What does the AWS Outage say about the Public Cloud?
For those in the tech industry, you would have to have been asleep to miss the recent coverage of public cloud outages. More specifically, the AWS outage that took place last week and affected hundreds of thousands of users around the world, causing downtime for more than four hours.
Amazon Web Services has blamed this on “human error”, saying one of its employees was debugging an issue with the billing system, which subsequently took many more servers offline than it intended. This clearly shows how vulnerable the public cloud can be.
Why is this so damning?
Now, the impact of a website and IT downtime can be simply catastrophic. Revenue loss for some can swim into the deep abyss and the loss of a retained customer base can be evidential. With the IoT and countless smart apps controlling the way customers live, such downtime has more effect than now than it ever has. This specific public cloud outage caused things like security cameras and smart home technology to go down; meaning people could not control things like their thermostats and light bulbs. First world problems, it would seem.
In the wake of this, organisations may now need to look at a different strategy. Those relying heavily on the public cloud will need to consider how to reduce any downtime risks.
Whilst cloud services, when compared 1:1 with classic colocation often appear to provide cost-savings; these comparisons are often made with a single ‘Cloud’ presence. There is implied/assumed resilience within the Cloud, which does not directly compare to traditional dual-site colocation. For a direct comparison, it is essential that one consider having a presence in multiple ‘Clouds’.
Once this is in action, the costs of specialist application re-development, inter-cloud data replication and (obviously) being billed by two different cloud providers, start to mount up.
This is especially true when a large amount of data needs to be replicated between the two Cloud providers. Each one charging for data-in and data-out respectively. A dual-site colocation package (with a high-capacity unlimited usage Ethernet link between them) will often swing the costs significantly towards the traditional colocation solution.
What’s more, if any potential problem with an organisations infrastructure were to arise, the modern data centre will have the skilled engineers on hand 24/7 to alert you and deal with the issue. Much unlike the AWS outage last week, whose breakdown was so bad that they could not even update their own status dashboard to warn the world.