The I/O stack in modern servers is emerging as an important bottleneck. On one hand we can increase both processing capacity and device throughput in modern servers, especially with the projected improvements in multicore systems and storage device technologies. On the other hand the I/O path of modern systems is not able to scale.

IOLanes examines related trends and designs mechanisms that will allow the I/O stack to scale as we go forward.

• IOLanes enables partitioned I/O to scale on dense multicores and to eliminate interference across workloads and VMs without compromising a single-system image.
• IOLanes embraces multicores for on- and off-loading purposes in the I/O path.
• IOLanes reduces host-guest virtualization overheads in the issue and completion paths.
• IOLanes offers a single-point for I/O performance monitoring and visualization.
• IOLanes targets OLTP, OLAP, and Data streaming applications.