Vyom

कर्मण्येवाधिकारस्ते॥

Why Gartner Analyst is wrong about Cloud Computing and OpenStack

Its been over two weeks since my last article, I have been busy with work and family. Yesterday I read an article Why vendors can’t sell OpenStack to enterprises, while I agree with some of the observations, the overall article seems biased and do not reflect the real cloud computing and how OpenStack is trying to address those challenges for enterprises. Nowhere in the article they mentioned CloudStack, oVirt, Eucalyptus, OpenNebula or any other major cloud computing platform.

Why Gartner Analyst is wrong about cloud computing and OpenStack

On the contrary article make a reference to VMware vCloud Suite, BMC Cloud Lifecycle Management, Cisco Automated Suite for Clouds, Microsoft System Center, IBM SmartCloud Orchestrator, HP Cloud Service Automation, CA Automated Suite for Cloud, most of which are not really true cloud management and orchestration platform in real sense, but are more a combination of Data Center Management and Server Virtualization Management (i.e. Virtual Machine Life Cycle Management). If any enterprise needs to built a private cloud computing platform using these tools, they will need to invest significant amount of technical resources (both devops and programming) compared to building it on OpenStack, CloudStack and its other open source siblings. Indeed what most enterprises call Private Cloud is in essense a Server Virtualization Platform either with or without Data Center Management Platform which helps them to manage virtual machine life cycle on a set of physical servers with some analytics.

Moreover article claimed “for one promising or successful deployment there are several that fail and that will forever remain undocumented”, which is a hyperbole giving readers an indication that most OpenStack implementations fail without any supporting statistics. Based on my personal experience of working in enteprises for over 15 years, it seems more true for most of the large IT projects done using commercial technology solution and platforms from tier one provider (some of which by the way are classfied as visinary and leaders in Gartner Quadrant). If you want to see more details please see Why big IT projects crash and Why Project Fail

I also attended the OpenStack Summit in Hong Kong [1] and there are valid points in the article, which I also discussed with few people and vendors. But in no way this implies Vendors Can’t Sell OpenStack to Enterprise, indeed on the contrary found OpenStack to be very suitable for enterprises and private cloud than public cloud for which I feel Amazon compatible CloudStack and Eucalyptus are more suitable. Also during my own research in Cloud Computing, I came across multiple definitions of what is meant by Cloud Computing, so instead of focusing on a narrow definition of Cloud Computing littered with marketing terms, I used the definitions by NIST which I explained in my earlier post What is Cloud Computing. In the following sections I will try to address each issue raised by Gartner Researcher and also try to explain why OpenStack and for that matter even CloudStack or Eucalyptus or OpenNebula are ready for enterprises.

1) Lack of Clarity about what OpenStack does and does not

What is OpenStack

In this point the researcher contends that OpenStack is positioned as an alternative to Commercial Cloud Management Platform (CMP), as they are called in Gartner. This is indeed untrue, if you look at the figure above and read the definition of OpenStack [2], it gives you a clear idea of What is OpenStack. In the same paragraph Gartner researcher contends it needs to be augmented by a commercial CMP for being useful in enterprise private cloud and here I smell a rat.

Indeed instead of taking a narrow CMP view and definition of Gartner, I use a more neutral cloud computing definition from NIST and try to explain how OpenStack platform can help enable true private cloud with or without commercial CMP. This will directly address the first point regarding what OpenStack does and doesn’t.

Definition

Cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction. This cloud model is composed of five essential characteristics, three service models, and four deployment models. Using this definition let me clarify how OpenStack can directly be implemented in Enterprise and vendors can sell it.

OpenStack and Five Essential Characteristics

  1. On-Demand self service

    This is one of the areas which is part of OpenStack since its inception. All the computing resource which includes compute, storage, networking, identity management, image store etc. are available and exposed as an API. This means they can be provisioned on demand without any or little human intervention. Now each enterprise use different policy and workflow for enabling and provisioning computing resources and OpenStack provides tremendous flexibility. This capability can be implemented without any Commercial CMP with long term support, moreover due to open nature of the API’s and availability of the source code enterprise will avoid vendor lock-in.

  2. Broad network access

    OpenStack provides Pluggable, scalable, API-driven network and IP management. It is part of OpenStack platform and with Neutron, it is possible to implement broad network accesss of any kind complying with enterprise network topology. Obviously every enterprise has its own network design which is unique, so it does need efforts to design a blueprint, but its true for any cloud computing project be it using commercial or open source. But compared to any commercial CMP which are mentioned in article I feel OpenStack and CloudStack are way ahead. Indeed VMWare need to acquire Nicira and still Nicira integration with OpenStack and CloudStack is much better then VMWare vCloud itself (obviously it will change with time but there are more then one option for enterprises).

  3. Resource pooling

    OpenStack and other open source cloud platforms support multiple mechanism like baremetal, light weight containers, hypervisors, storage, networking technologies which makes them ideal at resource pooling. So if enterprise use OpenStack or some other platform they have more freedom pooling disparate resources for their enterprise cloud computing services which is a typical situation in any enterprise be it large or small. Again due to open source nature of these cloud computing platforms they will always remain ahead of any commercial CMP’s in terms of the diversity of technology and the level of resource pooling available.

  4. Rapid elasticity

    By virtue of being open source OpenStack and similar cloud platforms offer unprecendented rapid elasticity to private clouds to scale horizontally or vertically for any application workloads without tying the enterprises into heavy investment and licensing restrictions. Moreover if the internal enterprise cloud cannot handle application or computing workloads, these platforms gives flexibility to enterprise to use any public cloud services automatically to scale without human intervention or typing them into specific vendor lock-in. On the other hand Commercial CMP’s severely lack in this area due to inherent licensing restrictions tagged on in the form of server instance, cpu cores, number of devices or storage capacity etc. So indeed in this aspect most of the Commercial CMP are detrimental for Private cloud then useful.

  5. Measured service

    OpenStack with the new project Ceilometer which provides a very scalable metering and billing service that can be customized for enterprise based on their own charging mechanisms. If Information Technology (IT) organization or division within an enterprise is cost center, it can be configured to automatically calculate chargebacks for individuals and departments and post it to internal finance and budgeting systems. If IT department or division is a revenue center it can be configured to calculate the computing resources and generate invoices for specific department or individuals within an enterprise. Again since each organization have their own unique budgeting, chargeback, pay per use charging mechanisms, OpenStack and for that matter even CloudStack with its cloud-usage plugin provides all the necessary data for enterprises to do this measuring and billing service without human intervention. Obviously many enterprises prefer not to build it themselved but rely on third party vendor to help them design their chargeback or invoicing mechanism and integration. Commercial CMP do not have such level of flexibility in this area.

OpenStack and three service models

  1. IaaS

    This part of current OpenStack offering and can be implemented fully within an enterprise with more flexibility and openness then can be offered by any commercial CMP’s in the article.

  2. PaaS

    Obviously since OpenStack supports multiple mechanisms like containers, bare-metal, hypervisor based compute resources, multiple different storage technology, different pluggable configuration management tools etc. it can directly support enterprises to build OpenStack PaaS cloud, commercial CMP are not as successful as open source platform in this area. Personally I am not in favor of extending core Cloud Computing Orchestration and Management platform to include PaaS, but provide a plugin mechanisms to build them. So in this regard there are many projects like Pivotal CloudFoundry, Redhat OpenShift. In this regard my personal opinion is that OpenStack should not make Solum as part of the OpenStack core project but keep it separate similar to OpenShift or CloudFoundry.

  3. SaaS

    OpenStack and CloudStack are more flexible and open platform for building enterprise private SaaS platform and indeed are very successful with many SaaS companies. But since this is not the focus of the Gartner article I will not explain it in details how it helps SaaS platforms.

OpenStack and four deployment models

Obviously OpenStack supports all four deployment models and there are successful case study in all four of them. I haven’t seen a single Commmercial CMP with successful case study in all four deployment models. Since Gartner analyst’s article was focused on Private Cloud in this area also there are several successful case study of OpenStack and its other open source cousins but it did not receive as much press coverage as the Commercial CMP’s do, obviously due to amount of marketing dollars spend on them. If press don’t know it does not mean it does not exist.

2) Lack of transparency about the business model around OpenStack.

OpenStack is a very young project and business models around it are building and some of them will bear fruits in time to come. This does not mean its not ready for enterprises today, as I covered in earlier section OpenStack is ready for enterprise and vendors can sell it today. Indeed in OpenStack Summit there was a video presentation from Redhat Red Hat Keynote - Truth Happens (For China users Red Hat Keynote - Truth Happens - China), that succintly can answer this concern. Readers can find more success stories at Open Source Stories.

Also one thing which every vendor needs to understand is that private cloud implementation is not simply install, configure and manage a piece of software. Its much more and requires proper understanding of Cloud Computing and how it will generate business value for enterprises. Since cloud computing is so new it will take some time for vendors to digest and understand, as it is disrupting their traditional models with which they worked for decades. Being part of the open source eco-systems and projects within OpenStack, it helps them to prepare for demonstrating business value and at the same time improving OpenStack to deliver that business value. Vendor participation in OpenStack is not philonthropy, every vendor participating in the process is not just contributing, but learning, preparing, training and improving their own team to be ready for next disruption and in directly helping the overall cloud computing eco-system. This investment is far less then the money they will spend on consulting and market research firm to tell them what to do, without any tangible outcome. Participating in open source cloud computing platform helps them with tangible outcomes, not just with a piece of powerpoint, pdf or a report.

As for low adoption rate and few deployments, it can directly be attributed to lack of knowledge and awareness about Cloud Computing within IT organizations of traditional enterprises. I will be very happy if market research and analysis firm (like Gartner, Forrester) can take lead in this area and promote awareness about Cloud Computing within IT organization of traditional enteprprises. Hopefully they can steer them away from narrow cloud computing definitions and visions promoted by Commercial CMP’s. I am sure once the IT Organization understand the basics of Cloud Computing, they will adopt and launch multi-million dollar projects with OpenStack or its other siblings for implementing private clouds and integrate them with public clouds in ways unseen before, its just a matter of time.

3) Lack of vision and long term differentiation

In this I do agree partially with the Gartner Analyst that many vendors just focus on making OpenStack installers and differentiate themselves by offering a piece of software which is easy to install and manage. This is due to lack of Cloud Computing knowledge within the industry both among vendors and enterprises. Hopefully organization like Gartner, Forrester, large vendors, governments, large enterprises in time will fix this.

But with respect to long term vision I do not see any issue, indeed in last three years, since the time I am following this project I have seen constant progress and improvements with some issues in between which are normal for any large project of this size.

Also I don’t agree that vendors need to differentiate OpenStack to sell it to enterprise. If vendors remain true to demonstrating the business value they bring to enterprise with OpenStack platform for Cloud Computing, they will be able to sell. Obviously here the age old Survival of fittest will apply, vendors who are able to demosntrate a better business value will survive and the one’s who could not compete in demonstrating value will slowly die. So I will suggest OpenStack community to focus more on improving the platform and provide a demonstrable business value then superficial differentiation based on unneeded features.

4) Lack of pragmatism

Based on my personal interaction with the vendors in OpenStack Summit I did not encounter this lack of pragmatism. Some vendors were already offering OpenStack configurations which integrates hyper-v and VMWare to build Cloud Computing within enterprise using OpenStack. Indeed VMWare is another tool for managing virtual machine lifecycle and some features for data center management. The way I see it is, if enterprise has invested in VMWare and want to extend this platform they can very well implementa a Cloud Computing Infrastructure using OpenStack (personally I work with CloudStack and know that it can be used with VMWare directly and assuming it is also true for OpenStack).

Since OpenStack supports pluggable architecture for hypervisor, storage, networking I do not see any issue in having multiple different hypervisors including Microsoft Server Center as part of the larger OpenStack eco-system. Indeed based on my latest information the OpenStack from Suse, support hyper-v from microsoft along with VMWare out of the box. OpenStack is not just a virtual machine life cycle management platform like vmware or other so-called commercial CMP, but is a large platform which provides a pluggable, API based interface.

I feel the main reason enterprise choose a vendor in long term is based on how a vendor is able to demonstrate the business value of the services or products its providing to enterprises. In any eco-system there are good and not so good vendors its the same problem with any commercial technology product be it from SAP or Oracle or IBM. So its not lack of pragatism but different camps in a large project which is normal.

Conclusion

Comparing with any commercial CMP mentioned in the article I still feel OpenStack and its open source siblings like CloudStack are better at addressing Cloud Computing needs of enterprise with pluggable, distributed, scalable, REST based services driven architectures. Lack of interest from enterprise for Cloud Computing has more to do with:

  • understanding it,
  • difficulty in deriving business value,
  • lack of quality analysis, articles and materials available to understand it,
  • vendors are not yet ready to demonstrate real business value they bring by implementing Cloud Computing in enterpirse,
  • traditional IT vendors sold many technology platforms, products in the name of commercial CMP’s to enteprise under the disguise of Cloud Computing without demonstrable business value,
  • and last but not the least narrow definition of cloud computing by large vendors to suite their own product portfolio.

All this aggraveted the situation of cloud computing with enterprises and made the job of new platform much harder. But this will change with time, when vendors will be manture enough to demonstrate the business value and enterprise understand how they can derive business value. Until then its a normal trajectory of market development. But again I do not agree with the analysts conclusion and underline my initial thought that Vendors can sell OpenStack today, obviously it will require them to demonstrate business value to enteprise customers, but that is true for any technology product or solution.

Footnotes:

[1]About OpenStack
[2]The OpenStack Summit Hong Kong 2013