While the virtualization of servers running custom applications is mature, we looked at the next logical step: moving those applications to the cloud. In this MarketPulse report, we looked specifically at compa-nies’ confidence and readiness to deploy their custom Java applications in the cloud. We gauged their cloud adoption stage, evaluated data management technolo-gies and how they affect cloud adoption, and consid-ered technical and business factors that will shape their cloud initiatives.
The transition to cloud computing is clearly an evolu-tionary process. Step one starts with virtualization of the application. To fully harness the opportunities provided by this new architecture, however, step two in-volves developing your applications in a lightweight ap-plication development framework such as Spring. Step three is deploying in a lightweight runtime like Tomcat or VMware’s enterprise version of Tomcat, tc Server— which are ideally suited to run on virtual platforms. Step four involves implementing data management technolo-gies specifically geared for the cloud.
The Cloud is For Real
As evidenced by the research report, cloud computing is not pie in the sky. For example, 57 percent of respon-dents currently have one or more applications running in the cloud, 31 percent plan to deploy one or more applications in the cloud during the next 12 months, and 24 percent have plans to deploy one or more cloud ap-plications during the next 12 months or beyond.
a new IDG research Services report reveals that
virtualizing custom applications is just the first step to
enabling development and deployment in the cloud.
Cloud computing is gaining widespread adoption, and for good reason. Lower costs,
increased flexibility, and the ability to scale IT resources up or down as needed are
among the primary benefits that have a majority of CIOs leveraging the cloud. In a recent
survey conducted by IDG Research Services, IT executives reported they’re well on the
way to application development in the cloud, but significant challenges await, especially
around data management in the cloud.
Systems ride the Cloud
Despite its rapid growth and acceptance, the transition to the cloud for custom applications is not an instant fix. It requires the evolutionary approach noted earlier: virtualizing applications, implementing a lightweight de-velopment framework for quickly building applications and deploying them in a lightweight runtime environ-ment, and the data management component. The clear conclusion of this report is that a large major-ity of respondents are already following this evolution-ary approach. For example:
n 89 percent are either already virtualizing their appli-cation infrastructures or planning to do so. n 83 percent have begun developing applications in a
lightweight application framework or have plans to do so.
n 82 percent are currently developing applications on a lightweight application runtime framework.
n 89 percent are either investigating new storage tech-niques or planning to take this step.
While many organizations are proceeding down this path, it is still early in the cloud rollout, and most users are only cautiously moving from the comfort of their established server virtualization environments toward virtualizing their data centers.
Next-Generation Application
Development
Spring, from the SpringSource division of VMware, is a lightweight runtime framework that exemplifies the progressive state of application development and the steady migration away from complex application methodologies. Most often, Spring is used in conjunc-tion with Apache Tomcat, an agile, fully funcconjunc-tioned application server that has rapidly grown in popularity by providing an alternative to traditional, monolithic application servers such as WebLogic and WebSphere. Unlike traditional application servers, which may have a memory footprint of hundreds of megabytes, one of the core differentiators is the lightweight nature of Tomcat: it has a 10 megabyte memory footprint that is ideally suited for modern virtual environments. VMware provides an enterprise version of Tomcat—vFabric tc Server—that provides the operational management, advanced diagnostics and mission-critical support capabilities businesses need. Together, Spring and Tom-cat or Spring and tc Server provide a fast and flexible application development and deployment environment that reflects the speed and dexterity users have grown to expect—and demand.
This lightweight characteristic becomes even more important when you consider that these applications will run on a virtual platform; it is difficult to realize the benefits of virtual infrastructures when there is no choice but to allocate gigabytes of memory from resource pools just to run an application server that is already overly burdened. Another benefit: the abil-ity to rapidly provision these lightweight application servers to accommodate the unpredictable spikes in demand that come from data-intensive, Web-oriented applications.
Tomcat has also created a highly compatible environ-ment for customized Java applications in the cloud, a trend that the report found to be gaining momentum. When respondents were asked to describe when they would be deploying custom Java applications to the cloud:
n 21 percent said they would do so within the next six months.
n 29 percent said within 7-12 months.
n 38 percent said within the next 13-24 months.
The MarketPulse research report found that four business factors rated very or somewhat important when addressing the elastic application demands of infrastructures:
n Reducing capital costs (97 percent) n Business agility (97 percent) n Reducing operating costs (96 percent)
n Simplifying the deployment process (96 percent) Says David McJannet, Director of Product market-ing at VMware, “What we’re really talkmarket-ing about here is time to value. How long does it take me from idea inception to actually being able to deliver an application? Pure business value is about getting applications out there more quickly. Developer productivity is the thing. There’s really no question that deploying applications is much, much faster using Spring and Tomcat.”
But business value continues to grow after ap-plication deployment. That’s where agility kicks in, by quickly allocating resources to accommodate volume demands. As an example, McJannet cites a trading application in a financial services envi-ronment. If there is a spike in trading activity—say, the price of gold has gone way down, and there is a big rush from people who want to buy it as a result—and the trading application can’t handle the spike in demand, buyers go elsewhere. Although there are certain applications, such as daily inventory reconciling, that are not prone to spikes in demand, people are progressively build-ing applications that are more Web-oriented, more data-intensive, and more likely to be asymmetric in terms of spikes. By deploying a lightweight framework, users are able to service periods of high and low demand without having to go out and buy new servers that will rarely be needed in the future.
Minimizing Time
to Value
expressed that same level of confidence.
McJannet cautions that we are still in the early days of cloud computing, and therefore many organizations are still in the process of converting applications to the cloud. He goes on to say that we do know that Tomcat users are confident in their application server runtime because it is lightweight, can be quickly provisioned, encourages the creation of much lighter-weight ap-plications, and is ideally suited to run on the pools of virtual infrastructure that are foundational to the cloud.
Workloads Matter
More than one quarter of respondents described their process loads as periodic (29 percent) or periodic spikes (29 percent), while 23 percent described them as asymmetric, and 19 percent reported a steady increase. Those with periodic or steady state process loads, or steady state increases are increasingly likely to have more applications currently running in the cloud (67 percent vs. 48 percent among all others). This group also expects a higher percentage of Java applications to be running in the cloud within the next six months (26 percent on average vs. 16 percent among all others) and within 7-12 months (35 percent on average vs. 24 percent). This seems to contradict the conventional wisdom that application loads with large spikes are the prime candidates to move to a cloud.
McJannet addresses that seeming contradiction by “Tomcat has become the most widely adapted
applica-tion server out there, period,” says David McJannet, Director of Product Marketing for VMware’s Application Platform Division. “There is a general realization that you don’t need these big, giant application servers because they are cumbersome and unwieldy to work with. Therefore, people have gravitated toward these lighter weight approaches, which allow them opera-tionally to do things much more quickly.”
Runtime Ready for Primetime?
One of the key findings of the survey is that Tomcat users are clearly the most advanced in their journey to cloud adoption. The survey found that Tomcat is one of the top primary application servers in use for the development of custom applications, as well as within companies’ production environment. Confi-dence is a key to success in any endeavor, and when people were queried about their confidence in their application server runtime suitability for the cloud, Tomcat users rated a much higher confidence level than non-Tomcat users. Specifically, the report found that 38 percent of Tomcat users are significantly likely to be very confident that their companies’ current application server runtime is suitable for the cloud, while only 11 percent of non-Tomcat users
Source: IDG Research, June 2010
Source: IDG Research, July 2009
Application Server Runtime for Cloud
How confident are you that your company’s current application server runtime is suitable for the cloud?
(Net) top 2 Box Very Confident Somewhat confident Not very confident Not at all confident
total 81% 24% 57% 12% 4% Tomcat users are significantly more likely to be very confident that their companies’ current application server runtime and data management technology is suitable for its cloud application infrastructure.
Why Java Matters
Within the next 6 months Within the next 7-12 months Within the next 13-24 months 0 5 10 15 20 25 30 35 40
saying it’s not an either-or situation. “I think the reality is that people are gaining confidence in the cloud ap-proach for a broad range of applications. While ideally suited for workloads with usage spikes, this deploy-ment model is clearly attractive across a range of applications.”
Data Management Confidence
While most organizations have invested in a shift toward a lightweight approach, when it comes to the choice of data management technologies required for this new deployment model, they were not as far along. The report found that respondents’ confidence in their data management capabilities such as data-bases and caching capabilities is not uniformly high: n 72 percent of respondents were very confident or somewhat confident that their data management technologies such as databases and caching are suitable.
n 25 percent were not very confident, or not at all confident, in this key area.
In the case of Tomcat users, responses were signifi-cantly different, as 30 percent of them—compared to 11 percent of non-Tomcat users—are significantly more likely to be very confident that their companies’ data management technology is suitable for their cloud application infrastructure. McJannet says that is
likely because Tomcat users tend to be early adopt-ers of technology and are therefore further down the investigation path.
McJannet also says the demands of cloud applications have exposed challenges for traditional data manage-ment technology because a typical application with a Web server, application server and database is bound to eventually become stressed as an increasing amount of data is accessed and returned to the database. “At some point, the data access challenge becomes the bottleneck,” he notes.
In response to these challenges, McJannet says a growing number of people are putting in some kind of caching layer between the database and the applica-tion server to allow the applicaapplica-tions to more simply and quickly scale to meet usage demands—essentially allowing the applications to interact with the very fast caching layer instead of the database. This caching layer allows them to provision capacity much more quickly, as opposed to spinning up another database every time application workloads expand.
SpringSource tc Server provides enterprise users with the lightweight server they want paired with the operational management, advanced diagnos-tics and mission-critical support capabilities busi-nesses need. Designed as a drop-in replacement for Apache Tomcat, tc Server ensures a seamless migration path for existing custom-built and com-mercial software applications already certified for Tomcat.
Workloads Matter
More than one quarter of respondents described their process loads as periodic (29 percent) or periodic spikes (29 percent), while 23 percent described them as asymmetric, and 19 percent reported a steady increase.
23%
Asymmetric Steady Increase Periodic Spikes
29%
19%
Periodicinstances so they can be processed more quickly.” Data integrity (100 percent) and security (97 percent) are the highest-rated technical factors in terms of im-portance when addressing elastic within the infrastruc-ture. Cost reduction (97 percent) was reported as the top business concern.
Moving Forward
Getting started in lightweight application develop-ment means following an evolutionary path that may be obstructed by previous investments and legacy applications. Step one involves virtualizing existing ap-plications and moving them into pools of private cloud infrastructure. Step two is implementing a lightweight development framework such as Spring as a starting point for developing applications in the virtual world. Step three is implementing a lightweight runtime like Tomcat or tc Server, which has proven, cost-effective performance. Once these three pieces are in place, moving applications into this more modern application architecture will be a straight-forward process.
Conclusions
The survey results make it clear that a virtualized lightweight runtime environment based on Spring and Tomcat provides the scalability, agility and provisioning speed required to satisfy evolving business require-ments and move application development systems onto private and public clouds. Tomcat has established its superiority at companies with less than $1 billion in revenues, and its flexible approach to Web-based, data-intensive environments has won acclaim from compa-nies of all sizes that are tired of operating in expensive, inefficient, environments.
To discover more, go to www.vmware.com.
Elastic Demands and Technical,
Business Concerns
An elastic application spikes in demand either periodi-cally or regularly. A good example is a retailer that has massive— but temporary—spikes on certain days throughout the year for events such as back-to-school sales or preholiday sales. Before the advent of lightweight application development environments, that retailer would buy the maximum number of serv-ers required to handle the biggest loads of the year, and then 90 percent of them would sit idle waiting for those few-and-far-between spikes to roll around again the following year.
People are now moving toward creating pools of virtual infrastructure—a private cloud—so that as spikes come and go, they can apportion servers to those applications on demand. Once the usage spike wanes, those servers can then be productively deployed to other applications inside their organizations. Technical concerns that accompany the elastic data storage and elastic application demands appear to be just as important as the business concerns. Tomcat users were found to be significantly more likely to rate data isolation as very important for their infrastruc-tures when addressing elastic data storage demands than non-users (61 percent vs. 41 percent). Not surprisingly, WebLogic users are twice as likely as non-Tomcat users to cite “minimal memory footprint” as an important technical factor when addressing elastic ap-plication demands (44 percent vs. 21 percent) as well as “quick provisioning” (73 percent vs. 37 percent). Data isolation enables people to partition data layers so that only the data that is relevant to a single set of application servers is available to those servers. This improves data access performance and is supe-rior to the traditional heavyweight method of having all application servers share a single data source, and then have to compete with each other in order to retrieve information from it.
Says McJannet, “Tomcat users are more accustomed to building lightweight, modular applications that are by their nature more partitioned. If you extend to the cloud deployment model, it becomes more important to have these smaller component pieces; if there is a spike in a processing request, you can call up other