Long term readers of this blog know my views on public vs private cloud debate. I believe that private clouds may dominate over public clouds among the enterprises in the short run but, in the long run, most of the workloads will move to the public clouds. The cloud economics will be a major driving factor for this shift. Even though I believe in a future where public clouds will dominate, I don’t want to dismiss the current concerns of enterprise customers outright, much like many public cloud advocates. Security is still a concern (Look out for my next post, either later today or tomorrow, highlighting this fact vis a vis Wikileaks episode) for many in the enterprises but, ultimately, the public cloud economics will seal the deal. Of course, only after further maturity in the security and other related aspects of the public clouds. There are some smart companies who are seeing this trend and already making dramatic shifts into the public clouds.
Netflix is one such company. Even though they are in the consumer space, the size and scale of their operation makes them an useful case study for this topic. Recently, they moved their 24/7 movie streaming infrastructure to AWS from their own datacenters. Netflix story is not just a big win for AWS but also has the potential to replace “The New York Times story” which is slowly becoming boring and saturated.
Randy Bias, CEO of Cloudscaling and former CTO of public cloud provider GoGrid, published a blog post where he talks to Adrian Cockcroft, Cloud Architect at Netflix, on the business drivers behind Netflix’s move. It is a great interview for anyone interested in cloud computing and the associated trends. Don’t forget to read the interview in full. In this post, I am going to pick a few points from Adrian’s interview and argue how public clouds are going to eventually win the game.
Almost all the public cloud stories starts with how public clouds eliminates the CapEx part of the economics and then goes about highlighting features like high scalability, business agility, etc.. However, the private cloud spin machines counter these arguments by focussing on the security aspects and then argue that private clouds can also offer similar scalability (of course, with a good amount of spin) and business agility. But, I am going to point out to an argument in favor of public clouds which they cannot counter with any amount of spin and, in my opinion, it will eventually make the enterprises appreciate the value of public clouds.
In August of this year, Forrester Analyst James Staten wrote a blog post about how a good reverse capacity planning can help organizations leverage the cloud economics to the fullest. In the post, he even highlighted the case of how the US Federal government site, USA.gov, took advantage of this factor and saved 90% on their hosting costs. In fact, he even extrapolated this idea to argue that AWS Free Tier is a great opportunity for Amazon to drive revenues in the long run.
A few weeks back I posted a story about how one of our clients has been turning cloud economics to their advantage by flipping the concept of capacity planning on its head. Their strategy was to concentrate not on how much capacity they would need when their application got hot, but on how they could reduce its capacity footprint when it wasn’t. As small as they could get it, they couldn’t shrink it to the point where they incurred no cost at all; they were left with at least a storage and a caching bill. Now with the free tier, they can achieve a no-cost footprint.
Clearly, one cannot enjoy this advantage with private clouds and this is going to be the biggest attraction for enterprises as they ponder their cloudy future (pun intended). This same fact was highlighted by Adrian Cockcroft in the interview with Cloudscaling when he talks about not having to pay for the unused resources after scaling back. Also, he highlights how the future lowering of public cloud pricing will help the bottom line when organizations rely on public clouds instead of their own infrastructure.
The third point is that the costs are elastic, you start paying for a resource just before it goes live, and if you stop using a resource you stop paying for it. If you own a resource it sits around a long time waiting to be delivered and installed, and if you no longer want to use that type of resource you are still paying for it for three years. When Amazon cuts prices, your installed capacity gets cheaper. When they install new instance types you can be running on them in hours, technology refresh in real time.
In my opinion, the way scaling back helps in the cloud economics is going to be an important factor which enterprise planners should take into account when they look at their cloud options. Yes, most of the enterprises don’t have their peaks like the consumer facing sites such as Netflix or Amazon retail. But, this factor will nevertheless be significant in the long run. Irrespective of our political ideology, we want to cut the wasteful expenditure in the government. It is only natural that we would want to use the same approach (Business 101) in the private enterprise as well. Unlike in some other aspects of the government vis a vis private enterprises, US government is leading the pack here. I think the enterprise managers should also be smart when it comes to planning for their future IT needs.
Having said that, I once again want to revisit what I mentioned in the first paragraph. Public clouds are still not mature and we need to go a long way before we can consider public clouds for most of the enterprise workloads. This shift is not going to happen in the next 2-4 years but, eventually, the above mentioned advantage of public clouds will change the game for good. What do you think? Fire away your thoughts.
Update: Looks like some folks are upset with the use of the term game, win, etc.. Relax, it is just my style of writing. I see the markets, politics and, even, our lives as a game in a bigger stage called universe. If I have to get away from my style of writing and do a post, the title would be “Why public clouds will get more traction than private clouds in the future”. I hope it helps. I find such nit picking amusing as it is just my style of writing. Rather, I would welcome you to criticize me on the substance part.
Related articles
- Netflix Advice on Moving to Amazon Web Services (readwriteweb.com)
- Interview: Netflix Moves to AWS Cloud (insidehpc.com)
- The Cloud Economics : Emerging Signals (enterpriseirregulars.com)
- Private cloud discredited, part 1 (enterpriseirregulars.com)
I agree that cloud economics will be a driving factor towards a more important role for the public cloud in future IT. Certainly when taking into account the possibility to scale back to a no-cost footprint and the assumption that public cloud prices will decrease when the adoption increases and more providers enter the market. (assuming here that public clouds will be mature enough enough in the next few years)
However, there still a lot of room for improvement when it comes to private cloud deployment.. Mainly on aspects such as automation and scaling the OpEx along with the workload. Given the right improvements, I believe that the private cloud still has a chance..
A similar dilemma existed in the past, namely the choice between buying and collocating a server or renting a dedicated one. For a 3 year server lifetime, the former was and still is a lot cheaper. And when running lots of them, not only do the savings become more significant, the cost of maintaining them decreases. Off course this results in a higher CapEx, but if you have the cash and in the end the cost is equal or less, why would you bother?
For the private/public cloud, the situation is slightly different. Ideally you’d have a software stack that is able to manage your private cloud fully automated, leaving your sysadmins with the same responsibilities/work as they’d have in a public cloud. And while you can never scale back in a private cloud as you could in a public one, given the right technological improvements, this cost may be smaller than the surplus you pay to your public cloud provider.
You do not hear me saying that public clouds will dominate the cloud landscape, certainly not, however I do not believe in the opposite either.. Certainly not when considering internal IT hosting business application with limited peak-behavior, constant uptime and maybe even limitations on the location those apps have to run (i.e. business-critical applications).
To conclude, I strongly believe in a hybrid ecosystem, where public and private clouds co-exist and more importantly work together. Not all workloads are suited to run on a public cloud, but a lot are. Those could run on public clouds when the private cloud has high load and could come back home when the load decreases. If the size of the private cloud is well chosen based on the knowledge of the workloads, this could be a very elegant approach. However, there’s still a lot of ground to cover until this will be possible and more importantly straight-forward to use..
What you have said above is what I consider as the short-term trend (2-5 year period). In the long term, as I said, *most* of the workloads will move to public clouds. There is no way private clouds can match the public cloud economics. Just impossible.
” There is no way private clouds can match the public cloud economics. Just impossible”
I agree totally , but if the world was just driven by economics it would look a whole lot different. Everyone would have outsourced the IT, we’d all drive small economical cars, by $10 watches and have a simple mobile (and MS wouldn’t have existed)…
So other factors are at play, people playing politics, career dynamics, what ever…
The problem with private cloud is that you are investing in “undifferentiated heavy lifting”. To be successful, you either need to hire very good and scarce people who know how to build clouds from low cost components, or pay enterprise computing prices from IBM, Oracle, HP, VMware, or Microsoft. You also need to invest in enough unused capacity (headroom) to cover your future peak needs, or build an even more complex hybrid cloud system that combines private and public capacity. The problem with hybrid is the data, you can move compute around easily, but it takes time to move the data, and you have to solve hard problems in consistency and coherence. The way Netflix runs now is hybrid, we refer to it as “roman riding”, you have two horses, stand on the saddle of the datacenter horse, then put one foot on the cloud horse, and gallop along hoping you can keep them in sync. We are moving as fast as we can to take our foot off the datacenter horse.
The best way to figure out your total burdened cost is to take your entire ITops budget, from the CIO down, and divide it by the total number of datacenter instances you are running, then convert that to an hourly rate. Even based on internal chargeback, numbers that Randy and I have heard are an order of magnitude higher than published public cloud costs. However, the primary driver for Netflix is business agility, (i.e. opportunity cost), not infrastructure cost.
Many ITops organizations will spend the next year or two turning themselves into a private cloud, while their internal customers go public, especially with new apps, and leave them running a shrinking legacy.
Thanks Adrian for your insights. I agree with you.
Krishnan:
So, perhaps I am mis-intepretting your post, but there seems to be an inherent “either/or” aspect to your argument, which undermines your assertions. If you assume that in “X” number of years the typical data center will be a blend of private cloud, public cloud and legacy infrastructure, then your argument that public clouds will “win” stops making sense. The right conversation to be having is finding the right blend public, private and legacy to optimize cost while still meet business requirements.
There is always going to be stuff in the data center that is a no-brainer for public cloud. There is also going to be a class of workloads that might or might not be a good fit for public cloud–the lower the cost of public cloud, the more willing folks will be to take the risk and move those workloads to public cloud. Then there is the category or workloads where cost is not the primary consideration, but rather it is things like reg compliance, availability, security, transaction performance, etc. If these workloads ever move to the public cloud, its not because public cloud is cheaper, its because the cloud provider can deliver other functional or operational characteristics (which probably adds to the cost, but that’s another conversation). These workloads are usually a company’s crown jewels and I think they are going to be the last ones to go–let’s face it, if a company can’t close out the qtr because of a hiccup, someone is getting fired. Are there hiccups today in Enterprise DCs? Sure, but folks have greater control over their infrastructure which gives them a better handle on the risk.
I find we always end up with the same adage: cheap, fast, or good–pick two of the three. Most folks will pick the post office for their holiday cards but pick FedEx for their escrow paperwork. I don’t see any indication that cloud providers will not evolve along similar lines.
I think you should also assume the enterprise IT will continue to evolve and continue to learn how to drive down TCO–ironically taking advantage of public cloud might be a key strategy, but at the end of the day, for critical workloads, I don’t think the cost differentials will be as wide as you think.
Regards,
Omar Sultan
Cisco
Great post, Krish. To Omar’s point, I think what Krish largely means here by ‘win’ is that public clouds will dominate the majority of IT spend. I don’t see any conceivable outcome where that is NOT the result. The simple reality is that, to Adrian’s point, most enterprises are engaged in undifferentiated heavy lifting for 100% of their IT. And perhaps 10-20% of that IT is core to the business. This is typical inside of most large enterprise businesses. I can think of dozens of medical companies that spend billions on IT that isn’t a core competency such as running desktops, email, archival storage, and related. It seems inevitable that to remain competitive all of these business will move that 80-90% of undifferentiated IT to an external provider *over time*.
Now, granted, that time scale might be 5-20 years, but it will happen. The net result is that the procurement of IT will undergo a drastic change. This is what legacy enterprise vendors, such as Cisco, are largely unprepared for. It’s also a fundamental part of what the disruption is about. Can they adapt? I don’t know, but it’s going to be a fun ride.
Always find it interesting how the debate generally excludes human resources. At the end of the day the non-human resources are about 15% to 20% of the TCO, so it seems like more of the debate should be about the value of the delivered solution which is clearly based on people, not the commodity components.