Hi, @Paul @Jim
About the topic “[WIP]Add py-spdk client for SPDK”, I am sorry for that I didn't
mention it on the mailling list in advance. Indeed，this is unrelated to my proposal on
Trello about the blobkv proposal.
Here is my idea of py-spdk:
The py-spdk maybe we can regard it as client for SPDK, which provides a python binding
for the native SPDK operations. Take the example of OpenStack/Cyborg as follows：
Cyborg ——————>Cyborg SPDK driver (in Cyborg) —————> py-spdk ——————>SPDK
In the above example, OpenStack Cyborg as a management layer calls upon SPDK module via
(1) its driver that will talk to the SPDK client and (2) the SPDK client that provides
the python biding.
This is almost a canonical procedure. For example when Kubernetes needs to use an
OpenStack infrastructure , it will first calls upon the OpenStack cloud provider (k8s’s
openstack driver), and then the cloud provider will calls OpenStack SDKs which intern
calls OpenStack clients and then calls into OpenStack native APIs. In such case the
OpenStack clients are hosted within the OpenStack community whereas the cloud providers
are hosted in-tree in Kubernetes community.
Although it should be very clear that unlike the cyborg-spdk-driver, py-spdk should be
part of the general spoke community. However it does make sense as you guys suggested that
such module may not be appropriate to be hosted in the master repo of SPDK.
Maybe it is possible to have an individual “spoke/client” repo for all the client side
modules ? (We might have other language binding clients for java, go, perl, … in the near
Is there anyone else interested in the proposal? Maybe you can review it in