Missed part 1 of this article series? Read it here and learn everything about maximum performance and consistent performance.
Given the complexity and multifaceted nature of this topic I will try to point out some “generic” arguments and particularities that are typically true but have to be validated in detail in the respective context rather than attempting an exhaustive discussion of ‘Availability’ in a single article. For the sake of simplicity, in this section I will cover technical aspects regarding availability as well as the contractual counterpart: SLAs, agreements that reflect the guaranteed or rather promised availability.
Five key arguments on availability include:
- Sophisticated virtualization and resource management technologies in multi-tenant environments are complex technological layers. As every additional level of complexity is added, this effectively decreases the availability. Bare metal cloud offerings can be considered more robust in this regard as they omit technological layers compared to virtualized clouds. (Please note: that’s the theory as mentioned in the introduction)
- Unmanaged cloud providers typically offer low SLAs that cannot be negotiated. If you need somebody to take responsibility and compensate your losses in the case of a severe outage, the big players like AWS and Azure might not be your first choice.
- Managed non-virtualized cloud offerings typically offer a broad range of system configurations and corresponding SLAs. Many aspects can be negotiated individually according to your availability requirements. (That’s also true for EXACloud, EXASOL’s managed bare metal hosting offering with premium services)
- Furthermore, managed non-virtualized cloud offerings mitigate outages proactively as they monitor your application more thoroughly and they know how to handle and manage your application. This might increase the overall availability significantly when compared to unmanaged services.
- On-premise solutions can be configured and operated in a tailor-made manner to meet your availability requirements.
Based on these arguments, it is clear that managed hosting and on-premise are rated best whereas the unmanaged cloud offerings receive the lowest rating.
Configuration & deployment agility (including lead time)
Typical non-virtualized clouds like AWS or MS Azure offer full on-demand one-click installations. Create your account, register your credit card and your web server, database or cloud-based CRM systems can be up and running in just minutes.
Bare metal clouds are typically less agile as they have to manage physical machines instead of abstract virtual instances. Of course there are bare metal clouds that also offer a high degree of automation, but in general it can be assumed that virtualized clouds like AWS or MS Azure are much more agile and faster.
If you decide to go for a managed bare metal hosting/cloud service, you may lose again on agility given that you have to liaise with humans. You probably won’t find a button labelled “operate the database according to my requirements and my processes. (powered by artificial intelligence)” ;-).
You will have to discuss processes, schedules, perhaps skill levels of the employees who manage your application and all this has to be documented and agreed to in some kind of contractual arrangement. That will increase the lead time and is already different from the click-and-go-paradigm of AWS and others. However, it is still faster than clarifying and setting up all the stuff in your own data center.
Of course there are also managed offerings that come with fixed service specifications, that require you to adjust your IT to the service. You win on agility but the price you pay is that you adjust your processes to the ones of the service provider, not the other way around.
Agility is the real sweet spot of unmanaged cloud offerings like AWS and MS Azure.
INTERMEDIATE SUMMARY PART 1 and 2
So let’s wrap up the first two parts of our article:
Not surprisingly, flexibility is where fully virtualized cloud solutions really shine. When it comes to the flexibility factor, fully virtualized Cloud beats the other hosting options hands-down.
Of course, that also means that if you don’t need a high level of flexibility, there might be more significant reasons to rely on bare metal, either managed/hosted, or on on-premise solutions.
And while performance, availability and agility are key criteria for choosing a bare-metal hosting option, another major determinant are the costs associated with a solution.
In the next part of this mini-series we look at how the different categories of cloud offerings stack up against each other when considering direct and indirect costs.