Content Delivery Infrastructure Updates #4


Akamai and Limelight Networks Create a New Chapter in Their Relationship With Licensing Agreement

Global CDN leader Limelight Networks, Inc. has ended a long and costly legal battle of attrition with Akamai Technologies over the disputed ‘703 patent. The two companies have agreed to waive all rights to appeals and convert the $51 million judgment into a $54 million license agreement instead, reaching an amicable accord.

The license has been subdivided into twelve parts that will be paid in quarterly installments over three years beginning on August 1, 2016. As a part of the deal Akamai is releasing its claim on the $51 million letter of credit to Limelight and frees Limelight to operate its network without fear of litigation.

“We are pleased by the outcome of this agreement. It eliminates the continuing risk from the ‘703 patent and allows us to extend the $51 million payment over a three-year period at an attractive interest rate,” said Robert Lento, CEO of Limelight in a press release. “This agreement, coupled with the series of financial and operational improvements, further enhances our confidence in Limelight’s value creation opportunity.”

It is worth noting however, that this settlement does not resolve other ongoing legal disputes between the two companies, particularly with regard to patent infringement claims filed by Limelight against Akamai and XO Communications in 2015. The press release clarifies that “This agreement excludes patents that Akamai asserted as counterclaims in the patent infringement case filed by Limelight against Akamai and XO Communications in the Eastern District of Virginia in late 2015 and does not otherwise impact Limelight’s rights in that case.”

Google Open Sources Implementation of Wide & Deep Learning API

Giving computers the power of deduction would be a giant leap in machine learning, enabling them to observe and memorize facts and extrapolate generalizations from them just as humans do. Google, in a research blog post, broadly categorizes memorization as wide learning and generalization as a deep neural network (or deep learning), jointly dubbed Wide & Deep Learning.

The post notes that such a conceptualization allows for generic large-scale regression analysis and classification problems with sparse inputs. Google has taken the step of open sourcing its Wide & Deep Learning API as a part of its TF.Learn API in order to allow the general public to develop models themselves. Links to tutorials are available within the blog post.

By integrating both wide and deep learning into app development, apps are trained to utilize two different prediction protocols in order to return more accurate results, improving with time and learning from its errors: “During training, the prediction errors are backpropagated to both sides to train the model parameters. The cross-feature transformation in the wide model component can memorize all those sparse, specific rules, while the deep model component can generalize to similar items via embeddings.”

Open Source Compression Technology Brotli is Key Ingredient To Dropbox

Google released its Brotli lossless compression algorithm to public consumption on Github late last year, under an Apache open source license. At the time Brotli was touted as a highly efficient and more advanced compression algorithm, capable of 20-26% higher compression rates than its predecessor.

Smaller compression is useful and important for the simple, yet powerful reasons that it allows for “better space utilization and faster page loads”, wrote Zoltan Szabadka, a Google engineer, on Google’s Open Source blog. “We hope that this format will be supported by major browsers in the near future, as the smaller compressed size would give additional benefits to mobile users, such as lower data transfer fees and reduced battery use.” Compression allows for a faster and more seamless user experience which drives traffic and ad revenues among other benefits.

Dropbox, it appears, couldn’t agree more. In a recent blog post, Dropbox revealed that it had been making open-source improvements to the Brotli codec in order to drive down bandwidth usage and improve user syncing and collaboration on its platform. The team is predicting that such improvements will result in a 4.4% decrease in bandwidth usage per bit for business customers. Another benefit of Brotli, aside from the fact that it is open source, is that it is already compatible with and supported as an encoding format on most popular browsers, including Firefox, Chrome, Opera, and Android Browser.

Akamai Introduces 10for10 Service and Managed Delivery Service

Akamai has released two press brochures detailing two new consulting services. The first, dubbed 10for10, is a service consulting package that promises to improve customer’s web performance metrics by as much as 10%. In the web delivery business, latency can be toxic from a profit standpoint, Akamai argues, citing a statistic that Amazon lost 1% in sales per 100 ms of latency.

Users are demanding when it comes to the performance they expect from websites and are turned off when websites take longer than 2 seconds to load. As such, Akamai is touting 10for10’s ability to decrease Page Load Times, increase SpeedIndex (which measures the time it takes to paint the area of the webpage above the fold), and increase offload hits and page weight (which measures the magnitude of all the embedded resources received by the client). Akamai guarantees a 10% improvement in at least one of the metrics or your money back.

The Managed Delivery Service provides offload website performance management services, allowing business to manage, monitor, and improve performance websites with the aim of improving user experience. Part of the unique value proposition of MDS is that it promises to provide actionable insights and remedial action to problems that have been identified. In order to do so, the MDS team leverages synthetic test data and real user monitoring data to identify problems and conversely opportunities to optimize performance 24/7.

The team essentially babysits an assigned group of URLs, assessing web performance based on metrics such as industry-standard KPI’s and user data. Once an issue is identified, the team performs triage, locating the root issue and prioritizing issues for mitigation. Finally the team offers weekly reports that define optimization opportunities.

Mellanox Introduces Dual-port 25GbE NICs for High Performance Servers

Mellanox is unveiling 25G networking technology that promises to boost network performance and speeds at a time when mobile data demands continue to grow steadily. Increasing throughput and scaling cost-effectively are tough demands that legacy technologies are struggling to meet. Part of the suite of 25G tech that Mellanox is debuting includes dual-port 25 GbE adapters in both mezzanine and stand-up PCIe card form, which allows users to drive down total cost of ownership and improve network performance. A dual-port 25 GbE solution provides 2.5 times more bandwidth than placing 10 GbE on each port, and is optimal for a variety of use cases including high frequency trading, video streaming, and NFV.

In terms of specs, the new HPE 25GbE adapters utilize an SFP28 port form factor in order to expand the number of connectivity options that each port supports suiting a variety of enterprise consumer needs and physical networks. The adapters utilize copper and fiber optic cabling, removes the need for breakout cables, and can be built on existing 10GbE fiber structures.

Scroll to Top