The EIC LocaCloud project contains three technical and scientific work packages which are:
- WP3: Hardware and edge datacenter development,
- WP4: Orchestrator, software grid management development,
- WP5: Software, developing Qarnot's services portfolio
WP3 concerns our digital boiler QBx which has been designed to accommodate 12 OCP servers. We need to redesign a new heat recovery cold plate when moving to new servers. Last year, we have worked on 3 server integrations.We started to refine the boiler to ease manufacturing and lower production costs. We have developed benches for Qbx and electronic boards for manufacturing.
Aim of WP4 is to scale up Qarnot's middleware to handle additional orders of magnitude regarding cloud grid size (number of edge DC sites, of computing cores, of clients). At the same time, the goal is to set the right conditions for scale up: data management capabilities, monitoring and supervision and especially IT security. We increased the number of cores per pool of machines from 500 to 700, still with a reactive system and want to reach 1500 cores per pool within a year. We have worked on high availability and horizontal scalability with a complete refactor of our central scheduler changing our communication protocol to HTTP. Caching and storage are also a concern as our infrastructure must be efficient whilst geo-distributed. We imagined and tested several configurations, most of which were unsatisfactory (performance, security, etc). We have started to develop our own registry service for users so that they can manage all their data on Qarnot’s infrastructure. In parallel, we are working on monitoring and quality management, developing telemetry from core services to detect issues in the system early before business impact. This is crucial for the management of our geo-distributed datacenter. In addition, we continue to develop accuracy and usability of our carbon reporting tool.
WP5 concerns our software portfolio to widen our cloud services to enlarge the accessible markets. Scale-up implies more sites and more clients, and consequently more needs for connectivity:
- Ad-hoc experiments for direct connectivity and redundancy (direct interconnection between our computing nodes to customer internal storage services, redundancy of an interconnection with a client to access private cloud service through the client IT infrastructure),
- shaping future development toward VPN-based interconnection for small and medium customers.
Qarnot’s cloud services portfolio requests transverse technological components on which we worked this year:
- IaaS: a general driver in our developments
- Streamlining hardware mobility between cloud and bare metal
- RoCE, accelerate node communication
- Carbon Facts V2, environmental footprint data
- Bare-metal server energy consumption
We also worked on end-users software integration:
- 13 new softwares, 18 updates of past integrations to decrease support and automate new version integrations,
- on-going development of a plug-in directly integrated into the modelling software.
