Adobe's Emerging Rich Media Ecosystem, Part 3: Marketing, Service Level Agreements, and Security, Page 2
Escalation clauses are undervalued and should be more widely used. These provide for a problem to be escalated up the various tiers of management on both sides if it cannot initially be resolved.
Even the best SLA does not last forever and there must be a procedure for orderly termination and (if necessary) migration from the supplier's system to another system.
Migration is critically important in relation to facilities management contracts and, as a rule of thumb, a year is generally allowed for this. The supplier also should be required to provide all reasonable assistance to the client with the migration to another system.
SLAs Need Technical Expertise As Well As Business Savvy
The complexity of correctly mapping the requirements of a known media service workload into the corresponding system resource requirements and accurately sizing a media server cluster to handle the workload with specified performance requirements makes difficult the task of creating a good SLA.
Note: Because workload measurements of existing media services indicate that client demands are highly variable ("peak-to-mean" ratios may be an order of magnitude or more), it is not economical to overprovision the system using "peak" demand.
The SLA might specify the desirable system performance by stating two types of requirements:
i.) "basic capacity requirements" that define the percentage of time the configuration is capable of processing the applied load without performance degradation while satisfying bounds on system utilization; and ii) "performability requirements" that define the acceptable degradation of service performance during the remaining non-compliant time and in case of node failures during the service time.
To identify the number of nodes necessary to support the required service performance, the provider must consider bit rates, as well as amount of CPU, memory, and disk resources needed to handle a given workload. If you assume that a site's media content is encoded at a constant bit rate for a given workload, it is relatively easy to determine what network bandwidth is required. However, bit rate is usually variable.
As a result, determining the media service workload is a complicated, statistics-based task. The service provider must collect and then analyze media server access logs reflecting processed client requests and client activities at the site.
Sometimes, a high percentage of requests access a small subset of media files. Moreover, many clients do not finish the playback of a full video/audio clip. So, for example, you could discover that as much as 50%-60% of the accesses last less than 2 minutes. Therefore, many accesses to popular media objects can be served from server memory, even when the media server relies on traditional file system and memory support and does not have additional application level caching.
Scaling Flash Media Server 3
Servers have a finite capacity, so as traffic and throughput increases, applications need to be scaled to preserve quality of service. Flash Media Server offers several flexible options for the graceful scaling of high-traffic applications. Knowing that you can increase the capacity of your system when needed is an important detail when writing an SLA.
You can deploy multiple servers behind a load balancer to distribute the application load evenly. Flash Media Server clustering enables you to scale an application to accommodate more clients reliably, and creates redundancy, which eliminates single points of failure. This approach is generally best for live or VOD streaming, where clients do not need to communicate with each other from within specific application instances.
Clustering can be achieved by using either Flash Media Streaming Server or Flash Media Interactive Server. Both provide an enterprise-ready architecture designed to simplify load balancing, failover, and clustering to ensure maximum availability over large regions—the Edge/Origin architecture shown in Figure 3.
Flash Media Server Intelligent Balancing
With the Flash Media Interactive Server, you can intelligently direct traffic to a multiple server cluster using server-side scripting. This option would typically be used for multi-way communication applications that require connections be routed to a specific server. However, this option does require the development of rather sophisticated server-side ActionScript to manage connections.
Flash Media Interactive Server software can serve as either an Origin or Edge server to enable virtually unlimited scalability.
Figure 1: Edge/Origin architecture
Edge/Origin server configurations improve performance by distributing the server load among many computers on a network. With an Edge/Origin deployment strategy, all connection requests from clients are redirected to an Edge server. The configuration also lets you maximize your network if you are supporting a large local network. By placing Edge servers in remote office locations, the Edge servers will cache media files locally so each stream does not need to access the Origin (host) server for each stream.
Page 2 of 6