I believe that 2013 will be the year that hybrid Cloud application deployments go mainstream.
This prediction is based on EMA IT-facing Integration research conducted in mid-2012, which found that:
• 45% of companies surveyed had already deployed at least one “ hybrid Cloud” application, defined as “applications integrating to at least one public Cloud service”
• 26% planned to deploy such services in the “near term”, defined as 12 months out.
If these estimates prove to be correct, more than 70% of companies will have deployed hybrid applications by July, 2013.
I know it may appear that I’ve been obsessed with hybrid Cloud applications for the past year, but there is a method to my madness. In an industry which is already struggling to effectively monitor and manage application performance and availability, hybrid Cloud applications are a minefield of management challenges.
For years, we’ve been employing buzzwords such as “complexity” and “heterogeneity” to describe the challenges of managing enterprise applications. However, as such applications increasingly span both internal and external execution systems, new descriptors such as “opacity” and “lack of manageability” enter the fray as well.
With a wide variety of high-quality data integration products on the market, deployment of hybrid applications isn’t the issue. The issue is managing performance and availability, since the majority of application management tools on the market are geared towards on-premise-hosted applications. They rely on the fact that customers have access to hosted applications and their supporting infrastructure, enabling IT professionals to instrument and monitor them at will. This is not the case with Cloud-delivered applications, and this fact makes deployment of hybrid Cloud applications a risky business.
Enterprise Application Management
Relatively few companies — only 5% – can definitively pinpoint the source(s) of every application-related problem. Almost 30% report that between 25 and 50% of such problems are resolved by trial and error. Twelve percent (12%) report that between 51 and 75% are resolved that way, while 4% resolve “more than 75%” by trial and error.
At the same time, more than 50% of companies report the cost of an hour of downtime for the “most critical business applications” to be between $75,000 and $500,000. Needless to say, the potential for downtime and/or performance problems, combined with the cost of such problems per hour, generates significant risk.
And while these numbers may not be shocking in and of themselves, here’s the kicker: THESE PERCENTAGES RELATE TO ON-PREMISE-HOSTED APPLICATIONS. Once all or part of a transaction exits the traditional data center environment, all bets are off. Enterprise IT loses both visibility to and control of the Cloud-hosted elements.
Application triage/root cause analysis requires, at minimum, visibility to the components supporting the application, which is typically achieved via instrumentation and/or logging. However, SaaS and PaaS services, in particular (IaaS is in a class of its own here), cannot be instrumented by the customer because they are owned by service providers.
This combination of factors presents a particularly pressing end-to-end management challenge when transactions span multiple public Cloud “hops”. Without adequate instrumentation, it is virtually impossible to determine which “hop” is impacting performance and availability. As yet, relatively few SaaS/PaaS vendors provide the instrumentation and metrics necessary to make such determinations.
Three Keys to Hybrid Application Management
This problem isn’t that difficult to solve, but it does require action by both SaaS/PaaS and enterprise management vendors.
The three key elements supporting triage of hybrid Cloud applications include instrumentation, integration, and analytics.
• 1. Instrumentation: With the rapid development of public and private Cloud over the past two years, manageability has become a key requirement in RFPs and product evaluations. IT consumers understand that, although they may not control service delivery, they still need visibility to availability, performance, and service levels. They still get user complaints when the service is down—witness the multiple Amazon outages of the past year, which have impacted almost every industry—and need a way to determine “who to call”.
While some public Cloud providers—most notably IaaS providers such as Amazon and RightScale—have been proactive in terms of instrumenting their platforms (and providing APIs to their consumers), others have not. For hybrid Cloud applications to become production-ready, SaaS and PaaS vendors must instrument their platforms and make management APIs available to enterprise management ISVs for integration into application management solutions.
I see the lack of such instrumentation as the primary factor inhibiting monitoring and management of hybrid Cloud services. And this gap is particularly noteworthy since many of the hybrid Cloud deployments I have encountered span multiple SaaS providers. Integrations between CRM and marketing applications are commonplace, for example, and several IT organizations have told me they are running literally hundreds of reads and updates against integrated SaaS services.
Instrumentation could take the form of log files, audit logs, SNMP MIBs, or other metrics, but should, at minimum, track the entry and exit of a transaction. Even basic metrics provided to application management platforms could provide a basis for tracking performance and availability, and it is the responsibility of IaaS, SaaS, and PaaS vendors to provide such metrics.
• 2. Integration: APIs are useless unless the information they share is accessible, and Integration solutions can bridge the gap between management metrics supplied by public Cloud providers and traditional (or Cloud-based) application management systems.
There are so many great integration solutions available today that this is not a matter of “reinventing the wheel”. Click here to check out my research on the topic. In fact, many of the leading integration vendors already have connectors for public Cloud platforms, and some have connectors to enterprise management products such as IBM Tivoli as well.
While some companies are still trying to integrate via homegrown programs, virtually 100% of the companies that have migrated to commercial products tell me they have saved money by doing so. As software changes, so do APIs, and companies such as Informatica, IBM (Cast Iron), Dell (Boomi), Pervasive, and Adeptia work closely with Cloud vendors to ensure integrations are always up to date. In-house integration skills are expensive, and program maintenance in the dynamic Cloud environment can be cost-prohibitive.
Bottom line: the Integration tools are there, and many are “Cloud-ready”.
• 3. Analytics: My colleague Dennis Drogseth recently published an exhaustive (and probably exhausting!) study of Advanced Performance Analytics. Enterprise management vendors in the application and network management spaces have developed very sophisticated, analytics-intensive solutions leveraging industry expertise, heuristics, algorithms, self-learning, and other capabilities to distinguish between “normal” and “abnormal” application behavior.
With access to API-delivered metrics and/or execution history such as log files, many analytics solutions could adequately support hybrid Cloud applications with little or no modification. I believe, however, that Cloud support has become a critical capability for modern application management solutions, and the research bears this out. Almost 70% of the APM-savvy companies indicate that “Cloud-readiness” is a consideration in new product purchases.
In summary, hybrid Cloud application management is within reach. It does require, however, concrete action from both public Cloud vendors and enterprise management product vendors. Both need to be aware that it’s no longer enough to simply deliver a product. For public Cloud services to cross the chasm from Proof of Concept to Production, It is also incumbent for SaaS/PaaS vendors to build manageability into their solutions, and for application management vendors to support SaaS/PaaS instrumentation in their products.
Julie Craig is Research Director for Application Management at Enterprise Management Associates (EMA).