VMIX was eager to explore cloud infrastructure services when they were first introduced some years back. We wanted to take advantage of the cloud’s unlimited scalability and lower total cost of ownership, yet minimize negatives associated with the cloud like system outages and possible performance risks.

Were we ready? Yes, we’d been using cloud-like services for years with our CDN Akamai, who’d been around for over a decade and provides an extremely high level of reliability and service. We also saw new functionality in storage and computing power as logical next-level services for VMIX.

On the other hand, most of the cloud services available were new and unproven. And we couldn’t take any risks using a vendor without a solid track record.

Approaching the Cloud

At the time, we were using proprietary, redundant storage hardware to store all of our client data files. Our hardware was reliable and responsive. We’d had no issues with our equipment since we acquired them and no downtime associated with any of our servers. And so we wanted to see if it was possible to leverage this investment.

Our hope was to build a hybrid solution that made prudent use of the cloud but one where primary storage, processing and data management was still located at our datacenter.

Storage made a good first candidate for transition to the cloud because, one, we had high costs associated with data storage and, two, storage patterns are notoriously hard to predict.

Prior to moving to the cloud we had three options for expanding:

1. Purchase too little storage and try to throttle users’ needs. This would be neither optimal nor desirable for any of our clients.
2. Be VERY good at predicting storage needs and adding new hardware as needed. Since I don’t have a crystal ball, and because we’re constantly adding new clients, this would be very hard to do.
3. Overbuy, and hope that we could fill up new storage devices quickly enough for them to start paying for themselves, while at the same time hoping they didn’t fill up too fast or we’d need to make urgent, and potentially very costly, hardware purchases.

Option three was undesirable but really the only viable choice for us at VMIX—before the cloud.

Stepping Into the Cloud

Venturing in, our plan was to combine the high-availability, concentrated storage in our datacenter with distributed cloud services from Amazon S3. New content and frequently accessed content would always be stored on servers in our datacenter. Then, as content aged and was called for less frequently, it would be moved out to the cloud. Essentially, storage in our datacenter would become one large, highly reliable cache.

Using this model, only infrequently played content would be stored in the cloud, and in the unlikely case of a cloud-related performance issue, new, fresh content would always be available.

From Within the Cloud

These are the very early days of cloud computing for VMIX, and we feel it’s premature to put out a definitive assessment of our deployment to Amazon S3.

What we can say is that migration was quick and smooth and that our clients have experienced no downtime at all. Our hybrid solution employing both datacenter and cloud services is serving us well, and has the flexibility and scalability to allow us to continue to grow.

In short, cloud services can provide a very efficient extension to existing datacenters with a little thought, planning and testing.

As the cloud matures, we trust that the reliability and availability of these services will improve to the point where we need to rely less and less on traditional datacenter hardware.

Over the coming months, our Engineering team will be posting here on specific challenges and solutions associated with our implementation, and we look forward to engaging in a dialogue around any specific issues or questions readers might have about our experiences, or their own experiences, with cloud computing services.