Half the price + Twice the speed = Price Performance Leader

AASimply put, ProfitBricks founders Achim Weiss and Andreas Gauger have a history at bringing innovative, disruptive technologies to market.  At 1&1 hosting, with advances in data center efficiencies not seen in that market previously, they not only introduced a new cost model but did so with superior benefits as well.  Their gains were so great, people thought their claims were unfounded and their pricing a desperation move at grabbing market share.  Those people were wrong.

The same thing is happening again in the IaaS market.  Today we have announced our new innovation-fueled pricing that, when combined with our greater flexibility and superior performance, saves our customers 40%-80% compared to Amazon or Rackspace.  It’s an exciting time for all of us here at ProfitBricks and when you take advantage of our 14-day free trial you’ll see the difference yourself.  Half the price at twice the speed makes ProfitBricks the price performance leader.

TripAdvisor: A Real World Price Performance Example

New pricing sounds great in theory, but how about a concrete example?

After TripAdvisor made details of their architecture public, PlanForCloud published an excellent blog article detailing what it would cost on AWS.  This serves as a great real world example that demonstrates ProfitBricks superior value.

We ran our own pricing analysis on the 270 front end servers, 70 back end servers, 32 Memcache servers, and 5 database servers (with additional storage) outlined in the PlanForCloud article, omitting data transfer costs for simplicity.  You can check the math yourself, but here are our findings:

Comparison1

Trip Advisor Architecture, Comparison #1: ProfitBricks vs AWS

What’s the difference between the Aggressive and Conservative ProfitBricks estimates?  ProfitBricks utilizes InfiniBand to provide a more flexible (through customizable VM sizes) and better performing (dedicated CPU cores and RAM) public cloud alternative to 1.0 providers like AWS.  The Aggressive cost estimate takes full advantage of this by matching the capacity more closely to the customer need.  In a world where AWS instances see around 15% utilization (slide 9), this provides a tremendous opportunity for cost savings.

For example, the TripAdvisor Memcache servers were sized by PlanForCloud as m2.xlarge on AWS, which have 2 CPUs, 17 GB of RAM, and 420 GB of ephemeral storage.  In practicality, though, those Memcache servers do not likely need 2 CPUs given their role in the architecture despite the fact that AWS forces customers to purchase that amount of CPU when what they really want is the 17 GB of RAM.

With ProfitBricks the number of cores are independently scalable from the amount of RAM or disk space.  Ultimately that means that customers pay for what they need instead of what they are forced to buy.  Similar arguments can be made about the amount of storage in AWS instance sizing and cost savings can be achieved there as well.

On top of that, independent 3rd party benchmarking from CloudSpectator shows through Unixbench test results that ProfitBricks CPUs run twice as fast as AWS CPUs.  That enables us to halve the Database, Front End, and Back End server CPUs and save even more.

Some people, though, either don’t believe this flexibility is possible because AWS doesn’t offer it or discount its importance despite the amount of savings that can be gained.  Or maybe they don’t believe the Unixbench numbers even though there’s enough detail in the CloudSpectator report for anybody to duplicate the results for themselves.  For those folks, the Conservative estimate offers a more direct comparison that matches the instance sizes more strictly.

It is important to note, though, that the default PlanForCloud calculation does not take into account support.  ProfitBricks offers the equivalent of AWS Business Support free of charge.  When taking this into consideration, the comparison changes to:

Trip Advisor Architecture, Comparison #2: ProfitBricks vs AWS

Trip Advisor Architecture, Comparison #2: ProfitBricks vs AWS

What about Reserved Instance pricing?

The analysis above is pretty compelling, but you may have thought, “Hey, what about Reserved Instance pricing?  Isn’t that significantly lower?”  Yes it is, but free lunch is never free.

Among the key benefits of the public cloud is supposed to be flexibility and matching capacity with demand without the up front costs associated with traditional on-premise or co-location data centers.  Amazon Web Services Senior Vice President Andy Jassy himself made this argument in his keynote at the inaugural AWS Users Conference last December (jump to 25:10 where his slide literally says “You Don’t Need to Guess Capacity”).

A version of the same diagram from Andy Jassy’s November 2012 Sydney Customer Appreciation Day slides (slide 33)

A version of the same diagram from Andy Jassy’s November 2012 Sydney Customer Appreciation Day slides (slide 33)

The problem with the traditional model of procuring compute and storage capacity, so the argument goes, is that it relies on forecasting.  You have to predict how much resource you’ll need and if you are wrong you either end up with wasted capacity or customer dissatisfaction.  The promise of the public cloud is its ability to react more quickly and at finer resolution to changes in actual demand than a traditional data center approach.

But Reserved Instances ask you to predict demand, exactly what Andy Jassy tells you not to do.  This is in direct opposition to the fundamental concept that was supposed to draw you to the public cloud in the first place and even Amazon admits that it is a difficult task.

Market reaction to Reserved Instances supports the idea that this just doesn’t make sense in a cloud world.  In their presentation at Cloud Connect this year (slide 20), Cloudability revealed that only 20% of the accounts they help manage — and they are now manage half a billion dollars worth — make use of Reserved Instances.  In other words, most people don’t use them.

Suppose, though, you actually did have predictable load for a portion of your architecture making Reserved Instances attractive.  Is that necessarily the best solution?  Or are there non-cloud alternatives that might be even cheaper?

The answer varies greatly by specific need, but in many cases renting a portion of a cage at Switch’s SuperNAP and cross connecting to one of the many public cloud providers they host (including us) for cloud bursting needs can be just as cost effective.  Similarly, going with another co-location provider for your predictable load and direct connecting to a public cloud for the variable load, like Zynga has done, is another option.

It is worth pointing out, though, that among the reasons ProfitBricks is in SuperNAP is because of their unsurpassed options and price leadership when it comes to direct connections.  Their volume of network providers, as ProfitBricks has discussed before, inherited from Enron’s build out that was never released gives them unique capabilities that other co-location providers, and therefore public cloud providers that hare hosted elsewhere, simply cannot match.

What about Spot Instances?

Even more cost effective, on the surface, are Spot Instances. AWS sells off excess capacity at extremely low prices but with the caveat that your instance can be turned off when demand throughout the entire user base dictates that your bid price is too low compared to what other customers are willing to pay.  Essentially, you can get a great price on a VM, but you have to be prepared for the fact that it can be turned off at any time.

Again, though, this doesn’t come for free.  In order to take advantage of this, not only do you have to have a workload that is interruptible on a moments notice, but the onus is on you, like so many other AWS limitations, to code around the particulars of Spot Instances to make them usable for you.  Otherwise, you are at the mercy of other users, who impact the reliability of your instance.  While the price is great, that might be the noisiest neighbor problem ever.

Half the price + Twice the speed = Price Performance Leader

Cloud computing, as even Andy Jassy argues, is supposed to be about matching demand with capacity more quickly and with more precision than traditional data center alternatives.  Since our inception, ProfitBricks has taken this to a new level with per minute billing, flexible instance sizes, and live vertical scaling.  By dedicating CPU Cores and RAM to instances without overprovisioning, we complement that greater precision with better and more consistent performance.  Non-coders love our Graphical User Interface, with its easy to use drag-and-drop deployments and the developer community can interact with our API to create automations like our new integration with Chef for programmable operations.

Now, based on a leap in data center efficiencies that Achim and Andreas have proven in the past can change a marketplace, we have the best on demand price in the IaaS industry.  Simply put, half the price at twice the speed makes ProfitBricks the IaaS price performance leader.