There’s a lot of back and forth in the hosting world about the relative benefits of cloud platforms and dedicated servers. Proponents of each tend to use the old marketing trick of lining up the strengths of one against the weaknesses of the other, so it’s hard to get a decent grasp of which performs better.
An interesting case study from Ben Ubois, the developer of RSS feed reader Feedbin, provides a useful data point for those who are thinking about creating a web service but are unsure whether to build in the cloud or use dedicated servers.
Feedbin is one of the many RSS readers that found prominence following the demise of Google Reader. On the face of it, an RSS reader isn’t an especially complex service, but it is one that can require a lot of resources for syncing feeds and for providing a responsive experience to users of the web interface.
Ubois originally deployed the Feedbin front-end on Heroku, a cloud application platform. As the service became more popular, he turned on more servers and increased the database resources until he was paying $6,400 per month just for the PostgreSQL database. Even at that price, the site wasn’t performing well under load.
He decided to move away from Heroku to a hosting provider that offered dedicated servers with SSD drives. The performance difference was remarkable. Ubois said:
“If you’ve never used a dedicated server before it’s a breath of fresh air coming from AWS or any VPS. All those problems of noisy neighbors, poor I/O, weird LAN latency or other networking anomalies just go away. Oh yeah and it’s fast.”
In addition to being much faster, the dedicated server hosting was also much cheaper, less than half the monthly hosting costs in the month the transfer was made.
This is not an uncommon story. The cloud has many benefits, particularly its on-demand pricing and the ability to scale rapidly. There is, however, a significant fly in the ointment. The performance-cost ratio of cloud platforms is terrible when compared to dedicated server clusters.
In rare cases, that’s a payoff worth making. But for most traditional hosting scenarios, on-the-fly scalability is like a chocolate fireguard, nice to have, but not really very useful.
Dedicated server clusters are an inherently scalable. While it’s true that it would be impractical to add 30 servers to a physical cluster in the morning and remove them in the evening as one can with a cloud infrastructure platform, it’s also irrelevant. Almost no one needs to be able to do that with their web hosting. Demand tends to be fairly predictable – site owners know when they need to scale and doing so on a dedicated server cluster is straightforward.
The Feedbin story is illustrative of a larger point: the cloud is great for some, but for web hosting, it’s not a cost effective or efficient solution.
Photo credits: John Gilchrist