Here are my thoughts from testing out OnApp CDN so far:
Originally Posted by lockbull
One of my questions about these federated CDNs is how is performance guaranteed and what are the SLAs, if any?
There are certainly no strict SLAs with it. But, they do enforce specific hardware requirements in order for an Edge Server to be accepted into the global federation.
The dashboard mentions they do passive monitoring of the edge server every 10 seconds, and active monitoring every 60 seconds. So, hopefully an edge server failure is detected quickly and automatically taken out of service.
Other notes from testing:
Terminology and documentation has a lot of room for improvement. For example, they use the word "publisher" in the global dashboard to describe something called an "edge resource" in another place.
Setup was a little bit tricky. I followed the documentation, but had to get OnApp support to fix two things that I couldn't figure out from the docs. (#1 - cdn edge node template was not installed and #2 - my custom domain somehow just "didnt work")
The backbone of a CDN is the DNS infrastructure. So, to use the OnApp CDN you have to trust and use their DNS. I've done some testing on their anycast dns network and it works OK. Performance is not as good as DNSMadeEasy but they did take traceroutes from me in order to attempt improvement. However, during testing I noticed that one of their authoritative DNS servers was broken (not answering any dns queries) which was concerning. That was fixed within a few hours of reporting the problem.
Interfaces are still a little bit clunky since they are merging Aflexi stuff in with new stuff they are developing. For example, if you want to subscribe to another CDN node in the federation it takes 3 steps in 2 different systems: 1) subscribe in the global dashboard 2) edit your cdn edge group to include it and 3) edit and re-save your CDN "site" (a.k.a. CDN Resource). the 3rd step was not intuitive and I had to contact support to figure that out.
When you create a new CDN Resource, it takes "15-20 minutes" until it will work, according to support ticket 388327. This is not documented and counter to the "Status: Active" that the control panel shows you.
As a provider, you can't deploy edge servers in multiple geographic locations unless you also have a OnApp cloud deployed there. I wanted to add a edge server in another location, but that would require 2 physical boxes and $140/mo in OnApp fees. They said that is something they want to allow but not in the next few months.
Routing requests from clients to the Edge Servers works pretty well but I need to do more testing. It seems that the request will pretty much always go to an edge server in the same *country* however in the US, my requests do not go to the geographically closest edge server. It would be nice if they had documentation about how their request routing worked.
Interested in hearing what other providers think of it ...