Cloud this, cloud that – early thoughts

I’ve been meaning to put together a post on the whole cloud hosting industry for a few days now, but things are really picking up at work and have thus kept me busy. However, I really wanted to write this and see how people would reflect and perhaps also share their own views on this.

First of all, one of the first things I learned is that all sorts of companies who have something online (to exaggerate a little) call themselves a cloud business. Wikipedia for example, has all sorts of ?aaS acronyms listed for cloud computing. There isn’e even a clear definition on what cloud computing really is. Mostly it circles around the fact instead of investing upfront for resources as fixed costs in accounting terms, you are paying for the actual benefit delivered as a service in variable costs. Usually, these services are paid for on a pay-as-you-go, on-demand basis. However, most of the software-as-a-service products out there charge users on a monthly basis (which isn’t really on demand, as opposed to the common hourly based or shorter billing cycles).

Trying to attach cloud to many things that really isn’t cloud has created a definition for all this: cloudwashing.

Cloud washing (also spelled cloudwashing) is the purposeful and sometimes deceptive attempt by a vendor to rebrand an old product or service by associating the buzzword “cloud” with it.

Sometimes it is tempting for providers with longer histories to slightly change their offering and call it cloud hosting. Amazon Web Services (AWS) and Rackspace are the largest players in the infrastructure game, but I’m willing to argue that they aren’t taking cloud computing as far as they could. This is because, despite claiming flexibility, both companies offer set packages of server resources. AWS has 4 standard instances and some other instances built for compute intensive tasks and memory intensive tasks.

However, these instances come with a set amount of CPU performance, memory and storage capacity. In essence you’re still buying (in a more flexible way) a set of VPS products that you’re able to shuffle around. AWS’ offering of cloud computing on the infrastructure level (especially their EC2 offering) extends to only the pricing of their instances where clients are able to pay in an pay-as-you-go style. It’s not on-demand, because you are given a certain set of resources whatever your more specific demand might be.

Again, I can only talk for UpCloud, but we’ve put in a lot of effort (and to be honest, it’s the team before I joined the company) in building a service where customers are almost freely able to choose their CPU, memory and storage capacity based on their actual demand.

Overall though, there are tremendous changes going on in how services are built and offered in the cloud. Almost always this change is better for the client making the using of these services more flexible. In the points below, I will be talking more about my findings and thoughts on the offering on the infrastructure level.

Secondly, towards the bottom of the infrastructure stack – something where UpCloud is working, the offering can be very hard to understand. One of the key terms in cloud hosting is virtualisation, which means that instead of bound by physical limits – the client is built a virtualised set of server resources.

In this process, service providers are usually able to build their margins for their business. This is because when the clients are given resources, they might not be dedicated as in the case of buying an actual, physical server. In plain English this means that clients are sharing the same resources allocated to multiple clients. When you’re able to charge more money for the same amount of absolute resources, be it per month or per hour, your margins go up. On the losing side are the customers who pay for services that might not perform as well as expected.

Thirdly, when you offer clients a service related to server resources – you want to do it in such a way that they can run enterprise grade applications and solutions on your offering. This also means that you take redundancy seriously. Redundancy is essence are the measures you take to keep your services up and running if individual resources begin to break down. At UpCloud for example, our infrastructure is basically redundant from the network connections all the way to the software level.

Building a redundant stack isn’t cheap, but it is more effective in those cases when things break down. In environments with multiple software level solutions and physical devices, almost always something will break down at some point in time. This means that you are either willing to pay for the redundancy in slightly higher prices or then go for the lowest of the prices which might not pay that much attention to redundancy.

SLA, or the Service Level Agreement, is usually considered legal jargon for the end client. However, when you buy your infrastructure resources as a service – the terms on which the capacity is delivered, is very much part of the package. I’ve even seen SLAs that sum up as “best attempt”, meaning they give no guarantee on how much their services will be up and running. In a case like that, no amount of resources can justify a service that can’t be used.

Marketing and selling these services is of course something I’ve put a lot of focus on. In some senses the cloud infrastructure industry (if we can call it that) is still selling its offering with specifications. In many ways it is at the same point in its lifecycle where the mobile phone industry was in the early 2000s. Touch screens are still a long way off. However, services offering simplicity and ease-of-use are arising. One of the most interesting ones is perhaps DigitalOcean from the US. They’re a startup that joined TechStars last summer and they’ve put immense effort on keeping the service simple.

Also, when everyone is shouting with the same arguments of how much RAM, storage or CPU power you’re offering – it becomes very dull for the end customer buying the offering. Apple obviously changed this in the mobile phone and personal computer space with design. They didn’t completely let go of specifications in their marketing, but they are not the key selling points. Performance is more important than the resources you have.

Marketing has strongly affected the terminology of the industry as well, making it even harder for the client to understand what they are buying. Take for example the CPU resource – central processing unit. The core of the system. Many providers have their own terminology for this or then simply coin it as vCPU (virtual CPU) meaning it’s not a dedicated resource. Some providers even ask you to determine how much priority you want in terms of percentage of that CPU, affecting the pricing of that resource. This all makes it extremely hard for the client to compare what is the best offering for their requirements.

This is where comparison sites such as Serverbear come in. They run a set of tests that try to put different types of offering on the same line. Serverbear has come up with their own BearScore which takes the resources and their price into account and tries to simplify the best bang for the buck. I’m certain these kinds of services will become an essential way for buyers to evaluate the offering available.

In many ways, at least the cloud infrastructure offering is still very early in its infancy despite the market being multiple billions in size. Some studies have shown that Amazon Web Services’ revenue in 2012 was around $2 billion. This makes it the obvious gorilla in the room. For me, it makes the game ever more interesting.

There is money to be made for sure. Linode for example – which is extremely popular among developers, has grown from $1.9 million in revenue in 2009 to $22.3 million in 2011. They generated their 2011 revenue with just 20 employees.

Many have doubted our attempt in building a cloud infrastructure (IaaS) company that can honestly and profitably compete head on with the likes of AWS. After a little more than a month at the job, I can say this couldn’t be further from the truth. We’re extremely well positioned at UpCloud to compete in the game for the long term. The technology is super well built, the team is the best in Finland and we’re only beginning our international sales in the upcoming weeks.

It’s a super interesting industry and I’m very much excited to be part of such a talented group of individuals that UpCloud is made of. Keep your eyes open for our announcements in the coming weeks!

Posted on Thursday, February 14th, 2013.
  • Vesa Sironen


    thanks for a great read – very good points.

    Cloud washing is something that one bumps into pretty much every day; it’s an interesting thought exercise to try to piece together how any company with a single server farm would be able to provide cloud services :)

    Most companies display a lot of fluffy white images on their websites, but under the surface lies nothing but a regular hosting/virtualisation gig.

    Even some of the bigger players market a “cloud based” infrastructure for their clients’ services that ultimately – due to different reasons – ends up being a dedicated environment run from a single container filled with servers.

    Thanks for taking this up. Happy to see that you and the guys are proponents of the real cloud!

  • Stuart

    Thanks for the mention Antti, any ways you think we can improve we’re all ears ^_^

    • Antti

      Thanks Stuart, much appreciated! I love your concept and approach on this industry. We’re also getting geared to run our first tests on your site soon with UpCloud. Will actually e-mail you on some promotional items :)