Wizard
Wizard
> > I see the function > > https://github.com/tkestack/vcuda-controller/blob/5ec3fcbc58679bbbbd82ef9e21d647d8d7383876/src/hijack_call.c#L713-L731 > > > > so, should it be change to > > ```c > > request_size = base_size * pAllocateArray->NumChannels *...
> As if vcuda would copy lib to container after container start up. When start multiple GPU-resource pods simultaneously, this action is not fast enough. You can try to modify...
> Can you expand on your motivation for FIFO and causal boardcast? Would a wrapper around `GossipSub` suffice which caches events and only emits them based on your policy? Ok
> > You need to add the learner first by using `ConfChangeType::AddLearnerNode`, and then promote it to voter using `ConfChangeType::AddNode`. Adding a learner is similar with add a node. Thanks...
> Raft maintain the progress in leader, which can be query by `status` API. You may also want to check out #457. Thanks for your reply. That was great!
@rpbear88 你们遇到的问题和阿里 polardb 团队很类似, shard-stoage 的数据库复制的时候需要 raft 乱序提交 + 乱序执行来降低延迟, 他们提出了parallel raft 算法, 但是我看了论文对比了一下 multi-paxos, 这个对 raft 改造很大... 所以你们可以考虑不基于共识的方案, 简单来说就是放弃 raft...
I tracked kuhantech now