and our reports. Since our reports were just released and Google is, well… Google, our immediate concern was that we were doing something wrong. So we started to dive into the data to find out where to problem lies.
The Road to Exoneration
First of all, in order to make sure I was comparing apples to apples, I had to clean the data from the discrepancies resulting from differences in how Google Analytics and our reports categorized information. After cleaning out the data I compared the distributions of the top five operating systems, as can be seen in the table below:
| |
Google Analytics
|
Cotendo Reports
|
|
iOS
|
43%
|
31%
|
|
Android
|
38%
|
19%
|
|
Symbian
|
15%
|
35%
|
|
BlackBerry
|
1%
|
12%
|
|
Windows
|
1%
|
1%
|
|
Total
|
97%
|
98%
|
|
|
|
For both Google Analytics and Cotendo’s reports the top five operating systems constitute nearly all of the traffic originating from an identifiable operating system (97% and 98% respectively). By “identifiable” I mean one of the well-known mobile operating system (iOS, Android, etc.). Proprietary operating systems which did not identify themselves by name in the user-agent string were not included. What is immediately evident is that iOS and Android make up a smaller share in Cotendo’s reports than they do in Google Analytics’. For Symbian and BlackBerry it’s the other way around.
The latter observation is the key to understanding why such large differences exist between the reports. Google Analytics relies on JavaScript and cookies to collect information. While this may work well for desktop/laptop browsers, it is a huge drawback for mobile browsers. Not all of Nokia’s Symbian-based devices support JavaScript, and RIM has turned off JavaScript support by default on most of its BlackBerry devices. This causes Google Analytics to identify only a portion of their traffic. The result is that the traffic share reported for these devices is much lower than it should be. On the other hand, iOS and Android operating systems are shown to make up a much larger portion of the traffic than they actually are.
Bottom Line
Web Analytics tools such as Google Analytics are very valuable but should be treated with caution when drawing conclusions about mobile traffic. Mobile devices, in many cases, are not JavaScript or cookie friendly and reports relying on them could be greatly distorted. Log-based reports are much less susceptible to such “blind spots” and could be a more reliable alternative for collecting information and generating mobile traffic reports. If you are currently using web analytics tools to gain insight into your service’s mobile traffic, it would be advisable to cross-reference the results with additional tools, preferably log-based.
Posted by Corinna Krueger on Wed, Aug 03, 2011 @ 12:42 AM
Mobile load speed directly affects your mobile site's or connected mobile app’s ability to convert browsers into buyers. Culprits of slow loading mobile sites apart from varying network speeds are too many page objects, bloated images, and poorly optimized code. The dozen of requests needed for these objects to load compound latency, leaving your users with nothing but frustration and shooting your sales in the head.
I had Shlomi Gian, our General Manager for Mobile, interviewed by TMCNet's Group Editorial Director Erik Linask last week. We are diving into the challenges of mobile Web performance, use cases, Cotendo's vision for accelerating the mobile Web, and how we make it happen.
Click the video below to watch. Enjoy!
Posted by Shlomi Gian on Tue, Jul 05, 2011 @ 09:13 AM
DoAT (http://www.doat.com), a company aimed at revolutionizing mobile search, announced last month the availability of its breakthrough mobile search experience. The DoAT platform changes mobile search in three fundamental ways - the way results are delivered (apps instead of links), the sources delivering the results (direct from publishers) and the manner in which results are ranked (via DoAT's "social radar").
DoAT deconstructs the traditional search paradigm of web links and instead merges search and apps in real-time, delivering the top web apps that specifically answer any query a user posts. The result is a fundamentally different experience of search that shatters old concepts.
“From the minute we started working on our search experience, we’ve marked performance as a key factor in our overall experience. Cotendo helped us gain a significant performance boost we needed to be out of the gate. We’re excited to be working with Cotendo on their mobile acceleration offering and look forward to further cooperation” says Joey Simhon, CTO and co founder for DoAT
Posted by Shlomi Gian on Wed, Jun 22, 2011 @ 06:33 PM
Every day hundreds of millions of people post something on a Facebook "Wall". It can be on their own wall or on a friend’s wall. It might be a picture, a link to a Website or video, or just a simple comment. So why not record a voice message to go with your post to better communicate emotions and feelings in your own words and in your own way? Wouldn’t you rather here what your mother has to say about the latest video you posted of your son’s baseball game? Or a co-worker’s funny reaction to a picture you post of a beautiful sunset in Hawaii?
That’s the idea behind startup VocalWall, which provides a cloud based platform accessible through a website and/or a mobile application (iPhone and Android) for users to easily record and post voice messages to social networks, comment boards and other places where friend gather and emotions and opinions are expressed. For now, VocalWall is allowing users to record voice messages (VocalStatus™) and post them for private or public replay on their Facebook wall. Sounds good, right? Apparently that’s what lots of people thought.
“When launching the service earlier this year we planned for moderate growth based on viral marketing. Soon after we realized that we could not control usage patterns and that Facebook users access the service globally,” Says Meir Ansher, CEO and Co-Founder of Vocalwall. We needed a quick solution that will allow us to keep growing our community while providing good and consistent user experience all over the world. Cotendo helped us accomplish this in a matter of days. Using their content delivery platform allowed us to focus on our core business. We owe them part of our success so far.”
On short notice, Cotendo integrated VocalWall’s audio playback services into its platform and used Cotendo’s Mobile Acceleration Suite to dramatically scale VocalWall’s playback capabilities and provide the company with industrial-strength network delivery to ensure that no one, no matter where they are, would suffer from excessive playback latency causing a poor user experience. After a user initial uploads an audio recording to VocalWall’s data base, Cotendo’s CDN will cache the playback stream in the closest PoP to an end user and then accelerate the stream to make sure VocalWall’s recordings come back crystal clear. This also minimizes round trips back to VocalWall’s Origin server. From what we know, this is a novel applications for CDNs. “Integration with Cotendo’s network was very easy – almost painless,” said Ansher.
An added benefit that VocalWall gained was out-of-the-box real-time analytics with Cotendo’s reporting tools. Integrated into Cotendo’s standard management console, the analytics reporting allows VocalWall to see key details of traffic trends including geographic distribution, device distribution, and network quality (which may affect playback). We aren’t compressing the audio – yet. But that may change; we do offer dynamic adaptive image compression that matches network conditions with different levels of image quality we serve.
So how will they make money? We do like our customers to be profitable, of course. One possible model is based on message playback (there are others). Individuals will use the service for free to post on their Facebook walls. But brands with active Facebook presences will pay to add this audio recording and playback capability to their pages. Think of, for example, the NBA adding recordings of user-generated fan reactions to the Dallas Mavericks winning the NBA Championship. These brands would pay for each playback consumed by the end user. For now, it’s all Facebook but over time playback could come on any number of channels including a mobile application, web site or Facebook. Considering that all major brands are aggressively marketing on Facebook, we think this is smart play.
Those big brands, naturally, want assurances that their message is heard loud and clear by whoever pushes the VocalWall playback button. That’s what we’re focused on at Cotendo – making sure VocalWall’s messages are played back quickly and with crystal-clear sound. With our best-in-class CDN and industry leading Mobile Acceleration Suite, along with other entirely unique features like real-time analytics of content moving over a CDN, big brands can rest easy knowing that VocalWall can handle any traffic spikes when some audio clip goes unexpectedly viral. This allows VocalWall’s team can focus on marketing the company, closing deals, innovating on their own technology while we handle and accelerate the delivery of their content to any location on any device.