Search results

How are you going to feed storage to your VMware cluster? Microsoft talks SMB3 and VMware talks iSCSI or NFS. It's possible to use Storage Spaces Direct for erasure coding and replication and put StarWind VSAN on top to expose VMware HCL-ed iSCSI/iSER but... It's kind of a train wreck :) ...and...

1) Response time looks good (sub 15ms), sequential I/O numbers indeed look pretty solid (for in-VM test). Did you notice extra disk activity on transfer drops? Lazy writer flushing ARC cache? Does memory usage on host change? 2) StarWind V8 (currently in final beta, release is planned for...

Sorry to bump the old thread but did you manage to complete your project? How's FreeNAS doing? :)

Do you run native IB or IPoIB? Any reason why you need a dedicated hardware set and don't share physical servers between hypervisor and storage (VMware VSAN-style)? Any DAS is still faster then even native IB as reads are going to SAS/SATA directly bypassing network.

It's a bad idea to mix StarWind a CacheCade. StarWind turns RAM to cache and that's faster then CacheCade on-board memory (system bus is wider and higher frequency and lower latency compared to PCIe). L2 flash performance with SATA/SAS attached would be the same but StarWind can use PCIe flash...

If you've used Windows built-in iSCSI you need to keep in mind it's not cached (NFS is). Also run I/O Meter to see average I/O response time and measure IOPS for ramdom I/O (4 threads, 16 I/Os in queue and 100% random / 100% seq) I/O.

Don't even try to feed NFS from Windows. It's a performance joke :( iSCSI is fine from the other point of view. You may try to use StarWind Free stuff for both performance and fault tolerance as Windows target has N/A.

Microsoft is pusing SMB 3.0 because it has no clustered file system capable or comparable to VMFS. LUN owner switch with NTFS is VERY expensive and Microsoft ODX has no ATS equivalent coming with VMware / VAAI.

I did not ask about how to do this as I'm an iSCSI zealot :) Could you please elaborate on why you did skip DataCore? Except price? What kind of reporting is missing?

FC comes to play when you think about latency. Real world latency for FC is much lower compared to one converged Ethernet can do. SMB 3.0 is not going to be better. It's still point-to-point to does not scale with a single initiator talking to many targets (the same scenario works fine for both...