In Software Defined Network(SDN), we propose An Adaptive Read/Write Optimized Algorithm for Ceph Heterogeneous Systems via Performance Prediction and Multi-attribute Decision-making.The reference link is as follows: https://link.springer.com/article/10.1007/s10586-022-03764-3.
In addition, we introduce Olsync, an object-level tiering and coordination system designed to enhance storage resource utilization and file access performance. Specifically, Olsync employs PIPO (Packet-In-Packet-Out), an innovative network communication framework based on SDN, to collaboratively optimize both the network control plane and underlying data plane.
Environment: Ubuntu20.04
- Let's download the original RYU controller
Execute the command: pip install ryu
- We modified the source code of RYU controller to obtain the new version RYU_M1.0, which mainly modified the functions of SDN_monitor. py. The modified RYU_M1.0 can actively obtain the network status of Ceph cluster nodes. It can also passively receive load information from OSDs collected by the underlying nodes.
Environment: CentOS7
-
When we complete the preparations, we use the CEPh-ansible tool to automate the Ceph cluster deployment.
-
We have added a custom OSD_INFO_MAP to the Ceph source code. OSD_INFO_MAP can be combined with SDN network model to maintain the load and network status information of OSD in real time and can be sensed by CRUSH algorithm. Finally, we use this information to build prediction model and multi-attribute decision mathematical model to optimize the read and write performance of the cluster.
-
Olsync can offer efficient object-level tiering and coordination services using the global views obtained by PIPO in SDN (e.g., data access patterns and interfering object requests) to make tiered storage and performance optimization decisions.
Olsync references Amazon Redshift clusters to achieve the optimal "performance-cost" balance. Links to Amazon Redshift clusters and their datasets can be found here: https://github.com/aws-samples/amazon-redshift-tiered-storage. The IBM Object Storage datasets that Olsync used for testing is linked here: http://iotta.snia.org/traces/key-value/36305.