Blog Post

Cloud Computing's Three-Horse Race

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

horseraceCloud Computing has hit the main stage, solidly capturing the minds of both the technology and business communities. But while three distinct deployment models have emerged, it’s far from certain which of them will go on to prosper. The three models are:

1. Renting raw hardware: compute processing, data storage and networking bandwidth.
2. Leveraging an integrated application development engine.
3. Ordering an application.

So in order to get a better sense of the prospects of each approach, let’s take a closer look at key companies promoting them and the market forces shaping them.

Cloud Hardware Rental (Infrastructure as a Service)
Consuming compute and storage services is more involved than the term “rental” may indicate, but this first step transforms and focuses most businesses on their specific value, which isn’t necessarily IT infrastructure.

The most well-known purveyor in this category is Amazon (S AMZN), but there are many other players as well, among them GoGrid, Layered Technologies, Joyent and Mosso/Rackspace.

The rental model provides one-to-one replacement of current enterprise infrastructure, except off-site and easily accessible. By choosing CPU type, memory amount and disk configurations of the lowest common denominator, customers of cloud hardware rental know exactly what they receive, can control the infrastructure at a fine-grain level and can easily compare pricing.

But it takes a certain skill set to load, operate and maintain applications in the cloud. Companies like RightScale and Elastra aim to streamline this process. And, unfortunately, Internet outages still occur, as evidenced by the Sprint-Cogent feud in late 2008, so everything may not always be at your fingertips.

Application Development Engines (Platform as a Service)
Sophisticated software offerings now easily tie together underlying hardware, touting “single-click” deployment and scaling of new application instances and machines. These engines also integrate common application components such as a data store, user authentication framework and cache. Some exist only in the cloud; others can be deployed privately.

In this category, Google (s goog) offers AppEngine, Microsoft (s msft) offers Azure, and Amazon has a suite of web services on top of basic server rental. Other software platforms that help speed up application development and deployment include 3Tera, Enomaly, and Eucalyptus, although all differ slightly in approach.

Compared with conventional application development, this strategy can slash development time, offer hundreds of readily available tools and services, and quickly scale. But developers still need to learn the platform, and such an approach may not support all programming languages, leading to concern over application portability.

Order an Application (Software as a Service)
Almost all companies use office supplies, but how many companies have their own paper and pulp processing facilities? The same question can be asked about installing, maintaining and operating enterprise applications. For most companies, ordering a cloud-based, software-as-a-service application lets them to get started in minutes, and provides plenty of readily accessible customization. (s crm) is the poster child of this category. But other sizable applications come by way of Workday and Aravo, both of which recently touted their largest customer wins, Flextronics International (s flex) and General Electric (s ge), respectively.

But for companies of a certain size and scale, the licensing model might not fit perfectly, and the quest for specific valued features, reports or integration may only be achievable through in-house deployment. Security has also been a big question around cloud and software-as-a-service applications, but as Alistair Croll points out, those questions deserve reexamination.

Placing Bets
So where are we headed next? Undoubtedly a combination of all three deployment models. Some companies, like data warehousing specialist Vertica, are taking the debate off the table by offering customers software, appliances or cloud services on top of Amazon’s infrastructure.

But I believe that for companies providing new cloud offerings, the software-as-a-service or “order and application” approach will prove the most rewarding.

ms_garyorenstein_018_72dpiGary Orenstein is the author of “IP Storage Networking: Straight to the Core”, host of and was a co-founder of Nishan Systems (now part of Brocade).

39 Responses to “Cloud Computing's Three-Horse Race”

  1. It is very nice to see that open source solutions will take over major parts within the cloud computing arena – as (some years earlier) within the internet scene. Projects like i.e. XEN, openQRM, XCP or Eucalyptus offer a wide range of new and dynamic high quality / enterprise class computing infrastructure from smallest to largest IT scenarios.

    From my view the most innovative ideas and technologies come out from open source projects like xen, kvm and others.

    It is a part wide obscure that i.e. A company like Microsoft communicates their new operating system Win 7 as a “typical cloud system” or a “cloud affine system” as most of typical cloud technologies are still reality for others – in many ways – but non proprietary.

    Just let’s do it.

    Syndicat IT&Internet

  2. Recently, i read your blog that suggested that there were NOW only THREE players in cloud computing. They were listed as, Platform vendors (PaaS), Infrastructure Vendors (IaaS), and Software Vendors (SaaS). It seemed logical to me, until I began to think is that what we all really want as consumers and enterprises, three services and a big sign reading “Some Assembly Required”. I have always hated those three words at Christmas time, but perhaps I hate them even more when I’m at work. Then why is it, that when I buy many of the “Software as a Service” solutions out there these days, I then have to go out and buy the platform (Cloud) and infrastructure to run them on. Why isn’t software, Infrastructure, and Platform all packaged and deliver as a single priced, easy to consume Service? Isn’t that what Cloud Services were intended to be? Seems to me, that we all need a fourth category!

  3. Gary Orenstein

    @Andreas. > Salesforce is not only a SaaS but also a PaaS Player with its and appexchange offerings in my opinion. > You are right. But I think they are where they are today (>$1B in revenue) because of their core CRM application. That said, I also believe that their platform and appexchange initiatives have taken some attention away from the core application. As a user, there are times when it appears ready for a refresh.

  4. The comments have been as informative as the original article. If I put my Information Architect hat on for a moment I have to point out that outsourced computing of any kind does not alleviate the organization from having good governance and risk management processes in place. This means that for any application deployment in-house or outsourced, the complete life-cycle should be addressed from pilot through end-of-life transition. So, data management whether for disaster recovery or platform migration should not be a surprise or unforeseen expense, but as @rohit pointed out it must be planned for.

    Regarding the author, @Gary_Orenstein, follow-up comment on creating an OSI like layer model for the various services – it’s not directly comparable and probably overly-complicated, but it could form the basis of a classification and rating system which does appear to be needed.

    Whether it s a service run out of someones bedroom with a front end purchased through Apple’s app-store or a major CRM implementation layered over Google or Amazon infrastructure is not the point, Both services can be valid but clearly both services would be in different classes (if a classification system were defined) and both services could benefit from ratings – both may even be rated AAA within their respective classes, for example if there was a rating system. Where we seem to be heading is the need for an independent body to provide said ratings and classifications.

    That said, this might actually be a cool and potentially quite lucrative start-up to help build if any entrepreneurs and VC’s are reading this :-)

  5. The economy sure is helping to act as a catalyst for the typically slow cultural shift that lifts new technology and ways of working (like SaaS), so there is tangible proof of a silver lining to the credit crunch. As a major “Order an App” provider, we have seen a lot more incoming leads in the last quarter because of this.

    Now taking a helicopter view on the popular “Order an App” space, one of the prime keys to success is in the SaaS provider’s ability to meld their offerings into the customer’s master data management (MDM) fabric. was lucky in that their offering can be operated by many businesses in an ivory tower, requiring only ‘human integration’ with the sales force. Cracking the code on SaaS supply chain and SaaS multi-enterprise networks is a whole different ball game and an integration strategy with ERP/MDM is a must – which will quickly separate the men from the boys when the sales cycle moves from the Business owner over to the more sophisticated tire-kickers of systems, over in IT. I have first hand experience of this, most notable when wrangling the Johnson & Johnson IT team into a deal recently, and they are super smart on SaaS i can tell you.

    In the end it’s all good and as one other commenter pointed out above me, a fantastic time to play in the IT sand box. Bring it!

    Nick Parnaby
    Founder & COO, RollStream Inc.

  6. Hi Gary,
    Enjoyed the article. We at Aster Data (data warehousing vendor) agree that giving customers options in and out of the cloud enables them to select a solution that best meets their needs. In addition to our software offering, we enable customers to run in either the Amazon or AppNexus clouds. One interesting point on scalability is that we are actually seeing customers using our cloud services at much larger scale than many people thought manageable at this time. For example, our customer ShareThis, is running a large DW on Amazon’s cloud. They have found that with Aster on AWS, they can backup up to 2.5TB data into a 500GB compressed backup under three hours. While taking these backups, ShareThis can also load data and perform queries seamlessly. Here’s some more info:

  7. We all know cloud is going to be on us soon. But whether we are equipped to work in cloud environment. Probably not. The recent failures of Gmail and other reliable servers during short span of time has raised lots of doubt over future of clouds. I am sure rollout of clouds either as a service or as a platform is not going to be smooth. Further there are many issues like security and safety of documents which have not yet been addressed. I believe the roll out will have to be in stages.

  8. Salesforce ist not only a SaaS but also a PaaS Player with its and appexchange offerings in my opinion.

    I think, all three models will have their place in the next couple of years, because all three models offer solutions for different kind of problems. Software users will buy SaaS and PaaS offerings, where sofware vendors will use IaaS and PaaS offerings. The borders between these three kinds will blur over time, especially between SaaS and PaaS.

    but everything moves to the cloud…somehow :-)

  9. Gary Orenstein

    @ Raj R. I could easily see more decisions being made in this sequence over time: what’s my top application choice for this task > what platform is most optimized for deployment > where can I have that infrastructure reside.

  10. Gary Orenstein

    @Erik, Thanks for clarifying. I was trying to draw the distinction between vanilla server rental and something more. But as Rich M. pointed out, that may have been unfair to both the Infrastructure and Platform ends of the spectrum. Perhaps we need a 7-layer cloud stack (similar to the networking world) that will help us make the distinctions. Clearly two categories between infrastructure and platform might not be enough to make sense of all of the emerging players.

  11. Gary,

    I see this a bit differently; All three models you outlined could co-exist and probably the discussion is less about much about which of three model will take off and more about what is needed to tighten linkages between the 3 models.

    One example- Let’s take SaaS-
    Consumer and SMB segment, it could be argued, are biggest end user segment in SaaS today ( Salesforce, Mozy, netSuite, webex, Postini, Sharepoint, online backup providers etc etc etc). Enterprises are moving towards SaaS as well ,though concerns linger and ramp will more gradual compared to SMB.

    Consumers, SMB would not care about leasing a virtual data center (hello..what on earth is that they would ask?) but they would use say Tax online or PayRoll online services In turn the SaaS provider would probably either run their offering on Dedicated DC, or use a co-located facility or host their infrastructure on ParaScale or Amazon etc etc. To offer a meaningful, SaaS offer to their SMB customers, the SaaS provider would need to plan for the “Cloud” infrastructure setup at DC i.e storage, multi -tenancy, Heating ,Computing etc etc ; Quite possibly they would seek a turnkey IaaS offer from a Cloud Service provider; Also, SaaS vendor would need to plan for Commerce capabilities – billing, service activation, enablement…this would be the platform as a service piece.

    The Network , security and service delivery component for all three models is currently under appreciated.

  12. Great article summarizing the three kinds of clouds and major players. Couple of clarifying thoughts:

    1. Rackspace/Mosso actually has products in all three categories: Cloud Files and Cloud Servers for IaaS, Cloud Sites for PaaS, and hosted e-mail for SaaS.

    2. IMO, 3Tera, Enomaly, Eucalyptus, etc. aren’t PaaS related but IaaS. They are software products designed for people looking to deploy their own IaaS cloud.

    As for placing bets, my bet is on all three. Each has it’s own unique benefits (I wrote a blog post describing them here – and there is a valid reason why a user/developer might want to use one over the other. I think we’ll see all three continue to grow. The future is cloudy!

  13. Gary Orenstein

    @ Winston. Could not agree more…that is the reason I referenced the Sprint-Cogent feud that left entire regions without Internet access late last year.

    In general, I think the networking aspects of cloud computing are woefully under-rated on two fronts. One we need much better ways to monitor and deliver multiple points of access as you suggest. Separately, I think many people forget that the glue holding these large scale-out data centers together is an increasingly sophisticated Gigabit and 10 Gigabit Ethernet infrastructure. There would be no “rack ’em and stack ’em” if we did not have sophisticated LAN capabilities in the data center. Hope to tackle both of these areas in future posts!

  14. Gary Orenstein

    @ Rich. Point well taken. I went back and forth with this, and there is certainly plenty of gray area between a sophisticated Infrastructure as a Service provider and Platform as a Service provider. I’m not sure myself that those categories can remain completely separate for long, but they have emerged as general categories. It should be perfectly acceptable for companies and service providers to position their services as a unique combination of the two, and perhaps that is where we will see the leaders emerge.

  15. One might argue that when you ask “Where are we headed next” and say, “a combination of IaaS, PaaS, and SaaS,” you are talking about a private cloud deployment on-premise or off-premise in a traditional managed or co-location data center connected to the corporate network just like any 3rd party site would be but managed using cloud management tools the likes of AppLogic, Appistry, OpenQRM, Eucalyptus, Enomaly ECP, Nimbus, Gigaspaces or others.

    You own or rent the hardware (networking and servers), you install the appropriate fabric management (cloud management tools) software for your needs. You mix in the platform as a service tools (like Appistry or GigaspacesXAP). Spread the SaaS icing on top of your own custom built software (as a service w/ an API) or use something commercial (like Zimbra) for your SaaS tier if it fits.

    Where we are headed next is to the broad realization that this rapid and expansive burst of information technology innovation into the main stream branded as “cloud computing” is the path forward to the next-generation of IT deployment, development, and management both inside and outside the data center with linkage to the public cloud computing vendors resources to use them as needed and pay as you go. Work being done along the lines of VPN-Cubed is crucial when we are headed is towards a public/private hybrid model that will push the boundaries of what our networks can do more than ever before. Let’s hope all that Internet 2 work and related research is panning out. We’re going to need it!

    It’s a great time to be in IT! Fantastic things happening all around all day every day.


    Kent Langley

  16. I think the cloud is ready/good to go. I cannot say the same about the network though. Cost, reliability (@Winston), and throughput are all issues that will need to be fixed. Like Talari says, T1s cost (and perform) about the same as they did 10 years ago.

    The only issue I worry about regarding large scale cloud/app deployment is — what happens when a provider goes away. Where does my data go? I cannot bring the App in-house (it has other companies data in any case) and transitioning data to another app or another provider will be painful and costly.

  17. In any attempt to simplify by bucketing, you’re bound to get some disagreement… and I’m afraid that this is one of those times.

    I’ll take exception to your definition and characterization of IaaS. Were IaaS simply to be ‘cloud hardward rental’, there’s be little to distinguish an Amazon AWS from a conventional hosting service / co-lo player from whom you could rent a particular hardware configuation.

    IaaS is considerably more than cloud hardware rental, as the purveyors you mentioned would be sure to tell you. A better approach to definition is “cloud datacenter rental”, in which the customer gets to select from a set of off-the-shelf and partially modifiable pre-fabricated configurations of servers, networks, storage and the core services (e.g. DNS, DHCP, LDAP, … ). A CONSIDERABLE amount of server, network and storage operation, administration and management is “taken care of”, though modifiable if your requirements demand something “out of the ordinary.” IaaS definitely can be considered BYOAP – bring your own application platform – as well as your own specific apps and data (though, in the case of Amazon, they have a nice collection of data sources you might want to use).

    If viewed this way, IaaS would clearly INCLUDE 3Tera, Enomaly, and Eucalyptus as players in this camp. Based on the catalog of prefabricated servers and “constellations” of servers that make up an n-tier application (e.g., a web-based application driven off a specific relational database platform), a 3Tera-equipped service provider could be considered both an IaaS and a PaaS player.

    I realize that this is still a matter of personal opinion, since there are no iron-clad definitions of the terms. However, you do the IaaS providers (and the efforts to which they’ve gone) a disservice to set the bar so low. You also detract from the very distinct sets of expertise in application platforms required to deliver PaaS by lumping the infrastructure providers with the platform folks.

    = Rich Miller

  18. Bottom line – you have no network – you have no app – Do you have multiple ISP’s with Multiple different connections to the internet? Even when you think you do you may not (one former client I dealt with had two different ISP’s – two different (redundant) T1 lines BOTH of which ended up 300/400 yards away in the same underground conduit which got blasted to bits in a gas explosion. Does your cloud computing vendor have similar redundant network access? Verified? Guaranteed SLA? If you can’t get to the cloud based app you’re in trouble. If your customers can’t get to your cloud based app because their network connectivity is gone then you may still be in trouble too. I think cloud computing, SAAS and grid computing have their place – just not sure its ready for mission critical deployment as yet.