TY - GEN

T1 - Database Kernels

T2 - 14th Conference on Innovative Data Systems Research, CIDR 2024

AU - Lee, Sangjin

AU - Lerner, Alberto

AU - Bonnet, Philippe

AU - Cudré-Mauroux, Philippe

PY - 2024

Y1 - 2024

N2 - Flash memory is the de facto standard for data persistence in dataintensive systems. Despite its benefits, this type of memory has atleast one severe disadvantage: it is offered only as part of tightlyclosed Solid-State Drives (SSDs). To access an SSD, applicationsneed to resort to one of many possible I/O frameworks, whichthemselves are wrappers around a block interface abstraction, theNVMe standard. These levels of indirection impact how applicationsare structured and prevent them from benefiting from the full powerof Flash-based devices.In this paper, we argue that SSDs should instead interact with applications via CXL. CXL is a new technology driven by an Intel-ledconsortium that allows systems to maintain coherence between ahost’s memory and memory from attached peripherals. With CXL,a device can expose a range of Flash-backed addresses through theserver’s memory. One implementation option is to allow applications to read and write to that range and let the device convertthem to Flash operations. In our SSD, however, we pick a differentoption. The device exposes what we call a Database Kernel (DBK)through a CXL-backed memory range. Read/writes against a kernelwould trigger database-centric computations that the kernel wouldperform inside the device. We show examples of DBKs to supportdifferent database functionalities and discuss their benefits. Webelieve that CXL and Database Kernels can support a new generation of heterogeneous database platforms with unprecedentedefficiency, performance, and functionality.

AB - Flash memory is the de facto standard for data persistence in dataintensive systems. Despite its benefits, this type of memory has atleast one severe disadvantage: it is offered only as part of tightlyclosed Solid-State Drives (SSDs). To access an SSD, applicationsneed to resort to one of many possible I/O frameworks, whichthemselves are wrappers around a block interface abstraction, theNVMe standard. These levels of indirection impact how applicationsare structured and prevent them from benefiting from the full powerof Flash-based devices.In this paper, we argue that SSDs should instead interact with applications via CXL. CXL is a new technology driven by an Intel-ledconsortium that allows systems to maintain coherence between ahost’s memory and memory from attached peripherals. With CXL,a device can expose a range of Flash-backed addresses through theserver’s memory. One implementation option is to allow applications to read and write to that range and let the device convertthem to Flash operations. In our SSD, however, we pick a differentoption. The device exposes what we call a Database Kernel (DBK)through a CXL-backed memory range. Read/writes against a kernelwould trigger database-centric computations that the kernel wouldperform inside the device. We show examples of DBKs to supportdifferent database functionalities and discuss their benefits. Webelieve that CXL and Database Kernels can support a new generation of heterogeneous database platforms with unprecedentedefficiency, performance, and functionality.

M3 - Article in proceedings

BT - CIDR 2024. 14th Annual Conference on Innovative Data Systems Research (CIDR ’24)

PB - CIDR

Y2 - 14 January 2024 through 17 January 2024

ER -