The short answer is really, “I don’t care.” Nor should I. My expectations are that any appliance I own should simply work effectively to do the job for which it was intended. I don’t know the details of how my microwave makes my Salisbury steak TV dinner; I just know it does. We’ve grown very comfortable with the appliance model in virtually all aspects of our life, yet the very same concept causes a great deal of consternation when applied to the datacenter. This unease has been highlighted recently as we march towards the general availability of the Dell Cloud for Microsoft Azure Stack. Which, as you may be aware, is an appliance.
Based on numerous customer conversations as well as internet discussions it is evident we need to have a broader discussion on the appliance model, and why it’s important; particularly where Azure Stack is concerned. Jeffrey Snover, Microsoft’s lead architect for Azure Stack, had a wonderful series of tweets on the subject (http://bit.ly/2srQGJc) in which he went so far as recommending that Azure Stack partners stop showing architecture diagrams because it confuses people. I’m not convinced it needs to go that far as, after all, “inquiring minds want to know.” However, I do think that we need to be very clear on the messaging as to why Microsoft and its development partners have gone the route of an appliance model for Azure Stack.
Let’s start by setting the table with the concept that successful cloud computing in general is based on rote standardization. Individual ‘snowflakes’ in large cloud environments (or even moderately sized ones) lead to management, sustainability, and support nightmares. In 2015 a Gartner analyst posited that some 95% of private cloud implementations had run into challenges to success (http://gtnr.it/1LTLIJc). His blog posited that several of the largest pain points were insufficient changes to the operational model, doing too little (e.g. – this isn’t cloud), and doing too much customization (I&O). The reality is that these points are tightly coupled. The operational model must move from one of IT management to one of cloud service delivery management; which in turn will involve offering a set of well-defined services for consumption rather than a la carte servers, network, storage, vms, etc. To go back to my analogy, rather than offering a set of on-demand diodes, timers, and thing-a-bobs the process and focus is on offering the ability to quickly cook dinner.
If this holds true for private cloud implementations the stakes are even higher for a hybrid cloud implementation like Azure Stack. Not only are we concerned with offering on-prem Cloud services but must also deal with compatibility to Azure Public. This is why the appliance model becomes extremely significant.
Over the past few years, Dell and others in the industry have been driving towards the software defined datacenter. The realization here is that software can take on take much of, if not all, the heavy lifting related to infrastructure configuration and management. Not only can it do so, but it can do so more efficiently and effectively; particularly when dealing with a system reliant to a large degree on standardization. Moving forward with the Dell Cloud for Microsoft Azure Stack we take that concept to the next level.
Azure Stack is offered as an appliance. By that we mean it is an integrated and closed system. All the underlying technologies to deliver the infrastructure services: networking, compute, storage, etc. are handled by the software. This means the deployment, configuration, management, updating, and all the trappings of traditional IT management are handled by software; the IT admin role shifts to that of cloud admin. When you look at the physical components it looks like a rack of servers with top of rack switching, but logically it’s a closed system delivering a slice of Microsoft Azure out of your datacenter. It is managed in a like fashion to Public Azure, you don’t have access to the underlying infrastructure configurations it’s all about the services being delivered. Do I care what individual components cook up my Salisbury steak? Nope, just that it’s cooked and cooked correctly in 3 minutes or less.
I’ll take a moment here to provide an example that has come up in several conversations of late: top of rack switching. We’ve had inquiries about if the switches can be changed to meet individual organizational “corporate standards” for networking – the answer is “no.” The top of rack switching contains some 3000 odd configuration settings being managed by the Azure Stack software, the traditional network administrator does not have access to these settings. Misconfiguration of these settings would pose significant risk to the services being provided by the system, so for the sake of sustained service levels and standardization they are locked down. The Network team is now responsible only for the connection to the backbone and beyond.
We need to start looking beyond componentry and view the appliance as greater than the sum of its parts. Rather than what is your corporate standard for a switch or a server, you should be asking what is your corporate standard for an Azure Stack appliance? By treating it in this way we can start addressing the causes for failed implementation I highlighted earlier. By leveraging an appliance model, operations, by default, must change from focusing on core IT componentry to delivery of cloud services and service delivery management. The cloud platform itself is being defined in alignment with Azure’s public offering which significantly mitigates the “do too little or too much” scenarios; there is no question that this is cloud and there are no snowflakes. This model allows the IT organization to spend more resources focusing on high value activities rather than keeping-the-lights-on type functions that are better handled by the software itself.
Many, if not most, organizations looking at Azure Stack have already accepted the cloud computing model and likely are leveraging Azure today for at least some portion of their IT services. The public cloud model is based entirely on consumption of services. The Dell Cloud for Microsoft Azure Stack provides an appliance-based approach to offering those services in a hybrid, on-prem fashion with the focus remaining on consumption of services. Managing an endeavor like this in a “build your own” fashion would be fraught with risk and likely have a high rate of failure and dissatisfaction. The appliance model allows organizations to maintain a consistent level of hybrid delivery of Azure services with a consistent management process across public and on-prem infrastructure, relegating core management functions where they belong in cloud, behind the scenes. Let’s focus on eating our Salisbury steak rather than how it gets cooked.
