Q&A: Interview with Pete Johnson discussing ProfitBricks and the public cloud

Q&A Series: ProfitBricks.  Author: Hoofer.

ProfitBricks is a cloud computing company that provides high-end infrastructure as a service (IaaS) for companies and IT professionals.  They offer fast and stable virtual data centers that are based on high standard virtualization technology, which allows customers to easily scale their individual infrastructure to fit their specific needs at any time.  To find out more about this company, I spoke with Pete Johnson, the senior director of cloud platform evangelism.

Q: With so many players getting involved in public cloud, what are differentiators that make ProfitBricks stand out?

A: First generation public cloud providers did a great job creating the market, but like any initial offering 1.0 is never the endgame.  At ProfitBricks, we consider ourselves to be the price/performance leader in the public cloud space which we accomplish with our Cloud Computing 2.0 architecture. By that we mean a more flexible, easier to use public cloud with superior performance and at a better price…


Q: What do you mean by flexibility and what advantages does that give a ProfitBricks customer?

A: Flexibility means several things to us, but the first has to do with instance sizes.  That first generation of public cloud providers offers instances in tiers.  For example, an AWS m1.medium has 1 CPU and 3.75 of RAM and the next size up is an m1.large with 2 CPUs and 7.5 GB of RAM.  What if your application runs best on 2 CPUs and 2 GB of RAM?  With tiered sizing, a customer is forced to choose between an instance that is either under sized or (over sized) under utilized so you end up with poor performance or overspend.

ProfitBricks offers custom instance sizes, allowing you to select the exact specifications you need.  Our customers can choose from 1 to 62 CPU cores and from 1 to 240 GB of RAM and anything in between.  Not only can you choose whatever size you want at instance launch size, you can live vertically scale an instance by adding CPU cores or RAM without rebooting the instance.  That enables our customers to scale up as demand dictates, instead of forcing a horizontal scale out by adding more instances.

Q: A lot of people think that public cloud has become commoditized, so does performance really matter?

It absolutely matters and can save you money, actually.  While you should always benchmark the exact application you’ll be running on any cloud, generic tests like Unixbench that exercise CPU and RAM can give you an idea of the kind of performance differences that are out there and they are significant.  We run more than twice as fast as AWS, for example.  A 1 CPU core, 3.75 GB RAM instance on our platform scores almost as well on Unixbench as a 4 CPU core, 15 GB RAM instance on Amazon.  Why pay for 4 CPUs when you can get the same performance out of a single? Consistency of performance is also important – with 1st generation clouds the “noisy neighbor” problem where one customer could affect the performance of another customer has been solved by our virtual machines having dedicated CPU cores and dedicated RAM. 

Q: How does ProfitBricks handle SDN?

Beyond the instance sizing flexibility, ProfitBricks offers the ultimate customization in private networking as well.  All our instances have 8 virtual NICs that can be connected to the public Internet or to a private network.  This lets customers create multiple private networks that isolate different portions of their virtual infrastructure however they’d like.

Q: Did ProfitBricks build its product entirely from scratch or was there leverage of existing components from elsewhere?

Our core enabler is InfiniBand, the network technology long popular in the supercomputing community.  We don’t just use it as a fast network between VMs, but have a customized version of KVM as our hypervisor along side InfiniBand to be able to offer that custom instance sizing and the live vertical scaling.

Q: ProfitBricks recently dropped prices by 50% on CPU cores and RAM, how is that possible?

Really it’s a combination of Moore’s Law and our modular design.  Our original infrastructure layout first went live in Germany in 2011 before being in the US in Q3 2012.  Moore’s Law predicts the improvement in hardware over that time that we’re able to take advantage of and chosen to pass those savings onto our customers instead of padding our margins further. 

Plus, since we don’t have temporary storage on our instances (all of them use dual redundant RAID 10 volumes across the InfiniBand network instead) we didn’t have to consider disk space on our new servers when we selected them.  Because we’ve separated hosts from storage instead of unnecessarily binding them together, our design is more modular and we can take advantage of hardware improvement in more specific ways than our competition.

Q: How do customers manage images on ProfitBricks?

We offer stock images for popular open source operating systems like Ubuntu, CentOS, and others as well as Windows.  But customers are free to upload custom images they may have from other environments, and we support a wide variety of file types for that.  We also have a very easy to use snapshot feature that enables our customers to save the state of boot or data volumes as needed.  So if you’ve installed custom software on top of a stock image, for example, you can take a snapshot of that and use it to launch subsequent instances.

Q: Why is ease-of-use so important?

If you look at public cloud spend as a percentage of global IT spend, it’s still at less than 1%.  The early adopters of public cloud are primarily developers, who historically have had a tolerance for difficult to use services because part of the fun for that group is figuring things out.  There’s this whole wave of other people – and it’s a much bigger population than the early adopters – who are less tolerant and need something easier. 
Think of the kind of people who used an Apple II versus the people who started buying the Mac.  The graphical user interface made the Mac more approachable to a wider audience and ProfitBricks has done the same thing with our Data Center Designer, which allows you to visually lay out machines, storage, and networks for drag-and-drop deployments.