2016年5月15日
By 1月时特约编辑
专题文章

The Cost of the Cloud for Video 编码: Crunching the Numbers

You need many different types of data—and lots of it—to make a good decision on cloud versus on-prem. I learned this recently while running a comparison for a consulting client with substantial encoding requirements. Multiple cloud and appliance vendors ran throughput tests to help determine pricing.

有趣的是, throughput doesn’t impact pricing for cloud vendors that charge via the SaaS model; you pay based upon minutes or GB of input/output. But multiple vendors—including Elemental, Telestream, and 编码.提供平台即服务模型(PaaS), through which you rent encoding capacity based on time, 在这段时间内尽可能多地编码. 1 .定价编码.com using both models, and for this client, PaaS was much cheaper.

然而, 在计算PaaS定价时, details such as single-pass versus dual-pass really matter, because dual-pass can increase your encoding time significantly, 价格也相应上涨. While there is no similar increase for SaaS pricing, there is a potential quality concern if dual-pass encoding isn’t offered. Ditto for quality-related adjustments—if set to max quality in the PaaS environment, they can increase encoding time and cost; if unavailable in the SaaS environment, 它们会对质量产生不利影响.

Obviously these same distinctions are critical when calculating how many encoders you need to meet ongoing demand in an on-prem scenario. 当要求供应商为您计算吞吐量时, be sure to specify both quality levels and the number of passes.

Another critical detail is how and where you perform adaptive bitrate packaging (ABR). 基本上, you have two options: You can either package while encoding or deliver MP4 files to your origin server and package dynamically with Wowza Streaming Engine, 基本流, 或者类似的产品. Packaging during encoding increases the price of SaaS encoding, which 再一次。 is based upon throughput in minutes or GB of video. 用于PaaS服务和设备, 影响很小, since ABR packaging is a lightweight operation that takes few CPU cycles, which is why you can (and probably should) perform it dynamically at the server.

一旦有了吞吐量细节, 你可以在云端为编码定价, and the cost of the required hardware to encode on-prem. 现在是时候考虑其他成本了. 最好的参考可能是编码.Com的白皮书题为， “Media Processing: Total Cost of Ownership, On Premises vs. 基于云的编码工作流是由核研究中心准备的吗. The white paper details three scenarios: open-source encoding on general-purpose hardware, enterprise-encoding software on general-purpose hardware, 以及企业编码设备. 第三点与我的分析最为相关.

在这个场景中, 该公司需要31台服务器, 而云计算要便宜200美元左右,4年，税后1000. Dig into the numbers and you’ll see that a major component of this savings was related to reduced staffing costs. 具体地说, the analysis assumed that on-prem encoding required 150 hours per month of IT time for “maintenance and upgrades,” 150 hours per month of IT hours on “capacity planning and development,” with another 150 hours of staff time lost each month ”by end users due to encoding delays.“根据白皮书, 而在云端进行编码则没有占用任何时间, essentially freeing the company to cut or reassign significant chunks of three employees’ duties. Under fully loaded salary assumptions of $80K each for IT employees, 给另一名员工7万美元, 这大约相当于200美元,每年裁员1万英镑, 再一次。, a significant component of the overall savings delivered by the cloud.

从我坐的地方, encoding in the cloud is a natural for spikes in encoding, such as for complete library retranscodes for a new format or device. It’s also a natural for companies that are committed to a cloud IT strategy for the bulk of their IT needs. 如果你对云编码持观望态度, 特别是对于一贯的, 持续的编码需求, it may be tough to financially justify the transition unless you can identify specific staffers that you can cut or reassign when moving to the cloud.

This article appears in the April/May 2016 issue of 流媒体杂志 作为“云的成本”."