Skip navigation
All Places > In the Limelight Blog > Authors smehrotra

In the Limelight Blog

7 Posts authored by: smehrotra

Recently, Volkswagen AG (VW) made the news not for its cars but for all the wrong reasons. A “discrepancy in installed software” in its diesel cars led to incorrect emission readings, violating several environmental laws across the United States and impacting over 11 million cars worldwide. How severe was the damage due to “dieselgate”? Within days VW announced its plans to keep aside over EUR 6.5 billion to cover costs and service-related expenses towards the software glitch in its diesel automobiles. The CEO resigned amid the scandal. The damage was not just limited to the environment alone – the brand itself is suffering from customer/prospective customer mistrust and potential abandonment. Even the stock market reacted sharply to the issue and the VW stock nosedived. While the software glitch could be patched up rather easily by bringing the car to a service center, it could be months before all the vehicles on the road can be updated to “fix” the issue.  This clearly points to a severe lack of Over-The-Air update strategy.


While this software discrepancy is neither the first nor the last of issues related to software glitches, there is a pattern on how the traditional automobile industry responds to situations like these


All of the below recalls in 2015 were due to software glitches:

  1. Ford recalled over 432,000 cars due to software issues
  2. Fiat recalled 7800 SUVs over software glitches
  3. Fiat-Chrysler recalled over 1 million vehicles to prevent them from being hacked
  4. Toyota recalled over 63,000 hybrids due to software problems
  5. Group of “ethical hackers” were able to remotely control and “kill” a Jeep Cherokee


What is the pattern that we observe?

  1. Manufacturer discovers an issue in the vehicle attributable to a software glitch
  2. Company issues a recall and notifies customers
  3. Customers bring vehicles to service centers for a “software update” or fix
  4. Process takes months to complete and millions of dollars
  5. Company’s brand and/or reputation is damaged


Is there a way to optimize this whole chaos, and in turn prevent customers from discomfort, and save the automobile manufacturers millions of dollars?


If we look around, the problem has already been solved. In February 2014, the U.S. National Highway Traffic Safety Administration published two recall announcements – for the same issue affecting automobiles manufacturers. The vehicles were at risk of catching fire, and there was a safety issue due to a software problem. While GM had to spend millions on 370,000 vehicles recalled over the next few months to update the software at its designated dealerships, Tesla conducted Over-The-Air-Updates of the software system - an overnight fix.


So, what did Tesla do right here?

  • Its cars are Internet enabled not just for the “infotainment” systems
  • It could troubleshoot the cars remotely
  • It could update/patch the software Over-The-Air (OTA)
  • It didn’t require customers to bring vehicles to dealerships
  • It expedited the time-to-service and potentially saved millions of dollars


There are various challenges to execute on an OTA strategy: development of these patches, distributing them to dealerships, making them secure, and finally delivering/updating them successfully without errors on a global scale at a moment’s notice. OTA updates need not be just limited in scope to vehicle recall issues due to software glitches either. A broad OTA strategy can enable the automobile manufacturer to expedite updates, patch intrusion or security risks, and enhance the user experience. OTA software updates for automobiles should not be an afterthought, but a well-defined workflow to optimize the whole process and make it seamless: for automobile manufacturers to transparently provide the update, and to the customers to seamlessly update their vehicles wherever they are.


See how Limelight helps companies deliver millions of software updates globally: both wired and over the air.

We would love to hear from you and continue the conversation on Limelight Connect. Keep the conversation going on Twitter with @LLNW and @_AllThingsMe.

If you have ever downloaded anything from the Internet – be it a mobile app, a software patch, or an application – there is a high likelihood you had a frustrating experience at some point.


That’s because most companies providing updates or software files only think in the context of the download speed of the file being delivered. They confine their performance and reliability measurements just to the file being downloaded, but fail to see the big picture.


While file download speed, reliability, and availability have a major impact on your end-user satisfaction measurement, they do not constitute of the whole download experience.


In our recent State of Digital Downloads survey, we found out that almost 30% of users blamed the brand (your company!) in the case of a frustrating download experience. In fact, 40% of users switched context or abandoned a webpage within 5 seconds of delay.


Believe it or not, your end users may be getting impacted by a slow download portal. Your Beta software version that was supposed to drive early adoption and pave the way for future product success? It’s entangled in the mess of a slow download portal.


Download the report


This said, it is a very addressable situation.


Software download portals should not be perceived as an afterthought. Everything that goes on in that portal – the download manager, product usage videos, security – are all part of the end user’s experience. For example, if your download portal requires a login page (or in other words is gated), how slow is the login page?


Using dynamic web acceleration technologies like Limelight Orchestrate Performance, you can drastically improve the time it takes for end users to log in and access the files they need. As a part of our latest solution to address these challenges for Software and Device Manufacturers, we not only accelerate the download portal, but also make sure your training videos are accessible for all devices, and the geo-restrictions are in place according to policies you set. It is the most advanced end-to-end software delivery and device update solution.


Every day, Limelight helps companies use our intelligent technology to solve their global business challenges. We would love to hear from you and continue the conversation on Limelight Connect.


Keep the conversation going on Twitter with @LLNW and @_AllThingsMe.

Devices exist all around us. In fact, Cisco™ predicts that there will be 15 billion Internet-enabled devices out there by the end of 2015.


That wasn’t always the case. The 2000s were the decade of mobile. Businesses focused on enabling smartphones as the most important way to reach an audience.


Today, however, companies that are focused only on reaching smartphones are missing out on almost 50% of their potential audience. As an example of the expanding universe of devices – and the business opportunity in reaching users across every one of them – the U.S. National Highway Traffic Safety Administration published a series of high profile recall announcements in February 2014. Two major auto manufacturers were forced to address the recall by making repairs at dealerships nationwide, at a high cost to business. Smart car maker Tesla™, on the other hand, conducted as an “over the air” software update to solve the problem, addressing the issue with dramatically fewer resources and on an accelerated timeline.


In fact, Tesla releases software updates every 34 days for their cars—achieving in a matter of days what used to take months or even years to accomplish – at a comparatively lower cost.



Today’s smart devices range from connected cars, to refrigerators, to televisions… There are just too many categories to list!


And your audience is not ready to compromise on their experience downloading software to any one of them. Taking full advantage of the business opportunity offered by reaching every connected device means delivering on that expectation every time.


It starts with receiving and pushing timely device updates, but companies face many complex challenges throughout each stage of their workflow: development, beta, and finally delivery and update.



Sharing large files or software binaries across globally distributed teams is no small task. Teams are more virtual than ever before, but your data centers are mostly centralized; according to one study, almost 64% of the companies surveyed had developers outside of their main campus. The inability of a distributed workforce to access files or share them quickly, reliably, and efficiently across geo-locations negatively impacts productivity.


Beta programs

Software beta programs are difficult to run at global scale. Beta programs are generally restricted to certain geographies or in some cases even certain companies. Protecting your intellectual property is critical to your businesses success. So is ensuring a great download experience. Your end users might not even download the beta software if the download portal is too slow; in fact, 20% of users report that they will abandon an online experience just after 3 seconds of delay. The ideal download experience is comprised of a download portal, fast download, and error-free files.


To learn more  download our research report


Delivery and Update

Delivering software and device updates both securely and reliably is a big concern. Gone are the days when end users would wait for their devices to update. The industry is moving toward a paradigm called “continuous streaming updates.” Companies like Sophos™ deliver thousands of updates every month to their end users’ devices; learn more about how Limelight is helping Sophos™ in this case study . What happens when a company like that is unable to satisfy the end user? According to recent Limelight research, over 30% of users blame the business brand for a failed or interrupted download experience.


new image pic10.png

To learn more  download our research report


Future proofing to be able to deliver and update to these devices and the devices of the future is complex, costly, and impractical to do yourself.


Today Limelight announced a new solution specifically designed to enable fast, reliable, and secure delivery of software and device updates with the most optimized workflow. The Limelight Orchestrate Solution for Software and Device Manufacturers was uniquely developed to address precisely the challenges that face the market today.



The solution offers network reach, intelligence, and cloud-based technology to enable your business more potential customers with greater efficiency than ever before:


  • Global Private Network: Limelight operates one of the world’s largest private networks with industry-leading cache efficiency, scale and availability that ensures a reliable, consistent download experience anywhere, anytime.
  • Device Detection: Smart edge device detection ensures the correct updates are delivered to every connected device irrespective of whether it is a wired or over-the-air (OTA) update.
  • Zero-Rated Traffic: A unique solution that enables toll-free content delivery over carrier networks, allowing end-users to access content without worrying about data-usage related bill shock.
  • Cloud Storage: A globally distributed origin storage solution that integrates with Limelight’s CDN to optimize worldwide delivery and management of media and software content by positioning content close to end-users, simplifying storage and lowering operational costs.
  • Robust Security: Global secure delivery (SSL) capacity, geo-fencing for regional access rights and DDoS Attack Interceptor for attack mitigation combine to protect digital file delivery against downtime, loss of business and negative impact on brand reputation.
  • Actionable Insights: Analyze user activity like content popularity, device usage and regional information to provide a better download experience.


Every day, Limelight helps companies use our intelligent technology to solve their global business challenges. We would love to hear from you and continue the conversation on Limelight Connect.


Keep the conversation going on Twitter with @LLNW and @_AllThingsMe.

Originally published October 1, 2014


Yes, it’s that time of the year again. The approach of the holiday season brings cheer to all faces – but if you are in charge of the operational side of a web application or website, it’s the time of year that you secretly dread. The 3 am support calls, the fire drills, the back-up server activation… we’ve all been there.


If this is you during the holiday season, you may not want to miss this blog!

holiday season shopping.jpg


Most companies put code changes under a strict lock-down mode (typically by the end of Q3) and rarely make changes that affect the stability of their systems until well after the holidays are over.


However, that does not mean you can’t test the operational readiness of your systems to make sure that servers, application resources, databases, and the coffee machine are all ready for the holiday traffic. In this blog post, we will review the various readiness tests you generally run and provide a list of key questions to answer as you ramp up for holiday traffic.


Before we jump into the details of how to conduct a readiness test, let’s do a quick glossary check; it is important to understand what kind of readiness test is best suited for your environment, and application, as you prepare for the holiday onslaught.  There are three types of tests to consider running.


  • Stress Test: Testing the limits of your resources by an artificial load over time to a point where things break
  • Load Test: Creating artificial traffic on your resources to make sure things work fine when traffic increases are anticipated
  • Spike Test: Trying to create a sudden traffic spike just to assess how the system handles sudden peaks


Each of these is considered a volumetric test; you can find more details about various testing types here.Now that we know what sort of volumetric tests might be useful, here are a few key questions that site reliability engineers should consider when planning a holiday readiness test for applications that use a content delivery network (CDN):


Am I testing the limits of my origin resources or the capacity of the CDN?Your test should measure the capacity of your origin resources. Because Limelight has virtually unlimited delivery capacity and built-in intelligence to distribute and automatically manage the load, there are virtually no CDN capacity limits to test.


Am I testing how much the CDN is able to insulate me from the sudden spike by offloading requests?While this is an important question, it is difficult to predict. There are many dependencies, like how much dynamic content your website has, and how the cacheability headers are set. We at Limelight often see customers achieve over 98% cache-hit ratio!


How am I going to measure the impact (of my tests)?Using third party testing platforms like Gomez and Keynote for conducting backbone testing and Real User Measurement (RUM) testing providers like Cedexis to measure user impact is the right way to go.


What are some of the metrics that I should look at?Some of the key metrics that typically get impacted during holiday traffic are:

  • Availability
  • Response Time
  • Errors

Am I conducting a truly realistic load test to mimic the real world traffic pattern?


Most times site reliability engineers are not conducting a true to life test. We may be sending traffic from one location (or even one server) in a specific city to the CDN. This is far away from the real world situation where your customers pound the CDN/origin from multiple locations simultaneously.


And what about multiple devices? Most users don’t necessarily sit behind a laptop and access your websites. There is a complex mix of devices like tablets, smartphones and even gaming consoles now!


Have I set goals and success criteria for my tests?


Many companies conduct arbitrary volumetric tests to gauge readiness, they don’t have set goals. Goals those are important to their business. For example:


“The pages should load in X seconds”


“The availability should not drop below X%”


The right approach is to set a goal that is closely aligned with your business metrics, like revenue goals.


Am I testing the right transactions that the typical users would execute?


When running tests, it is important to mimic what most real users on your website would do as closely as possible. For example:


“Are your customers able to view all the images on the catalogue—not just one?”


“Are they able to successfully add multiple items to the shopping cart without issues?”


If you are a retailer, testing just the home page may not be enough.


These are some of the key questions that you should consider before planning a holiday readiness test. Luckily we at Limelight have extensive experience and capacity to handle sudden traffic spikes and make sure the holiday season traffic translates into more revenue for you. We would be happy to speak with you and answer any specific questions that you may have about preparing for holiday traffic.


To know more about how we can help you solve today’s complex application delivery and optimization challenges please contact us or follow up with us on our online community!

Happy selling!

Originally posted July 9, 2014


This year, like every year, the best minds in the web operations and performance arena came together to learn, exhibit, and exchange ideas that impact digital experiences at the Velocity Conference in Santa Clara, California.


The key theme this year was – Building a Faster and Stronger Web. With 2000 attendees, over 100 speaker sessions, and 75 exhibitors, it was indeed the perfect platform for Limelight Networks, Inc. (Limelight) to showcase its exceptional technology, speak with experts at the front lines of web performance, and educate customers about the latest web performance optimization techniques.


The below social media post conveys the show’s significance to people responsible for web performance:


“Velocity is the conference I always wanted. Instead of focusing on one particular product or technology, it focuses on the true problem of keeping websites fast and available, which a lot of us have to deal with.” —Peter Zaitsev, CEO Percona Inc., co-author of High Performance MySQL


What We Heard


The attendees at the conference were enthusiastic to learn and share real-world best practices for optimizing their web applications, focusing on the complex topics of the performance of Ajax, CSS, JavaScript, and images. (These are the core components that a typical website is made up of, and what makes them dynamic and personalized.)


But how do you know if your applications are truly optimized? Performance testing was another widely discussed topic, with a focus on real user measurement (RUM).


Finally, attendees were hungry for best practices around optimizing performance for mobile and responsive design.


JavaScript Is the Assembly Language of the Cloud


Let’s focus on JavaScript for a moment.


In a keynote session by Scott Hanselman of Microsoft®, titled Virtual Machines, JavaScript and Assembler, he said – “JavaScript is the assembly language of the cloud.” (You can watch this session on-demand here.)


We at Limelight Networks see, optimize, and enable delivery of some of the most complex web applications and sites every minute. While JavaScript has empowered developers to do things at the client-side (your browser or device) that were not possible only a few years ago, it has also added more complexity and performance challenges to the webpages.


Many Velocity attendees came in thinking that their applications and websites were already optimized but found out about best practices and techniques that they weren’t even aware of. The also found that applying some of the best practices in a cookie-cutter way could actually impact the performance of their web applications negatively. Many engineers representing their companies were amazed to learn that their websites that they thought were already fast and optimized could be further fine-tuned for performance gains that ranged from 30% – 100%. They were eager to learn more and dive deep into the methods and what they could do better.


An intelligent service like the Limelight Orchestrate® Performance platform provides power to the developers to create more “client side heavy” JavaScript code without having to worry about optimizations, resulting in that superior user experience that Velocity attendees are hungry for… especially for JavaScript-heavy content.


Getting Beyond Page Level Metrics


Modern browsers are updating at a pace that is almost impossible to keep up with. While it is great to optimize code and create a powerful experience on the users’ devices and browsers, how can we make sure that experience is consistent?


In a keynote session on Real User Metrics (RUM) by Buddy Brewer of Soasta, he focused the attention on the complexities of browsers and page components. More importantly, how these components load at various times can point to which components are slowing pages down.


Real User Measurement (RUM) isn’t just about the page level metrics anymore. We can now collect real user data at the object level, find slow page components, and keep third parties honest. For example: If your users are complaining about poor performance, you have ability to isolate it to a much more granular level than possible before. Using the waterfall charts produced by a typical RUM test, you can find that small problem thumbnail image coming from an ad-network as the source of slowness on a specific browser, device or network.


RUM is an important counterpart to synthetic testing, where nodes sit on the Internet backbone to measure raw speed but give an incomplete picture of the user experience.


The key questions that RUM can answer are: Which page components affect perceived latency? Is that JavaScript object a single point of failure? Or is it that external CSS stylesheet causing the slowdown?


Using RUM to measure your content is critical to your business.




Web experts and prospects alike were keenly interested to know more about how to successfully solve the challenges of today’s complex websites, particularly mobile. Many speakers at this year’s conference focused on “Mobile First” and responsive design.


Sessions covered the challenges of mobile performance and the difficulty ensuring a great user experience on all devices.


One session covered building out a device lab and testing all the browsers. Tammy Everts (Radware) shared some interesting data on how views react to slow performance: 23% curse at the phone, 11% scream at their phone, and 4% throw their phone. (Our very own CMO has been known to bang his phone on the table as if hitting it like an old rabbit ear TV set would reset it.)


Overall, these sessions were very well attended and it was clear that attendees see web performance on mobile devices as one of their key challenges.


Our Take from the Show Floor




Major high-tech companies like LinkedIn®, Facebook®, and Google® were showcasing their work and invited people to challenges and trivia contests about web performance problems. We at Limelight were honored to be among them.


Web experts and prospects alike were keenly interested to know more about how to successfully solve the challenges of today’s complex websites. Our CDN and performance experts had key conversations around mobile, performance, and web application best practices. Many attendees were amazed to learn a CDN like Limelight can not only accelerate and cache static content, but also accelerate dynamic and personalized content. Front End Acceleration (FEA) techniques also caught they eye of many technologists. These web professionals were amazed to see the possibilities introduced by a more complete approach to web performance optimization, combining static object caching, dynamic content acceleration, and front end acceleration.


While JavaScript, RUM, and mobile conversations dominated the conference, one more thing drew long, inquisitive lines at our booth – the presence of a caricature artist. In a more lighthearted moment, one of our customers – a prominent social media network – brought in its whole team to have their caricatures drawn.


velocity caricature




Velocity is about the people and technologies that keep the Web fast, scalable, resilient, and highly available. From ecommerce to mobile to the complexities of cloud, companies need to create a faster, consistent web experience globally. And companies like Limelight are dedicated to enabling that for you.

To know more about how we can help you solve today’s complex application delivery and optimization challenges please contact us and follow up with us on our online community!

Originally published July 2, 2014


This is blog post #7 in our blog series #OptimizeDigital, where we explore themes based on our newly released book Optimizing the Digital Experience (available for download here). You can catch up on the previous post here.


Twitter_logo_blue Tweet this post: Currently reading: Are You Getting the Most Out of Your CDN?


Looking to make your website faster? Then you have probably come across content delivery networks (CDNs). A CDN acts as your instant global infrastructure, distributing and caching your content on its network of servers, which are located closer to audience members everywhere than your infrastructure could ever be—at least without spending unthinkable capital and developing deep technical expertise. (Consider that Google is one of a very few companies that operates its own global CDN.)


Caching is a key performance strategy for any CDN. The general rule: the closer your end user is to your content, the better. Shorter distances minimize latency and cut down on unnecessary network hops that can lead to packet loss and retransmissions. Caching enables those short distances.


Once upon a time, that was enough.


No more.


Today’s CDNs must be able to handle today’s content. That means addressing trends in media, user consumption, and web development so you can superbly deliver:


  • Content to mobile and other connected devices
  • Content for every browser
  • Dynamic (personalized) content
  • Video, video, and more video


Do you have any of these content types? A quick content audit  will tell.


This blog post introduces you to optimization techniques that traditional CDN providers cannot offer. These techniques accommodate the more complex delivery needs of organizations operating in today’s digital environment.


Dynamic Content Acceleration


This is important: most CDNs are capable of delivering dynamic content, but very few are capable of accelerating that delivery. The difference between the two has a real impact on your bottom line.


Unlike static content, which is the same for every user and can be efficiently cached in an edge server and counted on not to change for long periods of time, personalized content must be uniquely generated each time it is requested. That means every request for dynamic content must travel all the way back to the origin servers where you store it, and the object requested travels the same distance to your end user; a network round trip, at minimum. The network round trip can occur on the public Internet, or a private network. (See Chapter 5 of Optimizing the Digital Experience  to learn more about the difference between the two.)


Personalized content providers (anyone delivering content that has to be refreshed on request like a real-time bank account balance, custom search results, or personalized web experience) face a challenge: How do I achieve the kind of performance gains that static object caching can yield, given the fact that my dynamic content can’t be cached in the same way?


Specialized CDNs can help you overcome the dynamic challenge with specific optimizations across the middle mile—in essence, opening up a bottleneck-free world where data can travel freely between your servers and your online audience. There are two main ways to get this done: route optimization and TCP acceleration.


Route optimization means selecting a delivery path that produces the best possible performance. This delivery path can be the public Internet, or a private network like Limelight’s. (In addition to controlling the flow of your data, privately operated CDNs control everyone else’s traffic on their network as well to avoid congestion.)

Transmission control protocol (TCP) optimization is really a combination of techniques to improve the performance of your dynamic content as it travels across the networks between you and your audience. Since TCP sets limitations on how your content can move across the Internet, specialized techniques are required to either respond to or overcome those limitations. But rather than hand coding these optimizations into your content, a CDN can actually automate decisions and executions based on factors like your user’s browser, device type, packet loss, and timeouts all in real-time.


Front End Acceleration


Nobody can click, browse, search, or transact on your site until all of its components have loaded.


But components do not necessarily load in the order that the end user cares about them. A browser does not know to load the “buy now” button before all scripts have run. Specifically with respect to JavaScript, your page may behave differently when some script loading is deferred. For example, if the script is written to expect certain user actions, such as onclick or onkeypress, these actions will not be triggered until the page is completely rendered.


Think of it this way: If your audience could prioritize the loading of objects on your page, what would they put first? Probably the stuff they care about the most, like product images or headlines.


But even if your development team is already optimizing code to load content based on user preferences, the number and evolution of browsers on the market makes that a self-defeating exercise. Shaving milliseconds off load times for the latest version of one browser may add milliseconds on the next release. There is no cookie-cutter model to apply; some optimizations can actually slow down your performance.


Front end acceleration (FEA), also known as front end optimization (FEO), analyzes and optimizes your code to load content more intelligently, based on user expectations. This is done with a variety of techniques that can be applied to your content as well as any third party content on your site, like advertisements.

By placing the responsibility of staying ahead of the most recent trends and changes in the browser market on your CDN, you free your development team to focus on primary business goals like developing new features and functionalities. A mobile banking app developer, for instance, should be able to improve product usability without worrying whether downloading more client-side scripts will reduce application performance or introduce more latency.


Just as there is no cookie-cutter approach to optimizing code for every browser, compressing for all browser types is difficult. There are many browsers that are fairly old and do not support compression (for example, some variants of IE 6).


Your CDN should be capable of compressing content for the browsers your audience uses to free up bandwidth and storage; compression generally reduces the file size by about 70%. Anything text based can be compressed including XML and JSON; most websites compress their HTML documents as well. Compression reduces response times by reducing the size of the payload. It is also worthwhile to compress scripts and stylesheets, but many content providers miss this opportunity.


Currently, Gzip is the most popular and effective compression method. Approximately 90% of today’s Internet traffic travels through browsers that claim to support the Gzip approach. Note: Gzip compression should not be applied to image and PDF files because they are already compressed. Trying to apply it in these cases not only wastes CPU, but can also potentially increase file sizes. Still, compressing as many file types as possible is a simple way to reduce page weight and accelerate the user experience.


In Sum


Traditional CDN services alone are just not enough to improve performance in today’s digital landscape. New techniques have emerged, and R&D teams are working as you read this to optimize delivery for the next generation of devices, browsers, and user behaviors.


As anyone who has been using the Internet for more than a few days knows, it is a living network that does not ever stop evolving.


Before web pages grew from an average of 100 kb in 2005 to 1800 kb in 2014, in-house performance optimizations were enough for most organizations. Now, before things get any more complex, stop and think about what it will take to create a great digital experience for your audience.

Do you know what it takes?


If not, you can catch up on the #OptimizeDigital blog series or read the book Optimizing the Digital Experience.


In this blog series, we have covered the bases in terms of the optimizations required for content providers like you to create the best user experience. In the next and final post, we will talk about making the case for performance to others in your organization. Thanks for reading.


Want to chat? Find me on Twitter, drop me a note, and join the  [#OptimizeDigital] conversation on Twitter.


Want to learn more about web performance? Download the book Optimizing the Digital Experience for an in-depth look at the topics presented in this blog series.


Click to download


Twitter_logo_blue Tweet this post: Currently reading: Are You Getting the Most Out of Your CDN?

Originally published June 11, 2014


This is blog post #6 in our blog series #OptimizeDigital, where we explore themes based on our newly released book Optimizing the Digital Experience (available for download here). You can catch up on the previous post here.


Twitter_logo_blue Tweet this post: Currently reading: Instant Global Infrastructure? What a CDN Does for You.


Though Internet service providers (ISPs) and mobile networks provide increasingly fast connection speeds, a host of variables along every mile of the content delivery path means that those speeds are not consistent in the real world. And even if they were, other factors—including availability, scalability, and user device—could hurt your online performance.


Fortunately, there are plenty of optimizations to consider: caching, strategic storage, dynamic content acceleration, and front-end acceleration… to name a few.

Have you performed a web content audit? A content audit is a quick and easy way to map your content to these different optimizations, identifying which ones are right for you. (The previous post in the #OptimizeDigital series is all about how to conduct your audit. Read it here.)


In this post, you will learn about optimizations for:


  • Delivering static content
  • Large and small objects
  • Rich media (like video)


We begin with an overview of how content delivery networks (CDNs) work. A CDN connects two points on the content delivery path: your origin (the servers where you store your content), and the edge of the networks that your users rely on to access content in the “last mile.” A CDN connects your content to user access networks either by an optimized public Internet path, or across a privately owned network.


Recall that this middle mile is full of performance-killing threats. A CDN routes your content across an optimized path to its own servers, which are located right next to your end user access networks. CDN servers are located in clusters referred to as points of presence (POPs). POPs interconnect with one another across the public Internet or a private network; they also connect with last-mile access networks in a specific region or around the world, depending on the scale of the CDN. Private networks (like the one we operate here at Limelight) help you avoid the hazardous public Internet. Basically, a CDN can extend your infrastructure as far and wide as required to reach your audience on any device.


Static Object Caching


For large and small object delivery


CDNs offer static object caching: storing copies of your most requested static content in cache servers across POPs near end users, refreshing it as needed, and ensuring availability through replication and backup. (Advanced CDNs, which we will discuss in the next blog post, can offer more than just static object caching, but consider it the point of entry.)


But the real value of caching is in how much of your content a CDN can retain at a global scale. Since static content, like the logo on your website, does not often change, it can stay in the CDN’s cache servers almost indefinitely. If the CDN has a high cache hit ratio, meaning that the content is available most of the time it is requested, then your content is within immediate reach 95%+ of the time. Unfortunately, some CDNs have a low cache hit ratio—so when a user requests your content in cache and it is not available, the request travels through to the origin until the content is found.


Cache hit ratio is a true parameter of caching performance and depends highly on the CDN’s architecture. Because of our densely-architected metro POPs, we at Limelight Networks are able to maintain a cache hit ratio above 98% (one of the highest in the industry).


Static object caching with a high cache hit ratio gives you two important advantages. First, you avoid so-called round trips, or multiple requests traveling back to your origin for content, which conserves valuable bandwidth. Second, content is right there when a user requests it, reducing the latency associated distance between your content and the end user requesting it.


BOOM—faster website!


Both the small and large static objects that make up your website can be cached, from text to rich media on demand. “Whole site delivery” is a term used to refer to the delivery of both small and large objects as well as the containers of those files such as HTML. Delivering whole sites is more challenging because they present a heterogeneous mix of cacheable and non-cacheable content. The main HTML file could be non-cacheable, but the various components that make up a page (scripts, images, stylesheets) could be cacheable. This requires intelligent cache management and header parsing rules at the edge to differentiate among complex content needs.


Cloud Origin


For better delivery of media, software, and more


Recall that before a CDN can cache your content, the content has to be retrieved from your origin. If your origin is in Las Vegas and your users are in London, then of course you want to cache content in London. But the content has to get there from Las Vegas in the first place. Any time there is a cache miss—meaning your content is unavailable in cache for any reason, such as unpopular content that has fallen out of cache—the request travels back to Las Vegas. That distance comes with a performance penalty, especially for large files like rich media and video.


When it comes to performance, where you store your content matters.


While many CDNs offer traditional storage in their multiple POPs, cloud origin provides something different. Limelight’s purpose-built cloud storage that is optimized for cloud origin allows content providers to take advantage of Limelight’s global content delivery network—storing origin content in strategically located POPs and uploading data locally and replicating it worldwide.


Limelight’s cloud storage infrastructure can ingest content from you locally through the nearest POP, but when origin data is requested (like the first byte of your website or a repeat request on a cache miss), the request does not need to travel from London all the way back to your origin in Las Vegas; the content has been pre-positioned in the London POP, maybe even in the same rack as the server where it will be cached for future requests, cutting down on round trip times and latency.


Like caching, cloud storage that is optimized for cloud origin protects your origin from repeated requests, as well as improving availability and performance. Whether you are distributing large video files or releasing new software, cloud storage that is optimized for cloud origin is a key to better performance.


In Conclusion


A CDN provides access to global storage, delivery infrastructure, and optimization techniques that are otherwise inaccessible to most organizations. Moreover, a CDN overcomes a major delivery challenge: the inability to access or control the vast network that transports your content between you and your end users. Static object caching is a good solution to small and large object delivery challenges, including rich media; cloud storage that is optimized for cloud origin can further improve your user’s experience on a cache miss.


But there is more! Chapter 5 of Optimizing the Digital Experience goes into depth about how CDNs optimize your performance. And in the next blog post in this series, you will learn about how to optimize dynamic (non-cacheable) content, what front-end acceleration can do for user engagement, and more.

Until then, find me on Twitter where the  [#OptimizeDigital] conversation is ongoing or drop a note.


Want to learn more about web performance? Download the book Optimizing the Digital Experience for an in-depth look at the topics presented in this blog series.



Click to download


Twitter_logo_blue Tweet this post: Currently reading: Instant Global Infrastructure? What a CDN Does for You.