18.04.2019 I Author: Sabine Narloch
How do you create a virtual data center that even inexperienced users can manage? Henrik Hasenkamp describes in the funkschau interview what role a sophisticated UI/UX design and extensive automation of operations through machine learning play here.
funkschau: Mr. Hasenkamp, a large proportion of development projects in companies are implemented using agile methods. What about gridscale?
Henrik Hasenkamp: That's no different with an infrastructure service than with any other software. The core is not the iterative approach, but the strong user centering. The requirements and needs of software users are decisive. And this is precisely why it is developed in small steps and with constant consideration of user feedback. Each time, it is necessary to determine anew which problems the users have to overcome, which tasks they want to perform and how they can do this best. However, some basic statements apply to every development project: the user interface is always the first thing users see. This is why they have a vital interest in understanding the user interface as quickly as possible and being able to use it intuitively. Developers shouldn't create an artificial obstacle between customers and their tasks - this only costs valuable time for everyone involved.
funkschau: How do you proceed? Which criteria are important?
Hasenkamp: A user interface convinces through simplicity and reduction. Simplicity includes the following criteria: The design of the user interface must be clear, clearly structured and graphically clean. The reduction is about the fact that the user only sees functions that he currently needs. Everything else should only appear when it is necessary for operation. The Gridscale interface looks like this: It is clearly designed and not overwhelmed by "Featuritis" (see box). In addition, there is an expert mode that supports experienced users with high requirements.
Many years of practical experience in the provision of cloud infrastructures have gone into the development of the interface and less successful examples from other providers have been taken into account. Because many of the known vendors have confusing and overloaded configuration pages that make them difficult to use. But the cloud can also be so simple that agencies, small and medium-sized companies, for example, can quickly configure the desired servers without having in-depth know-how. These can then be adapted to actual requirements at any time, for example with more processor cores and more storage space. Additional resources are directly available to users in the production system. The user interface is self-explanatory. Even users without a deep understanding of details can intuitively find their way around it and can put servers and storage into operation within a few minutes.
funkschau: IaaS is a virtual data center and data centers are known as complex. How do you prevent users from failing because of this complexity?
Hasenkamp: Users never have to deal with the typical complexity of a data center if they don't want to. This is only possible in this way if new approaches are also implemented in data center operation and the greatest possible efficiency is in the foreground - through automation and machine learning. Gridscale does this in two ways: by collecting data in the data center and analyzing actual user behavior.
Possibilities of data analysis
funkschau: What data is this and how do you collect it?
Hasenkamp: The basis is the development of dynamic capacity management. It ensures that from the customer's point of view the availability is 100 percent. For this purpose, a concept is used that has rarely been used in the IT industry so far: Predictive maintenance through self-learning algorithms, known as predictive maintenance in industry 4.0. Sensors record important operating data. Software for machine learning examines these data for conspicuous patterns. In a data center, these could be temperatures inside and outside the housings, voltage levels or latency times. Deviations from normal values often announce malfunctions and failures. After some time of training, the algorithm can distinguish between normal and abnormal states. This allows for a high degree of automation of IT operations in the interest of the customer.
funkschau: What possibilities do the data analyses offer you?
Hasenkamp: This enables automation functions to be developed for the customer. The actions of the users in their dashboards permanently generate data. They can help to better understand the customer and adapt the platform in such a way that it is optimally tailored to solving the user's particular problem. Developers can draw conclusions from this and, as they revise the appropriate UI elements, they can also modify the user interface.
improve your experience.
funkschau: Can you outline a specific application?
Hasenkamp: An example: After evaluating the data, it is clear that some functions are frequently used one after the other or at specific times. So it could be that many users take regular snapshots, some perhaps daily, others weekly. Developers can now try to make this behavior easier by simplifying the user interface, for example by adding an automation feature. The practical result is a planning function that regularly saves snapshots as backups.
The statistical analysis of real-time user data can also be used to customize the user interface to the active customer. For example, each customer sees different selection and detail options according to their technical experience. In addition, the algorithms of the user interface make a kind of assessment of the user actions. They make different suggestions, depending on what the user is likely to want to achieve. In this way, users only receive the options that are relevant to them.
In this way, the evaluation of data is used to optimize the user experience (UX) and dynamic provisioning with individually customized visual interfaces is possible. Thus, customers from an SME can easily set up their own virtual data center.
The original article in German can be found here.