Mission & Motivation

The current cloud landscape is getting populated with many applications that are being migrated to the cloud due to its convenience and ease of use. However, there are still a subset of applications that are not frequent in the cloud. These are data-intensive and time critical applications. Data intensive applications, in the best case, experience bad performance in the cloud as current cloud database infrastructure fails to satisfy high loads and does not provide predictable performance. The latter is especially important for time critical applications, since they need to satisfy strict SLAs (Service Level Agreements) and without predictable performance it is impossible to satisfy them. Many critical applications run today on mainframes due to their high resilience and high performance. Unfortunately, there is no equivalent in the cloud till today.

Mission

CloudDBAppliance aims to create a cloud database appliance that can reproduce the performance and resilience of mainframes in cloud data centres. The goal is not only to match their performance and resilience but also to go beyond their current capabilities. Today mainframes only support operational workloads. For analytical workloads and business analytics a different kind of system is used such as data warehouse and Hadoop data lakes. However, this current approach is expensive and complex since it requires copying the data from the operational database into the data warehouse. CloudDBAppliance aims at producing an appliance with both capabilities, but also with the ability to perform analytics over the operational data.

CloudDBAppliance joins the forces of the main European hardware maker, Bull, with two promising database startups, ActiveViam that provides a fast analytics platform, ActivePivot, and LeanXcale that provides a scale-out database. The purpose is to build the next generation of hardware from Bull that will be able to provide a computer with 800 cores and 140 TB of memory with a re-engineered ActivePivot and LeanXcale platforms to exploit this extreme hardware. ActivePivot will be enhanced to be able to scale-up linearly with the number of cores, a difficult challenge when the number of cores is an order of magnitude of what regular servers provide today. The other challenge is how to be able to use 140 TB of memory efficiently in a many-core architecture with a hierarchical structure. LeanXcale will be more severely re-architected. Today, it is based on a distributed data store, HBase and HDFS, which are not adequate for this kind of hardware. Additionally, it is designed for data sets that do not fit in memory. Moreover, LeanXcale is designed to scale-out (horizontally), but not to scale-up (vertically), especially to extreme numbers such as the new hardware platform targeted in the project. For LeanXcale a brand new data manager will be built that will substitute HBase and will be designed to be able to scale-up efficiently for a large number of cores. Special care has to be taken to the fact that analytics will run directly on the operational data, that is, without making a copy of the database.

Motivation

CloudDBAppliance has been motivated from an actual business need, and therefore, it aims at exploiting the results via partnerships of the three industrial technology providers. The resulting appliance will be validated in three different sectors with data intensive applications and time critical requirements: banking, telco and retail.