by Yefim Natis | July 29, 2013 | 1 Comment
Choosing between public and private cloud as a strategic decision is a misleading distraction, a red herring. The strategic and transformative decision is the choosing of the Cloud (or not). Private or public is a tactical follow-up.
Clearly, choosing to host an application and its data in-house or with external provider is a choice of substantial immediate impact. But it is not transformative. The strategic value of Cloud computing for a business and its IT organization is not in the hosting and sourcing arrangements. It is the transformation of IT from a custom technology shop to a retail store-front of reusable pre-assembled technology stacks offered as services.
I like to compare this transformation to the difference between ordering of a new outfit from a tailor vs. buying it at a retail store. Custom tailoring can produce unique results, but is expensive, slow and depends on your and the tailor’s talents. A retail store will offer a number of different designs to choose from, but none are specially arranged for you. This approach is somewhat limiting, but allows faster time-to market, lower costs, greater agility and on the whole better quality of service as the economy of scales allows the provider to hire top talent and deploy state-of-the art technology. No wonder that the notion of Cloud computing has been championed, if not invented, by a leading retailer, Amazon.
Enterprise IT, quite similarly, is dominated by technology assemblies that are custom-designed for each new software project. It will, also quite similarly, transition to the Cloud arrangement where project leaders select suitable software stacks from the available pre-assembled general-use patterns, managed by the cloud provider on their behalf.
Both private and public providers will compete for the business. And even for the internal provider, the best practice ought to be “public-first”: even when deploying in-house, public cloud characteristics, such as
- design for potential massive scale,
- closely-controlled multi-tenancy,
- strict isolation from subscribers,
- tracked billable resource consumption and
- openness to change and growth.
are just better design principles. For the majority of IT organizations having followed these principles will turn out to be very fortuitous as they consider, in the future, offering some of their data and services to the outside communities or public subscribers.
In summary then:
- The transformational impact of cloud computing is not in the shifting from internal to external hosting of software, but in the way that IT resources are organized, managed and consumed.
- The core impact of the Cloud is the transformational shift in IT from custom to “retail”; from building custom assemblies of software for each IT project to choosing from a list of pre-assembled, integrated and certified stacks from a self-service “retail store-front”.
- Private and public, local and remote providers will compete to fulfill the requirements of new subscribers. A subscriber in a well-designed cloud environment does not know if the service they provision is fulfilled internally or externally, by a private or public cloud provider.
- This shift applies equally to IaaS, PaaS and SaaS services. A well-designed cloud provider and cloud subscriber environments will not be exclusively IaaS, PaaS or SaaS, but rather will support all three categories of services in a uniform way. All are simply the entries in the catalog of available services.
- The well-defined characteristics of cloud computing, including elastic scalability, sharing of resources, use tracking and others – are the essential means to deliver a well-functioning cloud-provider environment, but not the ultimate goal or outcome of cloud computing.
Category: Applications Cloud cloud platform Cloud Standards Platform as a Service (PaaS) private cloud Strategic Planning Tags:
by Yefim Natis | July 5, 2013 | 1 Comment
JBoss and Google are developing a new Java programming framework: CapeDwarf. The framework duplicates all (with time) Google App Engine Java APIs and implements them using JBoss application server, JBoss Infinispan in-memory data store and some other JBoss open-source components. Google provides enthusiastic support and is developing a collection of tests to certify compatibility.
I see a great potential lurking in this initiative.
The industry needs a standard programming model for cloud (public and private). Java EE or .NET have been designed 10+ years ago for a single-tenant environment. Java EE 8 is promised to add cloud characteristics, but it is some uncertain time away and no one can be sure if it would amount to a real thing.
Here comes CapeDwarf. A Java framework that is designed by cloud-native Google that really understands Cloud and implemented by enterprise-native Red Hat which really understand Enterprise. And neither the framework nor its implementation are theoretical designs that will need years to mature. The programming model is well-tested by Google’s customers and the underlying JBoss AS is battle-proven by JBoss enterprise customers.
And did I mention that the thing is Open-source from the get-go?
A new open programming framework for cloud-based applications that is designed by Google and Red Hat is an intriguing initiative . But it is not clear that either company is “big enough” for the job. Does Google really care about the enterprise? Is Red Hat ready to lead? Care and leadership would be required to turn a small niche project into an industry standard.
Time of course will tell. But for whatever it’s worth, I think there is a great opportunity here for the two companies to contribute to the industry and to their own reputation, influence, scope and future.
Why Dwarf? Think Canaveral. Lead.
Category: Cloud cloud platform Cloud Standards Google Platform as a Service (PaaS) private cloud Red Hat Strategic Planning Tags: aPaaS, cloud, Google, java, PaaS, private cloud, Red Hat, standard
by Yefim Natis | August 31, 2011 | 2 Comments
Yesterday at VMworld conference Tod Nielsen, a VMware executive leading its platform efforts, had announced that VMforce will not be delivered, CloudFoundry technology will not run in the salesforce.com data center and users of CloudFoundry.com will be enabled to access database.com in some unspecified way as a compensating feature. Today Byron Sebastian, salesforce.com platform executive, confirmed it. VMforce is dead.
As a long-standing Gartner maxim goes: “the only real partnerships are acquisitions”. Indeed, salesforce.com followed it to the letter by replacing the partnership with VMware with the acquisition of Heroku (and now Heroku is announced by Marc Benioff to offer Java — completing the swap).
The failure of the VMforce project and the salesforce.com-VMware partnership is a warning message to VMware and other vendors, that a strategy of a cloud enabler can be difficult: the partners that are essential for the success of such policy might decide to be competitors at some point (early or late) in the partnership.
I believe that the leaders in Cloud computing will aim to control the entire stack of technology underlying their services. The providers will trump the enablers.
Category: Uncategorized Tags:
by Yefim Natis | December 8, 2010 | 3 Comments
Salesforce.com (SFDC) is spending over $250 million for Heroku that has revenue likely under $10 million (adoption momentum notwithstanding since most of the adoption is of the free version of the service). Some question the wisdom of such expensive buy. However — such challenge is short-sited.
SFDC is not fighting to get a few $$. They are fighting to be recognized as a leading platform provider for cloud computing.
SFDC has faced two problems in the past:
- Proprietary nature of APEX/Force
- Perception of most prospects that they are a CRM vendor, not a platform vendor
In 2011 they are addressing both issues:
- Delivering Database.com is clearly separate from CRM and very interesting in the platform space – they are bound to be noted as having gone beyond just apps now
- Delivering Java (VMForce) and Ruby (Heroku) will give developers SFDC multi-tenant database with standard programming. I would not be surprised if they would look at Zend or Caucho next to add a PHP stack.
The casualty of this path seems to be Apex/Force. However – not so fast:
- Most of force.com is now called database.com (only 4GL development tools remain as a separate Force.com offering not overlapping with database.com.
- Apex code (the proprietary highly-optimized language) is now referred to as database.com stored procedures. A brilliant move – all stored procedures are always proprietary and always highly optimized for data access. These stored procedures are also highly optimized for multi-tenancy – a unique feature no one will match for some time.
I believe SFDC is executing on a sophisticated vision to transition form a proprietary cloud platform to an open cloud platform without losing its users or its differentiation.
Brilliant strategy. If paired with brilliant execution — Dreamforce 2010 will be an inflection point in SFDC story — form a successful niche player to a software industry leader.
Category: Uncategorized Tags:
by Yefim Natis | August 25, 2010 | 2 Comments
Red Hat announced “Red Hat PaaS” which is in fact a Red Hat “PaaS-builder technology” for private clouds and for new aspiring public cloud providers.
The key differentiators for this offering are:
- 100% open-source
- Near-comprehensive middleware technology suite (though component-parts are not all industry best-of-breed) — there is a Portal, ESB, AppServer, BPM technology, SOA registry, rules engine, plus plus, but not a DBMS
- The “PaaS” can be built on a choice of underlying virtualization/OS/Hardware including an all-Red Hat stack, Amazon stack, Microsoft stack and other (over time)
- The “Paas”, once deployed on one platform, can be migrated to another without change to applications — this includes migration between private and public cloud deployments
- The portability is also the basis for future support for cloud-bursting (running apps in private cloud and “bursting” onto public cloud VMs to add capacity to cover spikes in demand). This capability though requires resolving the thorny issue of uninterrupted data access.
- The entire approach is shared-hardware – a suitable model for migrating current workloads and current skills, however ISVs will find that managing/creating SaaS apps on top of Red Hat PaaS is much harder than doing same on Force.com, LongJump and other shared-everything PaaS.
- The programming environment above (JBoss) is unchanged, so most existing Java/JavaEE apps can be migrated with minimal effort (that includes many WebLogic/WebSphere apps, not just JBoss apps). (minimal change also means limited cloud benefit — as with all shared-hardware offerings (such as Windows Azure))
- Red Hat is not planning to be a provider of PaaS — only enabler of other people’s PaaS (for now)
- Red Hat is ahead of IBM and Oracle in announcing a comprehensive cloud strategy
Initial thoughts on the announcement:
- Red Hat shows vision in taking a comprehensive approach to their Cloud strategy (although the Cloud benefits at this stage are limited)
- They exert industry influence, forcing Oracle and IBM to follow suit.
- This is the first significant initiative where Red Hat Linux and Red Hat JBoss division act together and leverage each other’s capabilities – an important sign of successful integration of JBoss into Red Hat organization
- Red Hat’s emphasis on Apache and standardization of IaaS APIs (Apache Deltacloud) is a credible and progressive move
- Many Cloud providers prefer open source enabling technology because they are able to specialize it. This is an important advantage Red Hat has that IBM and Oracle cannot beat.
- However in real terms the new new in this announcement is only the “Cloud Engine” layer that plays two roles: (1) enabling portability to multiple underlying virtualization technologies and (2) managing the VMs on behalf of application instances (tenants).
- The real delivery of this vision will take 12 months or more: the self-service management of the environment and the automatic scaling are the most important capabilities that seem to be missing or minimal in the first release.
- How real this initiative is in delivering real cloud-enablement will depend on how good the “Cloud Engine” is.
- The success will depend on Red Hat’s ability to attract customers that will make the Red Hat PaaS-enabling technology into a real PaaS.
What do YOU think?
Category: Uncategorized Tags:
by Yefim Natis | August 19, 2010 | 5 Comments
In preparation for Gartner Application Architcture, Development and Integration (AADI) conference scheduled for Novemeber I am developing a PaaS Scenario presentation. We have recently published a PaaS Reference Architecture research , a PaaS market assessment research and are working on the PaaS road map and more. Below is a brief Q&A on PaaS for the prospective attendees that I wanted to share here.
1 What is platform as a service (PaaS)?
Simply speaking, PaaS is middleware as a service. PaaS is the middle layer in the big-picture Cloud architecture(Systems Infrastructure or IaaS below it and applications or SaaS above it). All of the middle-layer technologies you know from the traditional on-premises deployments apply in the cloud as well. These are the application servers, DBMS, ESBs, Portals, Messaging, BPM and other middleware technologies. There is more than 10 categories of middleware that we track that are available from some vendors as a service. However of all of these, two stand out: the application platform as a service (APaaS) and the integration as a service. APaaS is the extended application server (with development tools and a data store) and integration as a service is self-explanatory. So when you hear people talk about PaaS they may actually mean APaaS (like Force.com or App Engine) or they may mean any one of multiple specialized middleware as a service offerings. Or they may mean the complete comprehensive suite of all middleware.
2 Who should care about PaaS and why?
Technology Vendors: The middleware platform is where the programming model and basic architecture of applications is determined and where standards for interoperability and portability are established. Vendors in the business software space all care about the middle layer a lot for that reason. In the Cloud this is no different – the vendors that emerge as leaders in PaaS will have major influence on the evolution of cloud computing. There are no established leaders here so far although Salesforce.com, Google and Microsoft are already making strategic investments in that direction along with a few dozen small players like LongJump, Engine Yard and many others.
Application ISVs: Most application ISVs either already do or plan to offer all or some of their functionality as a service. To do so they have to find a technical way of converting their applications from single-tenant on-premises model to multi-tenant cloud model. Some will simply use hosting services and will make no changes to the application itself (managing each customer-tenant as a separate instance of the application). This model is not sustainable long term and, lacking any resource sharing, it is not cloud. Some others will develop the multi-tenancy support themselves. This can work well, but is expensive, time-consuming and creates long-term burdens that the ISV will have to live with for a long time. The more strategically-thinking ISVs turn to PaaS providers. This is the same as the developers of new on-premises applications would use a third-party application server and other middleware rather than develop their own. It is the only long-term sustainable approach for ISVs. Today there is still some risk in choosing a PaaS (or more likely an APaaS) provider as the standards are not yet established and as a result, all of the three approaches are used by some ISVs.
IT software projects: The difference today between using an on-premises development environment (a WebSphere, a .NET platform or another) and using a PaaS (Force.com, App Engine, Heroku or another) is that the PaaS requires minimal time for procurement and is priced by subscription rather than a large up-front license fee. The PaaS technologies are also in some cases much easier to use as many use model-driven metadata-based design and execution tools. Large projects “in a hurry” and small projects with limited budgets are the primary direct IT users of PaaS today. Integration as a service is also often used by IT organizations for advanced B2B integration projects. Many IT organizations also use PaaS today as pilot experimental alternative to traditional models – in preparation for the future.
There is MUCH more to say, some other day.
Category: Uncategorized Tags:
by Yefim Natis | April 29, 2010 | 4 Comments
It has become publicly known recently that the creator of the Java programming language (and a near-cult figure with some), James Gosling, has resigned from Oracle. This seems to suggest his displeasure with the Oracle’s road map for Java (although he makes no statement to this effect). However beyond the emotional side of things, the real impact of Gosling’s resignation is minimal in my opinion, especially in the shadow of the impact on java from its acquisition by Oracle. Here are some points to illustrate what I mean:
- Oracle strategically depends on Java. Their middleware and new applications are all 100% Java.
- In mainstream enterprise computing Java is the new COBOL with the only trailing alternative – C#. It’s not being replaced.
- IBM also strategically depends on Java (though not as much as Oracle) and IBM has custom favorable terms in its license for Java, so Oracle has a limited sway over IBM’s use of Java
- Gosling leaving Oracle is 100% predictable and in fact you may wonder what took him so long
- Gosling’s displeasure only has a symbolic meaning – he does not change or decide anything. Although he can influence some developers’ attitudes — he has minimal influence, if any, on enterprise IT policies
- Oracle asserted its ownership of Java twice: once by Larry Ellison declaring at the Sun Acquisition Q&A something to the effect of “We have Java and IBM does not have Java” (not an exact quote) and the second time by co-locating Java ONE and Oracle World
- IBM was unhappy with Sun leadership of JCP and they are going to be even more unhappy with Oracle’s. I would not be surprised if IBM never certified for Java EE 6 and beyond, just to avoid giving Oracle’s JCP any credibility
- Java may lose some portability not by IBM forking it, but by IBM not adopting any new stuff that Oracle is going to bring to Java CE/ME/EE. Whatever Java IBM offers now may be the end of the line for standard Java from IBM. I guess, you might call it forking, but they will continue to call it Java and Java EE, except all new development will be IBM layers over the 2009 state of Java standards. To fight that Oracle might rename the Java EE standard again, so IBM cannot claim compliance without adopting new specs. We will see.
- Red Hat will just watch the two fight it out. They are good at it.
- The winner in this battle may be Spring – an enterprise-worthy framework that is not owned or controlled by either IBM or Oracle.
Gosling’s departure in my opinion does not add or subtract from the enterprise prospects of Java because of the massive investments made by IT organizations and leading vendors in Java and because there is no new viable alternative. But the unique good times of cross-platform portability brought about by “vendors against Microsoft” are coming to the end (regardless of Gosling). C++ programs have never been portable from one ORB to another and Java programs will no longer be portable from one app server to another in 3-5 years.
Gartner has a series of research notes published in the aftermath of the Oracle acquisition of Sun. My colleague and friend Massimo Pezzini is the lead on many of them. Hope all of you can get to them to see the considered Gartner position on this topic. This, above, is just one man’s opinion.
What do you think? Let me know.
——————————– the hobby basement ————————————-
Category: Uncategorized Tags:
by Yefim Natis | March 24, 2010 | 3 Comments
Gartner just published an important Cloud/PaaS-related research note “Application Infrastructure for Cloud Computing: a Growing Market, 2010″. Here I just want to share a footnote from that paper and see if some might want to disagree, add-on or comment in any other manner. This text is trying to deal with the nuance of some flavors of elasticity and tenancy — all-important to enabling cloud computing.
Elasticity, Horizontal Scalability and Multitenancy
Elasticity, horizontal scalability and multi-tenancy are often mentioned here and elsewhere as key distinguishing characteristics (or requirements) of cloud-computing technologies. Without attempting to provide a full-proof definition, the common-sense nature of these terms is as follows:
- Elasticity is about the ability of software to meet changing capacity demands, deploying and releasing relevant necessary resources on-demand.
- Horizontal scalability is about the ability of software to automatically expand onto additional underlying resources, or contract onto reduced resources, without interrupting operations. (Typically, horizontal scalability refers to the ability of software to automatically spread over added, or contract over reduced, number of physical or virtual machines.)
- Multi-tenancy refers to the ability of software to be offered to multiple user entities (tenants) in a way so that each tenant operates as logically isolated, while, in fact, using physically shared resources. A tenant can be an organization co-using an application with other tenants. It can also be an application co-using underlying resources with other applications.
Multi-tenant elasticity is achieved when resources are allocated and accounted to tenant application instances. Multitasking elasticity occurs at lower levels when resources of one tenant are dynamically allocated to multiple tasks inside an application instance. Multitasking elasticity improves the performance of some applications. It does not amount to a new business model in business use of the software. Only multi-tenant elasticity is the definitive attribute of cloud computing.
… and photography is my hobby. Here is Long Beach Island, NJ on a less than bright and sunny day
Category: Uncategorized Tags:
by Yefim Natis | March 16, 2010 | Comments Off
According to CIO.com, IBM will soon announce a Rational/WebSphere test and development offering in the Cloud (IBM Adding Development and Test Offering to Public Cloud)
Everyone (justly) intrigued by the new opportunity should ask themselves: why does IBM not offer a production deployment option?
Your opinion is welcome here. I will comment back to your thoughts.
Photography is my hobby. Here is a scene caught in Brussels, Belgium
Category: Uncategorized Tags:
by Yefim Natis | March 11, 2010 | Comments Off
Talking to a gathering of a group of IT decision makers recently, I collected a number of reasons why they did (or may) choose the cloud computing option to replace the traditional on-premises approach for some of their application projects. The products considered were cloud application platforms like Force.com, AppEngine, LongJump and many others (see Gartner’s Who is Who in application Platforms for Cloud Computing).
The mentioned reasons to choose a cloud application platform included:
- Reduce project risk
- Assure a level of quality
- Improve time to market
- Escape legacy systems
- Improved productivity
- Incremental project results
- Reduce the need for long engagements with SIs
This list is not a Gartner research product, but a distillation of notes I took in one conversation on the topic with some highly-engaged IT practitioners.
What do YOU say? Do you agree? Do you believe any of it? Do you have more to add to the list? Cross something off?
Comments are welcome.
Thank you for your interest and happy blogging!
P.S. Photography is (still) my hobby. Here is a recent image from Phoenix, AZ
Category: Uncategorized Tags: