Leaseweb is currently seeing a change in the kind of orders we receive. While there is still a growing demand for cloud hosting, we are also noticing an increased demand for dedicated servers.
The likely reason for this trend is that more and more highly specialized Tech companies (most of them active in the SaaS, MarTech/FinTech or gaming industries), would like to have more control over their infrastructure, cost and performance.
For the purposes of this blog post, I will use the term “cloud” to mean a virtualized environment with additional features. HyperScalers, for example, offer these kind of platforms. When I talk about dedicated servers, it means completely automated delivered servers or physical servers in highly customized setups with NVMe’s, GPU cards or 10GE+ uplinks, for example. In many cases, the total environment of the customers asking for this sort of specific hardware ends up in a hybrid cloud, as where the specific parts of the infrastructure (or the base load) ends up on dedicated server, and the commodity part in the cloud.
I believe that in the near future, commodity hosting will move more and more towards cloud environments to offer fair rates at acceptable performance levels with lots of flexibility in the contracts. Specifically, when there are many peaks in the load and you do not want to spend too much time on the maintenance of the environment, cloud will be the preferred solution. My guess is that this will be the case for at least 90% of the IT load worldwide. However, if you really need to pinch out the last CPU frequency or lowest latency, you might see your cloud bill raising to unacceptable levels, or you might even not be able to find your target performance. And exactly for that reason, we now see a growing demand for dedicated servers, with high specifications or very customized solutions. Although this is just a small amount of the requests we receive, it is starting to become a trend.
Dedicated servers instead of cloud will only be cost efficient if you have a certain critical mass of infrastructure, meaning that for most startups, this is not the preferred way of hosting. However, for infrastructure heavy scale-ups or more mature companies, moving parts of the infrastructure to dedicated boxes might be an interesting option to consider if you have at least two out of six possibilities listed here:
- Your cloud bill is of a certain significance, meaning that you paying GBP 10K per month or more
- Your application is in need of ultimate network performance, meaning that every added millisecond is one too many. Or you really need the lowest latency to almost all parts of the world, or to very specific locations where the current cloud providers are not able to deliver this
- Your application needs very high CPU clock frequencies or very quick and predictable responses for the CPU
- Your application needs tremendous amounts of RAM, to deliver the performance you need. This is valid for example in cases where you have a big Databases running in memory
- Your application needs extreme fast disk access, or you want to move RAM towards NVMe disks
- Your application has a lot of base infrastructure, and does not need to peak (meaning that PAYG services won’t be of really value for you)
If you have at least a yes for two of the above six possibilities, it might be interesting to consider a move from cloud towards dedicated servers or probably even more interesting, to consider a hybrid cloud solution, in which your infrastructure is partly hosted in the cloud, partly on dedicated servers. Of course, this might mean more system administration, but with correct tooling, monitoring and maintenance this should not be of too much hassle. On top of that, I’ve now seen several cases where the cost improvement was between up to 60% of cost decrease, so spending a bit more time om server management might still be a valid argument to consider. Leaseweb looks forward to the future to see how this will further develop!