New approaches to DC management are needed to keep pace with changing priorities
Data center management systems have evolved into a smorgasbord of different monitoring systems, sensors, device alerts, and dashboards. We have reached the point where even the systems being deployed to oversee data center operations have become difficult to manage.
This is not efficient for today’s facilities. It will not meet the demands about to be placed on operations and IT by the latest compute intensive workloads.
In our last blog we spoke about the road that brought us to the modern data center and how operational inefficiency should be addressed.
Looking ahead it is clear that as well as addressing existing challenges, a whole new set of demands on IT, facilities and data center physical infrastructure are to be considered. These will fundamentally change data center operations management.
New workload demands include but are not limited to: Artificial Intelligence (AI), Machine Learning (ML), IoT (Industrial and Consumer), 5G and Multi-Cloud.
“For example, AI and ML are already changing data center operations at Google.”
Each of these transformative technologies has the potential to shift the IT needle in terms of scale, latency, responsiveness, performance, security, and for operations, energy efficiency, resilience and cost.
AI and ML
In AI and ML models, the larger the data set the faster and better AI and ML can ‘learn.’ This training data must be ingested into the data center. Consider how much East West data center traffic will be generated by the average AI workload? A 50 terabyte training data set is considered usable. But how quickly could this scale to 100 terabytes or half a petabyte?
For example, AI and ML are already changing data center operations at Google.
Google revealed recently that the demands of its Tensor Processing Unit Artificial Intelligence chipsets had pushed it to using liquid cooling in a data center for the first time. This revealed at first-hand how new workloads could force a change in data center operations.
As more and more enterprises adopt AI and ML those outside the hyperscale cloud community with responsibility for facilities and IT infrastructure will have to address similar issues.
“How can the demands for these applications be met through high performance data center infrastructure? What are the data center management operations challenges?”
IoT, 5G, Multi Cloud
AI is not the only workload that will change how our data centers are run.
IoT data sets tend to be bi-directional or multi-directional. This creates new communications patterns within the data center and between data centers, from hyperscale to edge data centers.
5G will see huge growth in mobile traffic. Picture billions of smart phones downloading at 10Gbps and faster and uploading data at 1Gbps speeds.
Multi-Cloud is about being ‘bursty’. Serverless computing is making the promise of true utility computing into a reality.
Data rates from IoT and smart applications can be unpredictable. Unless correct data center capacity planning based on actual data is achieved, firms face the prospect of being constrained by their infrastructure and burst thresholds. At the same time they will be challenged to ensure capacity availability for critical applications.
Already the giant hyperscalers, telcos, banking and finance, pharma and life sciences firms are adapting their data center operations to cope with the capacity demands of these applications. Unless properly managed at a physical infrastructure level, ingesting hundreds of terabytes or half a petabyte onto a machine learning platform could lead to a squeeze which could impact existing workloads.
Surging demand towards a zettabyte era
How much corporate and consumer data will be generated by these technologies? How much data will enter and exit data centers daily within five years? How much content will be hosted and transported between regional, local and edge data centers?
“InCommand enables DC operators to plan, deliver and optimize for new workload conditions. These can include IT asset changes and configuration, including lifecycle management, capacity and asset planning.”
Global data projections point to a zettabyte era. The implications cannot be underestimated.
How can the demands for these applications be met through high performance data center infrastructure? What are the data center management operations challenges? It will require maximum utilization and efficiency at the infrastructure layer.
Big changes to the role of the data center operations
As digital disruption accelerates this becomes a fundamental business issue which is driving changes to how critical data center assets are commissioned, run, optimized and ultimately, monetized.
This means data center management fundamentals are changing and require a new approach which manages both the physical assets and IT.
So, what are the implications? It means all physical infrastructure management must become digital.
Effective management requires the features that are unique to InCommand™, the software enabled service, that brings together process, people and portal-based service for data center operational efficiency. From multiple sites right down to individual devices and components InCommand provides granular insight to exactly what is happening inside the data center. Each infrastructure item (site, room, rack, power supplier, switch, cable, device) can be monitored to provide detailed information from which the operator can quickly scrutinize the efficiency of a single component or the entire IT infrastructure.
InCommand is fully integrated across 3rd party BMSs, monitoring, rack PDU monitoring, virtual machine, and application management, all the way up to cloud hosting environments.
Any data intensive workload deployment is reliant on asset availability, meaning operators requirements for valid and accurate data collection across every physical asset is more vital than ever.
InCommand enables DC operators to plan, deliver and optimize for new workload conditions.These can include IT asset changes and configuration, including lifecycle management, capacity and asset planning.
InCommand’s features include:
- Environmental and power monitoring.
- IT asset change and configuration management.
- IT asset lifecycle management, including maintenance records.
- Cable, connectivity and port management.
- Predictive ‘what if’ scenario planning.
- Capacity forecasting (space, power, cooling).
Open APIs to third-party ITSM systems (BMC’s Remedy and ServiceNow) and virtual machine management software (VMware and Citrix), enables customers to map virtual workloads to physical machines to drive up server utilization. At a business use case level it creates the foundation for show-back/charge-back and to meet compliance requirements.
(Download the 451 Market Insight Incommand Report LINK)
The scale of the data sets associated with AI and ML are already creating fundamental changes in how our data centers will have to operate. When seen in the context of other transformative technologies such as IoT and 5G, the reliance on IT, facilities and physical data center infrastructure will continue to grow. How we manage these operations will be the measure of success for all data center professionals.