X

The immediate popularity of Brick forced Alfred Xing to plan a migration from the shared-hosting on which Brick originated to virtual private servers. Brick handles more than 700,000 requests each day (~37GB of traffic). In its first year, Brick has served more than 200 million requests (almost 10TB) for its exacting, WOFF-compressed copies of the original webfonts, converted without modification to various formats for a better typographical experience. Demand continues to grow. Brick serves webfonts with full metadata and without subsetting, resulting in large files (ranging from 20KB to 700KB). Consequently, Xing sought a cloud host that could satisfy three primary infrastructure criteria: high throughput, sufficient CPU power to quickly generate CSS @font-face rules on-the-fly, and a reliable network. Additionally, Brick preferred a cloud service provider that was reputable, well-known for responsive customer support, and affordable.

As an online radio service, AccuRadio faced unique infrastructure challenges due to fluctuating usage rates and the need for flexibility and constancy. Daily usage spikes during lunch and coffee-break times, and listenership varies significantly during events like March Madness and the holidays. Any obstacle preventing users from accessing their favorite channels could lead to dissatisfied users, reduced usage, and revenue loss. AccuRadio's previous infrastructure was non-customizable, inhibiting the ability to measure and interact with its audience effectively. The company needed a flexible and affordable infrastructure to launch new ideas, scale up when necessary, and manage its 1,000-plus channels efficiently. Additionally, as a small company with a dedicated but small tech team, the cloud host had to be reliable and not slow down operations.

The Onion began its online presence last century, using on-site, dedicated servers. However, having encountered system reliability, performance monitoring and application tracking challenges with this hardware, the company shifted to the cloud - and more specifically, Linode. Today, The Onion’s virtual IT infrastructure, which includes 51 Linodes, functions on internal networking between those Linodes. The vast majority of the site’s 11 million distinct monthly visitors connect through this internal networking, Linode’s facilitation of which is critical to optimizing The Onion’s performance. That performance is rooted in a content delivery network, whereby The Onion stores an abundant amount of text, images and other content on origin servers. A dual layer caching system, comprised of individual Varnish servers integrates with the CDN provider, CDNetworks, and serves content for requests from four public-facing servers. These four servers include two application servers and two origin servers that host images. Any user request for content residing on an origin server gets channeled through the Varnish server, which caches the request and also acts as a failsafe in case the origin experiences downtime. Once the Varnish server pulls the requested content in from the origin servers, the CDN and Varnish servers sustain it for 10 minutes before flushing it and refreshing the content. Consequently, the system will deliver content more quickly to subsequent requests than to initial requests. With millions of monthly visitors to The Onion websites, Sinchok’s staff must be diligent in its system monitoring and maintenance. Traffic loads are balanced by using multiple Linodes, connected via private, internal networking. Systems monitoring is done with software such as Graphite and Nagios that runs on individual SSD-powered Linodes. To ensure redundancy and data resiliency, in the event of a failure of services, The Onion frequently backs up its systems using an exclusive, proprietary, backup process. One backup is stored in a server in their Chicago office, and the other is backed up directly to AWS S3.

ShowMyPC’s immediate connectivity of real-time communication relies on lightning quick server response for extremely complex algorithms. The application’s proper performance requires computational speed, network integrity and remote accessibility, three essential criteria for any infrastructure provider hoping to secure ShowMyPC as a customer. Like most rapidly growing startups, ShowMyPC possessed limited capital and staff early on. As a result, the team found the resource-intensive process of server provisioning notably taxing. The exercise of provisioning with many other cloud-hosts required a great deal of time, with complex billing statements and cost-prohibitive fees that were difficult for the company to absorb on a finite budget. As demand for ShowMyPC grew, the company began searching for an infrastructure provider that was fast and made server provisioning easier with the ability to automate common tasks. ShowMyPC needed the ability to provision servers in an efficient and consistent manner. Unwilling to accept deployment delays, the team grew frustrated with providers that either required a significant time commitment to bring new servers online or had an overly complex system or was expensive. Another challenge faced by ShowMyPC was the requirement for instantaneous data transmission between at least two remote sites. As a result, the team found it difficult to find an infrastructure provider that could maintain reliable communication between two or more devices such as PCs, laptops, smart phones or tablets. And finally the company needed a ready and responsive customer support team. Frustrated by slow-responding hosting providers, where the seemingly straightforward task of rebooting a server required the opening of a new ticket, ShowMyPC insisted on responsive, caring, expert and alert customer support.

Digital Firefly Marketing (DFM) faced significant challenges with client-website obsolescence and inadequate cloud hosting services. Over 50% of DFM's clients initially presented outdated websites that were unresponsive to search engine algorithms. DFM had to redesign these websites to make them more search engine friendly. Additionally, many clients were using low-cost cloud hosting services that resulted in slow website performance and frequent crashes. DFM itself initially used a $10-per-month cloud hosting service, which was insufficient as traffic grew. The company needed a more reliable and scalable cloud host with global presence and the ability to integrate cloud-based solutions.

As an online retailer, focused on offering a broad, competitively priced selection of street-wear and skate brands, MLTD.com first used an open-source shopping-cart software, x-cart, on a single, dedicated server. With annual site-traffic growth exceeding 50 percent, a single, dedicated server with its vulnerable, single-point of failure would not sufficiently allow the performance and dependability MLTD needed and its typical 18 – 32 year old male clientele demanded. With sales consistently exceeding previous levels, processing power and uptime needed maximizing to ensure MLTD’s renowned customer service would be maintained. After bringing online a dedicated server for its database and cloud-based web servers, MLTD recruited the Linode Managed team to closely monitor and, thus, secure its dedicated server. Around 2009, only two part-time employees operated MLTD’s IT department. At the same time, the company had reached a sales volume where a single hour of downtime would cost thousands of dollars in lost sales and potentially upset hundreds of customers. MLTD decided it would be a greater benefit to enlist the Linode Managed team to monitor its server, provide customer support and afford immediate crisis response, should it ever be needed. This additional security and peace of mind merited the additional cost. Since enlisting Linode Managed, the rare downtime encountered by MLTD has always been addressed immediately and rectified as soon as possible.

As a 24/7-365 online, global gaming diversion, Munzee has stored every capture since its 2011 launch in its inexpertly configured database. Seat-of-the-pants scaling built a functional but far from optimal platform, which hosted apps like Memcache, MongoDB, and IronMQ. Munzee has evolved from a single server handling growing traffic to scores of servers processing tens-of-thousands requests. As this online traffic grew, Foster and his crew identified a critical need for professional experience to help them determine adequate server allocation, identify sysadmin tasks contingent upon each server, and balance loads between servers. For its first two years of existence, Munzee operated on RackSpace infrastructure. This experience put into stark relief the three criteria a cloud host must exceed to capably serve Munzee’s round-the-clock, global, tech-savvy market. Foster clarified the first benchmark, “We set the bar very high in that we told Linode that downtime has to be kept to a minimum.” With a frequent global audience of tens of thousands uploading more than 100,000 captures even on a slow day, any extended downtime would demoralize players, stifle fun, jeopardize Munzee’s reputation, and seriously affect its bottom line. Downtime couldn’t be tolerated. The second benchmark, pricing, needed to be simplified and predictable. With RackSpace, monthly bills could surprise Munzee bean counters, exceeding allocations. Linode’s hourly billing, where customers pay only for the resources used each month, piqued Munzee’s interest. Finally, if Munzee were to migrate from RackSpace to Linode, it had to be done during the week, its slowest times, and not during the weekend, when most players are online.

Before virtualization, self-funded start-up companies like PlushForums would have had to build and maintain a monolithic computing architecture in a single location and most likely subscribe to a pricey Amazon Web Services to achieve a public-facing cloud presence. The overall cost could prove prohibitive. The threat, great: No matter how well an application could be segmented into services and components, if that infrastructure “silo” failed, it failed for all customers simultaneously. The cloud seemed to be the answer. Spreading risk across numerous virtual servers located in diverse regions should alleviate catastrophic downtime for a start-up. However, until data centers spanned the globe, cloud customers in Europe, Asia or the Middle East would incur extended latency for having to rely on only North American data centers. Nevertheless, cloud computing’s inception became a fanfare for cost-effective computing performance and assured uptime. But this hasn’t always been the case. Still and a bit ironically, many hosting companies pay dearly for annual subscriptions for cloud access, data integrity and reliable system performance. IT budgets bloat, which can sink many shoe-string budget start-ups. Founder Adrian Flitcroft and the gang at PlushForums wanted to avoid those pitfalls to which digital startups can be disposed. To capitalize on the advantages inherent with the cloud, the PlushForums team established four mission critical objectives, including: Immediate, positive cash flow generation. PlushForums sought a means to scale its business without capitalizing large and bulky infrastructure overhead. As a global forum platform, PlushForums also required geographically diverse points of presence, to ensure minimal latency. Responsive and impeccable customer service with as little manual intervention as possible, to sustain automation of new deployments and onboarding. Customer data integrity. PlushForums can’t emphasize and demonstrate enough its commitment to protect and preserve its customers’ data. Achieving these objectives, PlushForums recently completed a successful beta-test period and is now in full production with clients in North America, Europe and Asia.

Sensr.net captures video 24/7 from thousands of cameras across the globe. Its tech-savvy customers generally deploy several cameras, some setting up as many as 800. From those cameras, Sensr.net handles 70 million uploads every day and accordingly needs to process a great deal of video—nearly half as much as that processed by YouTube. For every two terabytes of incoming data Sensr.net processes, four gigabytes of data is sent out. Handling this digital load can be a challenging task for any server. All of the visual data necessitates abundant bandwidth, especially for incoming feeds. The most recent 30 days of video for every customer is stored in the cloud, with still-shot screen captures that enable users to quickly sort through and review footage. In 2008, as Beguelin and his colleagues began evaluating solutions for their Web-hosting and cloud-computing needs, they made bandwidth capacity their highest priority. Initially, Sensr.net considered ramping up its own on-site servers to manage the hosting themselves. Sensr.net quickly deemed this idea impractical because it would have demanded constant oversight and diligent maintenance and commensurate costs. With the expected growth of the business, Sensr.net needed a cloud host that not only could handle huge amounts of incoming data and traffic, but also could allow Sensr.net to scale server needs on the fly while guaranteeing minimal downtime, providing responsive just-in-time support with affordable, pay-as-you-grow pricing.

Back in 2007 – ’08, Brundin and Magnus Appelquist initiated a parttime business out of Stockholm, Sweden, that provided PaaS to a handful of developer and web-design friends and acquaintances. Demand for their service grew. In 2010, the two men incorporated as Cloudnet Sweden, AB. Since that inception, Cloudnet’s main objective has been to “provide our customers with the most reliable hosting service possible,” Brundin asserts. Web designers stand as Cloudnet’s typical user. In fact, Brundin and Appelquist collaborate with Swedish web-design agencies, which subsequently recommend Cloudnet to their customers. This vertical integration of app development services makes it easy for developers and web designers to confidently create - with little concern about system integrity or resource depletion. With agency backing and Cloudnet servers, a Swedish web designer can focus on the task at hand. Cloudnet’s managed product off-loads admin duties, including backups, system monitor support, OS upgrades and package updates. This comes in handy for developers who are not especially steeped in Linux. Cloudnet’s product is supported by a qualified team, about which Brundin takes pride in its responsiveness. These two complements permit end users to develop their apps in what Brundin describes as a “happy” environment. The inaugural Cloudnet service marshaled the power of physical servers and gave rise to the biggest hurdle Brundin and Appelquist would ever have to leap: server interconnectivity. While Brundin enjoyed working with physical servers and racked data centers, he and his staff found it difficult to address mounting issues within Cloudnet’s own data center. Moreover, physical servers constrained the breadth of its managed service. Shortly after Cloudnet’s incorporation, Brundin surveyed the hosting market and noticed a move to the cloud. Nevertheless, he remained a proponent of physical servers. He questioned whether the purported advantages of being in the cloud would continue to outweigh those of physical servers and data centers. In short, he was not certain that the cloud would prevail. Consequently, Cloudnet bootstrapped a parallel hybrid environment, relying on both physical servers in Stockholm as well as virtual servers in the cloud, many of which were and are Linodes. While the hybrid environment sufficed for Cloudnet’s needs and its customers’ demands, eventually, Cloudnet moved fully to the cloud, averting many interconnectivity and performance headaches.

SME initially faced two significant hurdles with regards to its hosted deployment of the Enterprise File Fabric. First, its SaaS platform demands constant uptime and high throughput, so SME sought the capability to balance traffic loads for its clients. Second, for some of its enterprise customers, SME needed to supply on-demand dedicated IaaS instances that could assume the advantages of its Enterprise File Fabric Appliance without requiring the client’s investing heavily in new technology or advanced system administration training, but which would remain cost effective. More recently, SME has sought to automate its client on-boarding for an IaaS instance.