What’s keeping us from embracing the cloud for our data storage? From my experience the primary hurdle has been cost. Because the cloud is so flexible on how you can store things, calculating that cost is different from one implementation to the next. One thing that seems to be pretty consistent, regardless of implementation, is the cost of bandwidth to upload data.
Most cloud services charge for upload, storage, processing, and download of data. Upload is next to free. Storage is the next highest cost. In the few scenarios I have priced out, it is cheaper to store data in the cloud than to host it yourself on site. So, from a perspective of total cost of ownership, if all you are doing is uploading and storing data, the cloud implementation is a REALLY good way to go. I had one instance where we had to retain nearly a terabyte of historical data, for the few times it was reviewed, less than 20 megabytes. We moved this data to the cloud and reduced our storage costs from 5k per month to $1.40 with Azure Storage.
The next two fees are where your costs can escalate. You pay for performance. If you have a lot of data, it takes more processing power to meet your requirements. The more power you need, and the more you need to always have that power, and not share the potential of the server with other consumers, the more that resource costs. This is a reasonable pricing plan. Still you have to take into account how much performance you require to be there when you need it. Guaranteed power has a premium cost associated with it.
The last item that can result in cost is the amount of data you take out of the cloud. So, if your database is in the cloud, but your application servers are outside of the cloud, it can be rather expensive for output bandwidth. Doing tasks like moving an azure database backup from one location to another for failover of site redundancy can cost a bit as you are extracting data from one site. It is next to fee to upload it to another site.
One way to keep costs down it to host your application and processing in the cloud along with your data. In this way your clients have little or no cost to access your data store. In fact, access speeds are usually phenomenal. Depending on your application, if it is a web site, you may have simply kicked the can down the road. If you are now sending the same data out, embedded in HTML, you didn’t reduce your cost much. You may have actually increased your download bandwidth.
Think of it from another perspective. If you are doing something like machine learning, your costs have moved from download speeds to that of resources to compute your data mining programs. Perhaps the final results you download over the web are not that great. The cost you incurred was performing your programs.
Time will continue to reveal best solutions. I don’t know that these costs are out of line. I would guess that there are some losses in the fees being charged today. One thing is for sure. The cloud is here to stay, technology is continuing to expand its power, and we should always ask the question if an app belongs in the cloud.
Cheers,
Ben