Accelio xio integration with kernel RBD client for RDMA support¶

Summary¶

Add accelio xio transport to kernel libceph messenger and integrate the new messenger with kernel RBD.
Accelio xio is a transport-independent messaging abstraction and currently runs over RDMA (infiniband or ethernet)

Owners¶

• Raju Kurunkad (SanDisk Inc.)

Interested Parties¶

• Raju Kurunkad (SanDisk Inc.)
• Vijayendra Shamanna (SanDisk Inc.)

Current Status¶

Currently ceph kernel RBD client uses the kernel libceph messenger. This messenger uses kernel sockets using TCP as transport to communicate with a ceph cluster.
Ceph support for using xio to communicate over RDMA is being tracked by blueprint. We are adding support in kernel RBD to communicate with ceph clusters that support xio.
Development of prototype libceph messenger to use xio for talking to a xio enabled ceph cluster is in progress.

Detailed Description¶

Add following changes to linux kernel libceph module:

• Interface with the xio kernel module for accelio session and connection creation(inprogress)
• Seperate libceph messenger and the actual transport(TCP, xio)
• Refactor existing libcepf kernel messenger to function in request-response fashion(TBD)
• RBD config and Interface to request use of accelio xio transport for communication with xio enabled ceph cluster(TBD)
• Measure performance with xio enabled libceph/kernel RBD talking to xio enabled ceph cluster(TBD)
• Verify all supported kernel RBD operations using xio transport in kernel libceph messenger(TBD)
• Verify basic kernel cephfs operations using xio transport in kernel libceph messenger(TBD)

Work items¶

Coding tasks¶

1. Task 1
2. Task 2
3. Task 3

Build / release tasks¶

1. Task 1
2. Task 2
3. Task 3

Documentation tasks¶

1. Task 1
2. Task 2
3. Task 3

Deprecation tasks¶

1. Task 1
2. Task 2
3. Task 3