Chat with us

Options for Boosting CDN Performance with Origin Storage

Blog Post by Charlie Russell, Senior Product Marketing Manager

August 8, 2018

Businesses are keenly aware that they must deliver great online experiences or suffer consequences like high abandon rates and customer churn. To help accomplish their objectives, many businesses use content delivery networks (CDNs), which can deliver content found in cache within tens of milliseconds. However, when the requested content isn’t in cache, the CDN must retrieve it from origin storage where response times and performance can vary wildly.

The good news is that storage responses within tens of milliseconds are possible with origin storage that is high-performance, distributed, dynamic and well-integrated with the CDN. Here are some of the factors that determine response time and performance when retrieving from origin storage.

Performance Factors Retrieving From Origin Storage

Multiple key factors can significantly affect response time and overall delivery for a CDN retrieving content from origin storage. One is transit path from origin storage to the CDN as well as transit within the CDN. Another is the number of locations where content is stored and how the CDN determines where to access content.

What is the transit from the origin storage to the CDN edge?

  • Long-distance internet connection from storage to the CDN edge. In many cases, the CDN must fetch the requested content from storage over the public internet, sometimes over significant distance. In this scenario, simply getting content from storage into the CDN can add considerable latency.
  • Regional connection. In some cases, the CDN has a PoP (point of presence) in the same region as the storage location, or may even have shared access within an internet exchange facility. In these cases there’s generally good performance in transit from the origin storage to the CDN.
  • Best: Storage and CDN edge server in the same PoP. The fastest transit performance occurs when the storage server and the CDN edge server are in the same PoP. In this case, transit from origin to CDN server is virtually instantaneous.

What is the transit within the CDN (PoP to PoP)?

The second leg of the journey from origin to user occurs within the CDN, from the ingress PoP to the egress PoP. If the user happens to be close to the origin, the ingress and egress PoP can be the same. Otherwise, the longest part of the journey may be within the CDN. For example, suppose a user in Europe requests content that’s stored in the Americas. If the CDN has a PoP on both continents, it’s likely the CDN will ingress the content in the Americas, then haul it to Europe within the CDN for egress. The channel for that trans-Atlantic haul can vary considerably depending on the CDN.

  • Open internet. Believe it or not, the vast majority of CDNs depend on the public internet for PoP-to-PoP transit. Their performance approach relies on serving content from cache. When they must retrieve from origin storage, their content transport is at the mercy of the congestion, rerouting, and buffering issues of the open internet.
  • Best: Private backbone. A CDN with a private backbone connecting its PoPs can provide the fastest PoP-to-PoP transfer, avoiding the congestion and traffic issues of the open internet.

Where is your origin storage located, and can you replicate it?

Replication refers to placing content in different regions or even continents. However it should be noted that replication alone doesn’t improve performance, it only improves availability and durability. Regarding storage location and replication, a few typical scenarios include:

  • Single location, not replicated. This is the lowest-cost option. It carries some risk of availability and durability problems if there is an issue with the single storage location.
  • Replicated to 2 locations. Replication mitigates the risk of a failure in a single storage location.
  • Best: Replicated to 3 locations. Replication in 3 locations has the lowest risk of disruptions due to network or equipment issues in any particular region.

Does it automatically serve from the fastest / closest location?

To provide performance in multiple regions, the system must intelligently retrieve from the fastest location for each user session. Scenarios include:

  • Not applicable, only one storage location. Clearly if there’s only one storage location, every session must be served from that location. Site visitors that are far from the storage location will have higher retrieval latency on cache miss.
  • Fixed retrieval from primary storage location. In most commercial cloud storage systems, replication provides backup protection but doesn’t improve performance. That’s because content is always retrieved from the primary location, even if the backup bucket is closer to the user. From a performance standpoint, there’s no difference vs. a single storage location. Site visitors that are far from the storage location will have higher retrieval latency on cache miss.
  • Best: Automatically retrieves from fastest / closest location. An advanced storage system not only replicates content to multiple regions, it also intelligently serves each session from the fastest location (usually the closest). For example, a site visitor in Europe will be served from Europe while a visitor in the Americas will be served from the Americas.

Summary

Why do businesses contract with CDN’s? To ensure the best performance for every user session. However cache performance is only part of the story. To complete the solution requires origin storage engineered for CDN performance, with multiple available layers of advantage to choose from. It minimizes latency from storage to the CDN edge by collocating storage and CDN edge servers. Instead of risking transit on the congested internet, a better solution transports content from origin to the user edge over a high-speed private backbone. And to dramatically improve performance and availability, it offers the option to replicate content to multiple regions and intelligently serve each session from the fastest storage location (usually the closest). Limelight Origin Storage offers all these capabilities as part of its CDN-optimized storage solution. By closing these performance gaps, response on cache miss can be improved and can even rival response on cache hit, at a cost that rivals ordinary cloud storage.