Concept

CloudDBAppliance aims at covering the need for more reliable infrastructure in the cloud for time critical and data intensive applications, providing a cloud database appliance with the same reliability and performance than traditional mainframes. Current cloud infrastructure fails to provide the predictable performance and resilience of mainframes and that is why time critical data intensive applications are not migrated to the cloud and remain running on mainframes. Mainframes provide predictable performance and a scale-up operational database. The goal of CloudDBAppliance is to provide a cloud alternative for applications that today can only run on mainframes.

CloudDBAppliance aims at delivering a cloud database appliance that matches the resilience and performance of the mainframe, and goes beyond current mainframes by providing real-time big data, while the mainframe has to rely on external data warehouses and ETL processes to deal with analytical queries and business analytics.

Hadoop data lakes are also becoming a standard in industry. However, today they only process analytics over historical data due to their lack of operational capabilities. CloudDBAppliance will bring a novel operation data lake that will run an operational database on the data lake and perform analytics with the data lake technology over the operational data.

The main challenges to be addressed to materialize this vision are:

  • How to architect the data managers (operational database, fast analytics, operational data lake and streaming analytics) to scale up efficiently to over 800 cores and beyond with 140 TB of memory or more
  • How to architect a hardware able to scale to over 800 cores and 140 TB of memory without internal bottlenecks and able to deliver maximum performance
  • How to be able to be scalable for both operational and analytical workloads on a single appliance
  • How to provide a resilience comparable to mainframes with the combined software plus hardware
  • How to reallocate resources to enable an efficient use of the hardware
  • How to reallocate resources across data centres