Hundreds of Columns Sorted by Topic

Home

Scaling Content Delivery for IPTV and WebTV by Gary Orenstein, Vice President, Marketing       8/6/2007 New challenges are underway in serving and storing digital content for on-demand viewing via IPTV or WebTV. As the amount of digital content accumulates, storing the data and making it instantly accessible to a large number of viewers are proving to be significant issues. That's because today's audiences want the best of both worlds -- a huge library of titles to choose from, on demand. And while content-centric companies and media outlets strive to meet those needs, it is easy to serve either a small amount of data quickly, or make a large amount of data available with slower access times, not both. But now solutions based on centralized storage caching give content businesses the best of all worlds: instant access to the most popular titles, a vast library archive to fit broad preferences, and the ability to effectively serve content from company-owned data centers. Background In order for content-centric businesses to achieve and sustain profitability, they must maximize the exchange between their large libraries and the highest number of users possible, as shown in Figure 1. But scaling this type of compute and storage infrastructure presents its own challenges. On the storage side, capacity and performance of the content library, such as a conventional disk array, are at odds with each other. The larger the library, the harder it is to serve content quickly due to disk contention and the need to manage multiple client connections. This limited capability hampers the ability to reach a large audience. As a result, content-centric companies typically outsource part of the distribution effort to content delivery networks (or CDNs) that replicate content across geographic areas to guarantee instant access to users. However, this model also has drawbacks with popular content causing CDN costs to skyrocket. Architectural Models for Content Delivery The base model for content delivery involves two main components: a user/Internet component residing outside the data center and the origin or "owned" infrastructure component residing inside the data center. The origin infrastructure typically comprises three tiers starting with load balancers and web servers that reside at the data center egress. That tier is supported by a set of application servers, which in turn access a central storage or database repository. With moderate scale, this model works very well in that the origin infrastructure is relatively simple to maintain, and there is only some reliance on a CDN which does not require excessive content replication or cost. To date, two models have been used to scale content delivery beyond the base implementation: the CDN model and the over-provisioned model. A third, new model is emerging based on centralized storage caching. These are outlined below and shown in more detail in Figure 2. CDN Model When demand spikes and the origin infrastructure can no longer handle response time requirements, content-owners often turn to outsourced content delivery networks. These providers offer the ability to host and/or replicate content to a distributed network serving users from multiple points-of-presence. This model is very effective at meeting user demand quickly and can be implemented with relative ease. However, large scale content distribution via a CDN, particularly with heavily-requested videos or music, can become costly. And while there will always be a need for edge-focused distribution, it makes sense to build origin infrastructures that can absorb additional peak load without sending a corresponding peak to the bill! Over-provisioning Model Fortifying an origin infrastructure in the data center to scale can also be costly, requiring added expenditures and excessive manual steps. Meeting high volume demand typically involves replicating content across individual application or web servers. For very small data sets this is possible, but it is not an option with larger libraries. Maintaining multiple content copies requires a tedious update processes and adds significantly to the capital expenditure at the application and web server layers. Another tactic to scale data center performance involves over-provisioning at the storage layer, replicating the content for performance needs. This too can add significantly to capital expenditures and software costs to maintain multiple, consistent copies. Centralized Storage Caching Model A centralized storage caching approach to scaling content delivery reduces the high variable cost of a content delivery network and eliminates data center over-provisioning. By caching content in a high-speed, high-capacity centralized pool of memory resources, lightweight application and web servers have instant access to a single content repository, with headroom to scale. Details of Centralized Storage Caching for Web Scale Distribution This approach uses scalable caching appliances with access to a single storage repository, and relies on the caching appliance to deliver the speed, performance, and and concurrent access necessary for the application and web servers. As shown in Figure 3, the underlying storage/database component holds a single instance of the original content. As application servers make requests, they are served from the scalable caching appliance. Caching by its very nature is dynamic. As requests are made, they are served from cache. If a request is made that does not reside in cache (known as a cache miss), the caching appliance simply retrieves that content object from the persistent storage device, such as a NAS server. Future requests for that previously requested object are now served from cache to deliver much greater efficiency and response time. In the event that the cache becomes full, the least recently used content object will be removed from cache, but still reside in the persistent storage device. This dynamic cycle requires little to no management overhead. By retaining frequently accessed data in a centralized pool of high-speed, high-capacity memory, the caching appliance can serve thousands of application servers simultaneously without sacrificing performance. This enables the use of lightweight application servers that have less internal memory and disk than what would be required without the caching appliance. As the user base grows and demand increases, the scaling requirements only have to deal with maintaining appropriate caching resources and adding lightweight and cost-effective servers. Conclusions With the rise of Internet-based distribution for music, videos, and other types of content, the ability to scale both the content store and the user base helps ensure profitability and success. To date, distribution models relying too heavily on content delivery networks and those relying on over-provisioned infrastructure have proved costly. Now options combining centralized storage caching deliver the performance and economics for scalable, web-based content distribution that is easily managed, maintained, and dynamically adjusts to meet user demand. About the Author Gary Orenstein is Vice President of Marketing at Gear6.  Gary Orenstein has been active in the IP storage networking industry since its inception with a career spanning multiple network storage companies and industry efforts. He was an initial governing board member of the Storage Networking Industry Association IP Storage Forum where he helped develop, promote, and deliver educational information furthering market growth. Prior to Gear6, Gary served as Vice President of Marketing at Compellent Technologies, a network storage company delivering affordable, modular SANs. Before Compellent, he was a founding team member at Nishan Systems, a leader in the IP Storage market, and spearheaded many of Nishan's milestone industry events such as The Promontory Project -- the first transcontinental IP storage network, and the first wire-speed iSCSI demonstration. Prior to that, he spent several years building international distribution and joint ventures for US companies, including a distributorship for Sun Microsystems in Asia. Gary is also the author of IP Storage Networking: Straight to the Core -- a book that outlines the business value of enterprise storage technology. He holds an MBA from the Wharton School at the University of Pennsylvania, and a BA from Dartmouth College. About Gear6 Gear6 delivers the industry's first centralized storage caching solutions that close the widening performance gap between increasingly powerful servers and lagging mechanical storage devices. By transparently delivering scalable, real time data access, Gear6 eliminates productivity-killing I/O bottlenecks and delivers guaranteed transaction response times optimized for time-sensitive Fortune 1000 enterprises. The company's CACHEfx appliances accelerate business critical applications, protect against peak load disruptions and maximize quality of service levels. Send us your response to this article. Learn How to Get Your Column Published on this Site