Don’t Double Down on Infrastructure – Scale Out as Needed

  • There has long been a philosophy in IT infrastructure that whenever you add capacity, you add plenty of room to grow into. This idea is based on traditional architecture that was complex, consisting of many disparate systems held together by the rigorous management of the administrators. The process of scaling out capacity has been a treacherous one that takes weeks or months of stress-filled nights and weekends. These projects are so undesirable that administrators and anyone else involved would rather spend more than they like to buy more capacity than they need to avoid scaling out again as long as possible.

    There are a number of reasons why IT departments may need to scale out. Hopefully it is because of growth of the business which usually coincides with increased budgets. It could be that business needs have shifted to require more IT services, demanding more data, more computing, and thus more capacity. It could be that the current infrastructure was under-provisioned in the first place and creating more problems than solutions. Whatever the case, sooner or later, everyone needs to scale out.

    The traditional planning process for scaling out involves first looking at where the capacity is bottlenecking. It could be storage, CPU, RAM, networking, and any level of caching or bussing in between. More than likely it is not just one of these but several which causes many organizations to simply hit the reset button and just replace everything, if they can afford it, that is. Then they implement the new infrastructure to just have to go through the same process a few years down the line. Very costly. Very inefficient.

    Without replacing the whole infrastructure, administrators must look to the various pieces of their infrastructure that might need refreshed or upgraded. This process can seem like navigating a minefield of unforeseen consequences. Maybe you want to swap out disks in the SAN for faster, larger disks. Can the storage controllers handle the increased speed and capacity? What about the network? Can it handle the increased I/O from faster and deeper storage? Can the CPUs handle it? Good administrators can identify at least some of these dependencies during planning but it can often take a team of experts to fully understand the complexities and then sometimes only through testing and some trial and error.

    Exhausted yet? Fortunately, this process of scaling out has been dramatically simplified with hyperconverged infrastructure. With a clustered, appliance-based architecture, capacity can be added very quickly. For example, with HC3 from Scale Computing, a new appliance can be added to a cluster within minutes, with resources then immediately available, adding RAM, CPU, and storage capacity to the infrastructure.

    HC3 even lets you mix and match different appliances in the cluster so that you can add just the capacity you need. Adding the new appliance to the cluster (where it is then called a “node”, of course) is as simple as racking and cabling it and then assigning it with network settings and pointing it at the cluster. The capacity is automatically absorbed into the cluster and the storage added seamlessly to the overall storage pool.

    This all means that with hyperconverged infrastructure, you do not need to buy capacity for the future right now. You can get just what you need now (with a little cushion of course), and scale out simply and quickly when you need to in the future. The traditional complexity of infrastructure architecture is now the real bottleneck of capacity scale out. Hyperconverged Infrastructure is the solution.

    Original Location:

  • I just mentioned this concept in a post elsewhere earlier today, in fact. This is an extension of the YAGNI Principle that is commonly known in software engineering circles but is often overlooked in IT ones.