Where's Wally World Record`

Why choose managed servers above cloud platforms?

Platforms are very much "the thing" at the moment, but there are other ways to develop and host a website. This series of blog posts offers a detailed look at the options.

Photo of Greg Harvey
Thu, 2016-09-08 16:07By greg

There are so many options out there in the hosting market, and the pricing models and benefits can be so confusing, it's hard to decide who to go with. So I decided to write a single blog about the perceived benefits of Platform as a Service (PaaS) hosting, when stacked up against what you might call more “traditional” models. However, as I wrote it became increasingly clear a single post was going to be way too long and complicated, so in the end this is an accidental blog series!

Full disclosure, Code Enigma doesn't do a platform per se. (We manage AWS for some customers, but that's a different proposition.) We provide a range of managed hardware options, and devops expertise to go on top, so we do have fairly strong opinions on the matter. So the goal of the series is to answer one question only: why would you choose to buy managed servers (such as our managed dedicated or managed private cloud machines) instead of something on a PaaS model?

The question is, of course, multi-faceted and complex - hence this being a series and not a single post! - and it always depends somewhat on context, but over these posts there are some key points that will be explored:

  1. Cost

  2. Amount and quality of resource

  3. Resilience

  4. Security

  5. Developer experience

Before we get started, the first thing to understand is the PaaS providers are not at all the same. You can try to make some general comparisons, you can do your best to match them up, but they are categorically not “like for like”. Resource availability is often unclear, virtualisation platforms and approaches are radically different, attempts by one to claim their approach is better than all the others are divisive and, ultimately, inconclusive - in effect “you pays your money, you takes your chance”, as we say in some parts of the UK.

While reading around what is available for the various platforms out there, it struck me so much of it is aimed at developers. But think on this: is what’s “cool” for developers really the benchmark for whether or not a hosting service should be recommended to a customer (internal or otherwise) who is responsible for the overall upkeep of the finished application? I would argue no. The two audiences do not have the same requirements and goals. And the latter is ultimately the real customer.

At the end of the day, the questions that really matter to website owners are not about development workflows and virtualisation technology. They’re not even really technical at all. They are typically:

  • “Does this represent good value for money?”

  • “Is my site going to be quick?

  • “Is my site doing to be reliable?”

  • “Is my site going to be safe?”

There are other considerations, naturally - and developer experience is a big one, because there’s a tangible cost to a system that impedes developers in their day-to-day - but the fundamentals are there for looking at performance, security, reliability and value for money.

As an aside, it’s worth noting, most of the “cool” from a developer perspective is eminently doable with more traditional infrastructure approaches, which will be explored in a later post.

Because we’re still fairly rooted in Drupal, I’ve decided to use three platforms offering Drupal support as a differentiator for comparison purposes, where applicable. There are of course many CMS-agnostic platforms out there, but these three are more relevant to our immediate ecosystem as a business:

  1. Pantheon

  2. Platform.sh

  3. Acquia

So, now we have the introductions over with, let’s dive in! In this post, part one of the series, we’ll start with the money.

TL;DR: as soon as you have two or more websites to host, managed servers from Code Enigma are usually cheaper than a PaaS provider.

You may stop now if you want. But if you want to understand how I come to that conclusion, stick around for the long version:

The following “plans” are all “medium” - they’re not directly comparable, because it’s not easy to calculate the available resource per plan with some providers, but they’re roughly similar - as close as I could practically get, given the lack of information from some PaaS providers regarding actual resource allocation. (Two important notes: clearly pricing is accurate at time of writing, but may be subject to change over time, so you should always check this for yourself and not rely on the table below; and the actual specific resource allocated to the selected plans priced below will be revealed in the next post on available resource.)





Code Enigma

Platform.sh

Pantheon

Acquia

£260

£80

£75*

£145**

* billing in USD, also seems to have 2GB RAM available to Platform/CE/Acquia’s 4GB.
** billing in USD, support appears to be included and you cannot remove it.

On the face of it we look pretty expensive. However, there are a number of points that play in favour of the managed VM approach, if you look more closely.

For starters, there are a few “gotchas” in the way some of the platform providers do their pricing. With us, what you see is what you get - a well-resourced private-cloud VM, with as many users, applications, IP addresses, tickets, etc. as you want. However, with platform providers things aren’t so straightforward.

Just by way of an example I discovered recently while handling our Great Escape from #AberdeenCloud, with Platform.sh you only get one “user license” for your £80. Want to add someone? £9.60 per user. Want to add a team of five? Suddenly it’s not £80 any more, it’s £128. Same goes for storage, to spec a Platform.sh “Medium” plan to (more or less) match our server (though you can’t know about CPU, I couldn’t find mention of that) for a five developer team, would really cost £156.50 (at time of writing). Suddenly that’s a whole lot closer to our pricing than it first appears.

But we’re still a ways off. Until we remember all of these platform services are charging per website. If you were a Drupal development agency with five developers looking to host three customers, reasonably low-traffic websites but with some fairly heavy admin views or Panels implementations perhaps. Things which won’t perform well if resource-starved, so you won’t get away with a “basic” plan. (Note, you might also be a company with three different websites for different products or divisions and an internal team of developers - same applies.)

With a Code Enigma VM you can run all three sites on a single server, no problem. Complete with zero-touch automated deployment, the total cost would be £300 (£240 for a server and £20 for each deployment). Platform.sh, your base cost for their “Medium” plan would be £240 (three times £80) but you have a team of five. Adding them to the three projects will cost you £144 more, so total price is £384. So it’s £84 a month more expensive, you don’t know how much CPU you’ve got and you still only have 5GB of storage per client, whereas you’ve got about 25GB storage per client with Code Enigma. Value for money not looking so great now.

And it’s a similar story for Pantheon, though it takes longer for the same maths to prevail - in their case you’d need five separate websites before Code Enigma is cheaper. And with Acquia the opposite is true, with just three customers Code Enigma is already significantly cheaper than equivalent Acquia services, according to the pricing on their website.

So while on the face of it a Code Enigma 4GB VM seems like the expensive option, it really isn’t if you have any kind of scale at all. You literally only need a modest development team and two sites to host before Code Enigma is cheaper than Acquia and Platform.sh, and a couple more and we’re cheaper than Pantheon too.

In the next post in the series and we look into the resourcing of the platforms mentioned here and how they stack up against one another.