StarWind Software

Hello

Firtly I have recently discovered starwind and that we can do with it is amazing, specifically on a cluster, thank you for that.

Secondly i have few problemes with ibm servers, i can not see any disks when i choose for example spti device

In the log there is this message :

4/27 14:19:34.756 620 SPTI: '\\.\PhysicalDrive2': The adapter supports maximum transfer length 61440 bytes less than minimal allowed (MinBufSize = 65536 bytes)!

So where is the problem, at the raid adapter level or wih disks ? the adapter is an IBM Serveraid 5i and disk 18.4 SCSI on a xseries 235. I have tried with 2 ibm servers without succeed but with 2 dell 2650 and 6600 it works perfectly.

Thank you in advance !

There are at least FOUR problems with what you want to do actually I'll try to list all of them if you don't mind of course...

1) Your RAID drivers are either broken or improperly configured. Having requests less then 64KB (MM_DISK_IO_SIZE #define) is a complete performance disaster. Issue has nothing to do with our software so please contact your driver vendor for either upgrade or letting you know how to trigger registry to allow at least 128KB requests.

2) StarWind Free has SPTI stuff DISABLED for non-evaluation period. So even if you'd manage to bypass Issue.1 you'll be stuck in 30 days after time is going to be up.

3) SPTI mode is neither recommended nor supported for hard disks. Reason is very simple: by design (Microsoft) Windows storage stack has SPTI requests non-overlapped. This means you cannot fire N+1 request before N did not complete. This means in a nutshell: NO QUEUE. No pipeline. No performance. That's why we've DISABLED our SPTI mode for hard disks leaving it for non-performance one-by-one by it's nature devices like tapes and optical discs. There's a way to hack down into StarWind XML configuration file and ENABLE disk drives support with SPTI but as I've told we don't recommend or support this mode. So bypassing Issue.1 and Issue.2 is going to lock you inside Issue.3

4) There's no guarantee your particular disk drive has command set tested with your hypervisor. StarWind has virtualization layer above hard disk and our virtual entity is tested against latest Hyper-V, ESX and Xen builds. Even 2 year old disk was obviously not. So even if you'd bypass Issue.1, Issue.2 and Issue.3 you'll have extermely hard times trying to find out why everything works until you start using Live Migration or VMotion or whatever.

So what you should do? Stick with DiskBrige, allow high-performance virtualization layer wrap your hard disk into multigigabyte write-back cached one (make sure you run High Availability configuration here as having WB cache w/o backup is like running on GT-R w/o brakes, acceleration is still perfect but cornering simply kills). Performance, compatibility and support

Hope this helped!

spack wrote:Hello

Firtly I have recently discovered starwind and that we can do with it is amazing, specifically on a cluster, thank you for that.

Secondly i have few problemes with ibm servers, i can not see any disks when i choose for example spti device

In the log there is this message :

4/27 14:19:34.756 620 SPTI: '\\.\PhysicalDrive2': The adapter supports maximum transfer length 61440 bytes less than minimal allowed (MinBufSize = 65536 bytes)!

So where is the problem, at the raid adapter level or wih disks ? the adapter is an IBM Serveraid 5i and disk 18.4 SCSI on a xseries 235. I have tried with 2 ibm servers without succeed but with 2 dell 2650 and 6600 it works perfectly.

Thank you in advance !