应该以任何方式工作 - 它与配置机制无关。
GPU passthrough或vGPU分配都是可能的。
我们使用MCS是因为它(1)更容易,(2)更便宜(不需要额外的配置服务器),(3)我们已经使用精简配置和缓存存储服务器,因此已经获得了很好的I / O性能。
附:
是的,通过溢出到磁盘缓存到RAM会很好,最终可能会来到MCS。
以上来自于谷歌翻译
以下为原文
Should work either way -- it has nothing to do with the provisioning mechanism. Either GPU passthrough or vGPU assignments are possible. We use MCS because it's (1) easier, (2) cheaper (no extra provisioning server(s) needed), (3) we use thin provisioning already with a caching storage server, so already get great I/O performance.
P.S. Yes, cache to RAM with overflow to disk would be nice, and will probably come to MCS eventually.
应该以任何方式工作 - 它与配置机制无关。
GPU passthrough或vGPU分配都是可能的。
我们使用MCS是因为它(1)更容易,(2)更便宜(不需要额外的配置服务器),(3)我们已经使用精简配置和缓存存储服务器,因此已经获得了很好的I / O性能。
附:
是的,通过溢出到磁盘缓存到RAM会很好,最终可能会来到MCS。
以上来自于谷歌翻译
以下为原文
Should work either way -- it has nothing to do with the provisioning mechanism. Either GPU passthrough or vGPU assignments are possible. We use MCS because it's (1) easier, (2) cheaper (no extra provisioning server(s) needed), (3) we use thin provisioning already with a caching storage server, so already get great I/O performance.
P.S. Yes, cache to RAM with overflow to disk would be nice, and will probably come to MCS eventually.
举报