Shockingly Fast Content Delivery When You Need It The Most

Today there is a massive, multiple-threat storm system moving through the DMAs of two of our vendors, and I want to brag a little. It's kind of hard to show how important it is to have a reliable, scalable, and fast CMS when there isn't an emergency. Today is a big emergency in Oklahoma. There are multiple tornadoes, multiple supercells, multiple warned areas, hail and flooding.

Consider what is happening during an event like this. All hands are on deck just trying to generate the content, create videos, images and text. Everybody is trying to support the meteorology team. They need to be able to focus and not worry if the CMS is going to break down under the load. On days like this, having a high quality CMS is crucial, and lapses of service, or slowdowns, can cost lives. People need to be informed of what's coming, and quickly.

So I'm up here in Alaska, watching it all go down, and at this point we've got the product so dialed in that it requires nothing of me, so instead I'm bragging about how well the CMS functions because the metrics we are showing right now are flabbergastingly fast.

Let's look at load, the sheer number of requests coming in for an event like this, versus "origin latency". Origin latency is the time it takes the system to reply to a request that isn't handled by the CDN. CDN response times are on average 15ms or less, and are handled locally on nearby nodes. That's easy. All CMS providers can do that at their "enterprise-scale" pricing tiers. Worpress VIP can do that, Brightspot can do that, BLOX can do that. I'm talking about requests for content that doesn't exist on the CDN nodes. They are in a sense, CDN overflow. Here is the traffic chart vs the origin latency for those types of requests, in 5 min intervals:

You can see on the first chart that it was a quiet night, we were down to 5k requests every 5 min (that's only 1.3 per second), and that's when our latency figures were the highest, but still far better than Wordpress VIP or any other enterprise scale CMS out there. Over the early morning hours, when few people were looking at the websites, we were seeing average latency around 130ms, with some spikes up to 266ms which quickly come back down as new servers are created. There were just 2 servers at that time.

When the storms came in, we jumped up from 46k requests/5min to 150k/5 min. Eight more servers fired up, all in parallel, for this single affiliate. Notice that that drives the origin latency down to 60ms. This is incredible, and it's not done yet. Another traffic spike all the way to 171k request/5min drove it down even further to 57ms, and there was a brief spike that was driven down again as the servers continued to scale. At that point there were 15 servers running in parallel to handle the load.

I'd like you to compare this directly to a screenshot of the sort of metrics that Wordpress VIP offers, directly from their documentation:

There are three things I want you, the reader, to understand here.

1. Viking CMS delivers content faster than anything else out there. Not even joking.
2. Viking CMS delivers content faster when you throw more traffic at it. A lot faster in fact.
3. Viking CMS has your back when you need it the most. When the sky is falling and everybodies pants are on fire, you don't have to worry about slowdowns or failures out of your control.

I'd like to add a final note as well. The content that the users are requesting, on a day like this also includes a great deal of off-site data (not just stories and categories, images and videos). In this case the forecast, METAR, and alert data is all coming from Baron, a private data bundler, for NWS and NOAA sourced-data. Our "Baron Forecast Task Runner" is a separate cluster that is also scaling independently to handle this amount of load, I just checked it, and it is running 7 servers per affiliate just to satisfy the need for real-time forecast data. Check out the amount of traffic it is handling, and the near-zero origin latency it is offering for these requests:

If you need a complete stack to handle extraordinary events, as well as "normal" days, Viking CMS is the best you can get.