The most popular and comprehensive Open Source ECM platform
Cloud computing is supposed to make things cheaper and simpler. In many cases that’s true. But bigger businesses are finding that as they move more operations and services into the cloud, there begin to be difficult problems in globally managing both the complexity and the inter-dependencies of the components of cloud infrastructure and the services that run in it.
A survey by Ovum last year found that 41.5 percent of IT managers think that increasing the number of applications and services running in public clouds results in a simultaneous increase in the complexity. Part of the complexity has to do with project management of the many applications. But beyond project management issues there are additional complexities related to performance. Applications in the cloud add a new twist to the problem of troubleshooting performance — with applications no longer under your total control, monitoring server and network performance factors becomes more difficult and getting a true detailed picture of the entire infrastructure is nearly impossible.
Interdependent services running in the cloud can be risky. Bryan Ford, Yale University researcher and assistant computer science professor, wrote in a paper for USENIX HotCloud ’12 conference, that “As diverse, independently developed cloud services share ever more fluidly and aggressively multiplexed hardware resource pools, unpredictable interactions between load-balancing and other reactive mechanisms could lead to dynamic instabilities or ‘meltdowns’.” Ford argues that because cloud infrastructure is so complex and because interdependencies can be created when combining cloud services and applications, it is possible for systems to fail in totally unexpected ways. Ford says that complex systems like cloud services are a little like the financial system — generally things work smoothly, but there can be occasional glitches where the results can be catastrophic.
Ford warns that “While cloud computing “is promising and attractive in many ways … at the very least, it would be prudent for us to study some of these risks before our socioeconomic system becomes completely and irreversibly dependent on a computing model whose foundations may still be incompletely understood.”