Organizing the AI in the Data Center.

As many of you know, the company I belong to, Bjumper, which is my home and where I was born, is currently undergoing a transformation process. 

During the week of January 8th, we also had some amazing sessions at our central offices in Madrid, sessions of reflection, not about who we are and how to be better, but about who we want to be!! 

Many ideas came out of it, but above all, a clear vision emerged, we will build the Data Center Autopilot omething that many of us have had in mind for years, from automatic asset management scanners, real-time climate control systems, automatic load balancing based on predictive systems, and points we haven't even identified yet. 

Years ago, bringing all these ideas to the market was possible (we had the technology and the knowledge), but not at a cost that the market could afford; they were ultimately just dreams.. 

However, something has changed today, and it's not AI (which has been around for a long time, albeit perhaps less noisy); it's that advanced AI-based technology is now available to everyone, including Bjumper.Today, it's easier and cheaper to access it. It allows us to improve products quantitatively at a cost that is now applicable to Data Centers. 

While I don't know much about the development world, I'm learning a new language ☺! There's something about this sector that fascinates me, and it's the ability to share knowledge. When a large community works in unison, the results are what we're all seeing around us today: exponential growth. Which leads me to ask, why not create a much more collaborative Data Center community? (Another small dream, but for another article).. 

So, to sum up, thanks to this so-called "collaborative" community, today we have access to free LLMs in the market so that others can develop applications more quickly, cheaply, and effectively, and, most importantly, with an interface designed to understand and generate human language. I believe that's where the difference lies. 

Does that mean we are creating an Alexa for the Data Center? No, or at least not the Alexa that we know today.

Today, it seems like standing out will be difficult; the market is flooded with AI (at least on paper). But, what does AI bring to the Data Center today? I delved deeper into this, and here are some of the initiatives I found.  

** La Desagregación de Memoria en Centros de Datos**: un estudio de Ewais y Chow (2023) revisa propuestas recientes sobre desagregación de memoria, destacando cómo esta puede abordar el problema de la subutilización de recursos al desacoplar los elementos computacionales de los recursos de memoria, si bien esto para el rendimiento de las aplicaciones en las máquinas, es muy interesante, en el día a día de quienes operan los Data Centers no es que les afecte, más allá de que esta es una línea de optimización en espacio ,o no 😔, de los Data Centers ( máquinas más eficientes, necesidad de menos máquinas para hacer lo mismo!). 

**Storage and Data Management**: Moreno et al. (2021) propose a theoretical model for a storage management system for virtual learning environments based on big data, using AI technologies to optimize the processing of stored data. This line (truly interesting because the data storage growth model is not sustainable) has a real effect on Datacenter management similar to the previous one.. 

**Unified Storage Securit**: Hussain et al. (2020) address vulnerabilities, threats, and attacks in centralized storage systems, proposing a framework to mitigate these problems and improve security based on AI. This is super important, and I'm sure these lines of research will have a long way to go. This is truly helpful for those of us involved in Data Center infrastructure management to minimize risks.. 


However, although these lines (they are just an example of what I have found) are very interesting, I haven't been able to find specific cases or studies focusing on improving the management of critical physical infrastructure of DCs using AI..


The closest have been several cases related to consumption prediction and early fault identification, which, if you allow me, we are talking more about machine learning than AI.. 

They may try to sell us on how AI will help us optimize energy consumption and become more efficient, but on the contrary, these applications will be hosted in Data Centers where power requirements and cooling needs also continually increase. So, I wonder, will AI help the Data Center or give us more headaches? Because what is undeniable is that it will be very present 

Unfortunately, I can't answer that, but where I do see a fundamental role for AI in Data Centers is in automating the management of physical infrastructure, in turning tedious tasks into automatic ones that free up time and mental space for the people working in the Data Center to focus on tasks richer in planning and analysis (we're going to need it!).) 

Entonces ¿Qué rol puede tener la IA en la operación del día a día del DC? 

We can't expect to have the DC rabbit right away (if you haven't seen it, take a look here https://www.rabbit.tech/keynote it's definitely worth it), but we can ask for it now and only to start: 

  •  Virtual assistants to automate responses to frequent queries
  •  Managing technical support tickets   and guiding staff through diagnostic and troubleshooting procedures. Additionally, this point is essential for addressing the lack of qualified personnel that exists in the market..
  •  Automated planning and execution  of IT infrastructure and Networking migration projects with predictive algorithms to plan migration, identifying the best time to make changes with minimal impact on operations..
  • Evaluating workload by teams and execution times to optimize project implementation times.
  • Automating basic access permission tasks  to Data Centers based on planned projects..
  •  Documentation! Sometimes we seem more like administrators than technologists. For audits, reports, etc. Being able to automate these tasks and make them much richer in content would give time and visibility... 

Really, we just need to do one exercise, and that is what I am doing today and what I should do to meet my company's business objectives or service? In many cases (it happens to me every day), what I'm doing is not adding value and is taking me away from the company's objectives.. 

If we identify these points, we have an improvement space in the Operation where AI can do a lot for us! 

At Bjumper we already have some identified and in the process of automation! Step by step, we will create a different way of Operating the Data Center, better, safer, and above all, designed for people. We are not machines, and that's precisely why what we can bring to our companies is much more than they could ever do. Let's use them to allow us to be more human!                                                               

 With affection,                                                                                                                                         Pilar Alcázar

                                                                                                                                                                                            Let It Work for You!

 

 


Share post LinkedIn