golang-kv VS heketi

Machine-Learning(ML), CortexVirtual Machine(CVM) & Model RepresentationTool(MRT) R&D 1. Use the AdaRound method on the Image Classification task and integrate it into the MRT; 2. Refactor the operation and usage of MRT, adopt the argparse library, and support the specification of environment variables and command line parameters; 3. ML network needs to be quantified, and Pytorch network quantization operation is supported at the same time, the model input range is changed from 0–1 to 0–255 to retrain; Cortex ZkRollup 1. Add JS API (provider, signer and wallet at the L2 layer) to interact with ZkNode, and JsonRpc to interact with ZkServer; 2. Modularize JS functions, write the necessary SDK, prepare for the subsequent front-end development, and complete part of the code test; 3. ZKNode builds the Server module, uses Flask tool to design the Web backend, and connects with the frontend JsonRpc interface; 4. ZkNode separate module function, realize some functions of log, command line and transaction pool; 5. ZkNode adds Layer One signature method and supports OnChainDataHash auxiliary storage; 6. The ZkRollup contract is improved, some problems are fixed, and the contract logic function is tested; Cortex Full Nodes 1. Update the DNS discovery API and repair the block header chain synchronously; 2. Optimize RPC subscription thread closing synchronization logic; 3. Optimize the startup detection and allocation logic of full node and CVM virtual memory; 4. Model Cache repeated loading optimization, try SSD Cache to replace Memory Cache; 5. The separated storage engine is a switchable module, and it is expected to support more storage types. Refer to the link: https://github.com/ucwong/golang-kv/; Miscellaneous 1.Update address data statistics on the Cortex chain;

Something I have been blissfully ignorant about is the glusterfs support that seems abandoned. There was gluster/gluster-kubernetes with the goal to easily create gluster clusters and heketi providing a restful api for that.