14 High Performance Throughput for Rich Content Delivery Customers consistently choose the Scality RING for: Hardware Freedom: Pure software, runs on ANY x86 standard Linux based servers High Performance: Line-rate performance, scales linearly across mixed workloads 100% Reliability: Reliability with minimal intervention at massive scale It Just Works: Production proven in most demanding environments Content Distribution Networks (CDNs) are driven by Media and Entertainment and a growing affinity for high-resolution videos everywhere. Content Distribution refers to providing digital data to geographically distributed end users. This can be in support of scale-out web applications or for digital media streaming services, including new non-linear content distribution models such as catch-up TV, NPVR/NDVR or VoD. At a high level, CDN consists of two essential storage components: Origin Servers, which can contain petabytes of data and are typically deployed in a small number of data centers, and Edge Servers, which are smaller cache servers, deployed closer to users. Many content providers are building their own Origin Server infrastructure for better TCO, and with the Scality RING we have a storage solution to help them run their own infrastructure. This Performance Spotlight highlights why high throughput performance of large files matter for content distribution. The RING’s performance matches perfectly with content distribution workloads, and can do so over both REST and file system interfaces. Large, fixed file sizes using the RING’s FUSE (File System in User Space) can scale throughput performance from approximately 500 MB/s to 2500 MB/s, demonstrating that a single connector is therefore capable of saturating a 10GbE link, and scales further to saturate the 20GbE available bandwidth. By expanding the connectors and storage nodes, 56GB/s of read aggregate throughput is achievable from a single rack, and the throughput performance can scale linearly with additional racks to support an increasing number of requests in the millions. Modeling read workloads intended for content consumption from origin servers – the hardware agnostic RING shows throughput scalability under increasing load without degradation. Read throughput performance for large file sizes show a high degree of consistency and predictability essential for content delivery. The RING can support a wide array of file sizes at massive scale that enables the content owners and distributors to manage large catalogs of rich media titles encoded in many formats. Content Distribution refers to providing digital data to geographically distributed end users. This can be in support of scale-out web applications or for digital media streaming services, including new non-linear content distribution models such as catch-up NPVR/NDVR or VoD. This benchmark workload is intended to model access from the file based applications to a RING used in a content distribution use case, represented by an Origin Server providing the source data repository for a Content Delivery Network (CDN), or a video server providing media content to content delivery engines. In this workload, we model a read access from the applications serving data to the edge servers in a CDN, using file-based connectors such as FUSE. The file sizes here represent media content, with larger fixed size files beginning at 128M, 512M and ranging up to multi Gigabyte sized 8192MB files. This is intended to represent larger video media assets, as would be common in a video server use case. Much of digital data delivery is the service is provided by CDNs. The high level architecture of a CDN consists of two essential storage components: origin servers, which can contain petabytes of data and are typically deployed in a small number of data centers, and Edge Servers, which are smaller cache servers, deployed closer to users. Many content providers and owners are now starting to build their own Origin Server infrastructures instead of paying storage subscription fees to CDN providers like Akamai and Limelight. By doing so, this empowers these companies to offer new content services, stream content through multiple CDN providers using least cost routing, leverage more content streaming providers, and reduce their overall TCO. Non linear services such as NPVR, time shifting, program restart, catch up using VoD asset from live broadcast, all will increasingly stress the bandwidth to access the files stored via the origin server. Media files that are streamed are compressed for content distribution. The adoption of 4K into the mainstream is still at the early stages; however using the public cloud to distribute 4K content will not be affordable due to high bandwidth costs and massive size of data transfer. Private cloud service providers will have a significant edge in providing 4K content to their users in the near future. Why choose Scality RING for Content Distribution? Build scalable distribution infrastructure Meets performance for HD media streaming High throughput for HD streaming Limitless storage for rich media content Lower TCO than AWS S3 Flexible, geo-aware platform Enables content owners to stream data through multiple content providers The performance was validated at the Scality lab, and there are more details available upon request. Large file sizes ranged from 512KB to 8192MB to model access from file based applications in the media space. Using a single FUSE (file based) connector, throughput performance was capable of saturating a 10GbE link, and scaled to maximize the utilization of 20Gb available bandwidth. Photo by ShareGrid on Unsplash
