I recently posed a question to eBizQ and got some good answers, both reasonable and unexpected perspectives. See my eBizQ SOA forum question here: Are Servers Bloating in Pre-Production, but Shrinking in Production due to SOA?
K. Scott Morrison and Joe McKendrick discussed how Virtualization is a tester's best friend in reducing this phenomenon - and yes we have seen virtualization of hardware reducing some infrastructure needs. What about the ones you can't virtualize? I vote the best answer (received by 5:00PM CST) goes to Avi Rosenthal, whose comments highlight why this phenomenon is happening:
"In distributed Service Based, Virtualized and Cloud based environments the number of other environment servers is skyrocketing vs. the number of Production servers. This is true as well for virtualized server instances. The reason for that phenomenon is that Production environments are more controlled because no one can allow uncontrolled Production. The risks of non controlled Development or Test environments are less obvious. As the number of instances grows the number of servers in the non-controlled environments grows more rapidly than the number of servers included in the controlled environment. However, logically the number could be reduced at least to a ratio of 1:3 or 1:4 and not 1:20 if the non-Production environments will be managed and virtualized similar to the Production environment."
Really good point Avi.
However HP's Kelly Emo then came up with the gist of an issue I wanted to address further: the huge rift between measuring costs and effort for Production vs. Pre-production environments. It seems these activities are usually owned by two completely different departments and cost centers within the enterprise as she notes:
"In a majority of customers I speak with, their organization, culture and funding models are no where near having this happen... different silos are managing hardware and software resources and different SLAs are associated with them. In fact, in some cases, SOX compliance keeps organizations from being able to mix and mingle their networks and data between pre-production and operations. That said, I see one place where this can happen.. in the cloud..."
That's a next-generation view. Indeed the cloud is a big piece of the puzzle. However, many of the companies who get excited about the value of deployment savings using VMs and Cloud-based infrastructure to replace servers, simply do not have a reason to care about the number of groups maintaining and consuming dev, test and integration environments.
The same goes vice-versa for pre-production - when you break up development into smaller, agile dev & test teams, those teams each provision and set up their own hardware and software environments in a largely unregulated fashion - and often find out that there are many dependencies they can't copy or replicate via conventional means. This leaves a lot of cost and effort unaccounted for in the wash of the software lifecycle, or as a cost of per-employee infrastructure or support.
You will see a lot more on this topic from us - while we aren't in the business of virtualizing servers and hardware, we are particularly curious about why companies aren't recognizing the value they get by optimizing Production vs. optimizing their infrastructure at every step along pre-production. Is a ballooning number of servers a side-effect of distributed, agile approaches, or is the symptom we are reporting simply a matter of perception? We'll keep you posted as we do more research in this area.
Comments