The rush, or rather stampede, to place all services in the cloud is in full swing. The cloud has become the Promised Land – where services “never fail,” and you can be sure that they will always be available.
Let’s take a closer look. An honest look reveals that if something sounds too good to be true, it usually is.
Everyone is moving to the cloud since it is viewed as “always up”.
But everything can go down. Whether it’s Microsoft 365 or a Google region or AWS’s network. And as cloud platforms become more and more complex as providers add more features and services, foolproof maintenance will get harder and harder. That means fewer and fewer cloud providers’ staff will truly have a handle on how everything works and is interconnected, or how to debug a major outage.
So don’t fall for the “always up” myth. Unless you can live for 5 hours while a critical service is down due to a flood in a data center in Paris. But if being down for 5 hours is unacceptable, it is best to design a backup system into your architecture to head off that big cloud outage just around the bend.
There’s a lot to learn from network design on how to do this best.
For example, how IUCC operates Israel’s NREN. We have rolled out numerous networking technologies over the past 30 years – from frame relay, to ATM, to MAN, to Ethernet links. Our axiom has always been to build our network with redundancy. Every university is connected via two different links. There is always the primary provider and the backup provider – which we insist always be different providers.
Over the years, circuit providers have approached us offering huge discounts if we would just let them provide both links. They promise that the circuits are totally diverse and “fail-proof.”
But our experience has shown that there is always some common infrastructure that can be hit by Murphy’s Law.
So we always order links from different providers and always terminate the circuits at different routers and different buildings inside the campus. Over the years this strategy has proven itself. Time and again.
So be wary and don’t let your institutions put all their cloud eggs in one basket. Encourage them to err on the side of caution and opt for redundancy and strong backup infrastructures whenever and wherever they can. For sure you can go ahead and help them put “their heads in the cloud.” But show them how to do it carefully, cautiously and intelligently like only an academic NREN can.