Wang lu
2008-Nov-12 18:10 UTC
[Lustre-discuss] May the crash caused by network is faster than disk array?
Dear list, In my configuration and test 1 OSS has 10Gbit/s network, however, the disk peak performance is around 300MB/ s. I think it is the peak performance, because before the OSS crash, the %util is 100% and read request queue is more than 1000, and at that time, the disk reading performance is around 300MB/s. I have 50 clients, each has 1Gbit/s network, which may reading/writing Lustre simutanously at any time. Aggregate client Network bandwidth> Aggregate Total OSS Network bandwidth> Aggregate Total OSS disk peak throughput May the OSS crash because disk cannot handle too much request comming from network?