search

quiver-team / quiver-feature

High performance RDMA-based distributed feature collection component for training GNN model on EXTREMELY large graph
Apache License 2.0
48 stars 5 forks source link

测试随着数据集的增大导致的寻址开销带来的影响，以及PageSize带来的影响 #10

Closed eedalong closed 2 years ago

eedalong commented 2 years ago
eedalong commented 2 years ago

随着数据集增大，单边读带来的性能影响，在2机2节点的情况下，每次Collect 80K个节点，特征维度为600维度，这里的测试是网络带宽的测试，即理论上限是12.5GB/s

With TLB Optimization

2.2G: 10613GB/s 22G: 10668GB/s 44G: 10665GB/s 89G: 10663GB/s 178G: 10765GB/s