The rise of Infrastructure as a Service got the world used to on demand, utility services that correlate perfectly with an organization’s particular demand profile Indeed one of the key value propositions that all cloud infrastructure vendors articulate is that, whereas in the fast infrastructure needed to be purchased in distinct blocks that didn’t relate to gradual demand increases, with IaaS an organization can achieve that beautiful thing, a demand line that correlates exactly to a usage profile.
But just how honest is this sales pitch? Are vendors truly offering customers perfect flexibility in how they buy infrastructure? A recent blog post looked at the issue in-depth and suggested that the reality is somewhat different to the hype. In reflecting upon Amazon Web Service’s offering in particular, CEO of IaaS vendor ProfitBricks, Robert Rizika wrote that:
This is in obvious contrast to past models in which you buy hardware for your on-premise data center or your colocated space, then live within whatever means you own. That capital expense was slow to react to shifts in demand and put you at risk of over- or under-capacity. Hourly billing freed us from that – most users saw a clear progression toward per minute billing to even more closely match the demand curve. Why, then, when you use an Amazon server for 62 minutes do you still get charged for 120 minutes? Multiply that by a couple hundred servers over a year and the difference adds up. How come AWS gives its best prices only with 1-3 year contracts, with up front fees? That sure seems a lot like that step-wise CAPEX spend cloud was supposed to avoid.
Rizika goes on to suggest that it is technology that is causing this problem and that technology makes it difficult to deliver fine grained options to customers. I’m not convinced that it’s purely a technology problem and have spent time in the past few weeks with two companies that are looking to solve this problem – one from a technology angle and one from a business one.
It’s About Technology – ProfitBricks Offers Vertical Scaling
I’ve written before about ProfitBricks, a company whose main selling proposition is that they offer variable instance sizes, along with the ability for customers to scale those instance sizes on the fly. The ProfitBricks view goes that simply scaling out and adding another virtual server unit as demand increases is little different to adding another physical server in a data center – it means that supply increases in chunks and, while the chunks on AWS are smaller and more granular than with physical infrastructure, it’s still sub-optimal.
ProfitBricks has, for the past three years or so, been building the technology to run what they call a virtual data center – essentially users can custom define their instance sizes and scale those instance sizes in real time. In an effort to answer the criticisms around chunky billing, Profitbricks are also offering minute-based pricing, a departure from the compute/hour norm in the industry. The obvious question here is how big can a ProfitBricks virtual server scale, since this upper limit determines, to an extent, how useful the service actually is.
By utilizing their proprietary software, and the Infiniband technology they use, ProfitBricks now offers a maximum size of server of a massive 62 cores with 240GB of RAM – that’s a monstrous (virtual) machine. They’re aiming this sort of server for high performance computational processes and organizations that run huge databases. By scaling up the size of the instances available, not only does ProfitBricks improve speed and performance in comparison to regular IaaS, but it also reduces operational burdens as less individual instances need to be administered – while it’s true that automation significantly eases the administration burden of IaaS control – any savings in this area are positive.
The graph below is a UnixBench test result showing comparative values between ProfitBricks, Amazon and Rackspace – more details on the test methodology here.
It’s About Business – Cloud Options Delivers Financial Solutions
On the other side of the same coin lies Cloud Options. I’ve been spending some time recently with Cloud Options in my role as a mentor for TechStars Cloud in San Antonio where Cloud Options’ parent company, Strategic Blue, is a company in the program (disclosure – the mentor role is unpaid and I have no financial relationship with Strategic Blue). Strategic Blue is a company chock full of people with financial services experience. The CEO, Dr James Mitchell, was formerly at Morgan Stanley and this financial industry experience is a good indicator of the company’s focus.
Strategic blue aims to fill the often-promised, but yet to truly deliver space of cloud broker. For customers the value proposition is obvious – they allow customers to de-risk their cloud infrastructure. Rather than having to try and guess what kind of mix between reserved instance sizes and on-demand infrastructure they need, Cloud Options allows them to simply buy IaaS as, and when, they need it. Cloud Options fills the role of financial intermediary between vendor and customer – offering customers capacity on a truly flexible basis.
For vendors however there is a significant value gain here – by, in theory at least, enabling vendors to sell all their potential capacity, Strategic Blue helps to smooth the capital outlay cost curve of the vendors – by giving them a way to derive revenue from all of their potential capacity, a forward options service for cloud does what brokers for other commodity goods do, gives the supplier a degree of certainty. True they may be selling some capacity at a discount price – but it’s certainly better than having it stand idle. It seems to me that, given the massive investments in cloud infrastructure, and the fact that the market is still in its infancy, anything that helps the CFO o an IaaS company to smooth its investment and return curves is a good thing – this is the very value prof that Cloud Options offers to supply side organizations.
Chunky is Fine, Visibility is Key
Of course many people argue that the slight negative impact of only being able to buy infrastructure in fairly chunky blocks has such a small impact on the economics of cloud that it’s hardly worth worrying about. Rather they suggest that gaining visibility over the true cost and usage is the real issue here. They point to the old saying that “from knowledge comes power”. This is the space that a number of young vendors such as Newvem and Cloudability (disclosure, I’m an investor and advisor) alongside the IaaS vendors themselves who are starting to give customers more options around exporting their usage data so they can gain some visibility over their real usage and cost.
The graph above is generated from an AWS usage export and gives the user information regarding what they have running, and when. It gives them insight into the timing of their customer demand, and it helps them understand their historical usage with a view to estimating their future usage. I t gives them the ability to make some decisions around purchasing (in the case of AWS) reserved instances, pre-priced and committed chunks of capacity for a future need. These graphs are also useful for the finance department as it can show just how spikey their load profile is, and can demonstrate exactly how many physical servers would be sitting idle in the basement if they weren’t “in the cloud”.
Summary – Different Strokes for Different Folks
There are no black and white answers here beyond the fact that gaining some visibility over an organization’s cloud spend is a no-brainer. The fact is that in some instances, just having that visibility – be it from a third party cloud spend vendor or the IaaS vendor themselves – and taking a bit of a punt on some reserved instances will do enough to smooth out the economics for most organizations. Those who have a specific high-performance compute requirement, or those who strive to be managing as few individual instances as possible to reduce networking performance impacts will like the story that a company like ProfitBricks articulates. And for those who have very particular usage patterns, and whose situation demands a very clear forward expenditure path, a cloud broker can give significant stability and security to what they do.
Cloud is a journey and not a destination – when it comes to supply and financial issues, a culture of continuous assessment and continuous improvement will deliver the best outcomes for both users, and vendors.