HA with write-back cache and other questions

[sjutras]

Hi,

My first post here. I have been trying out starwind in trial and i must say i am very impressed with its ease of use and performance I have a few questions, been trying to find out about them in documentation or this forum but didnt find my answers, hopefully i didnt over-look too quick!

1) If i understood correctly, with write-back cache enabled, if the SAN fail (hardware failure , etc) there is good chance (or its 100% sure) that the data in the .img file or the physical disk if using disk bridge, will be corrupted?

2) But using cache is appealing for performances, and i recall having read on this forum that with HA configuration, with write-back cache, if one node fail, the data is still "clean" on the other node? What happen if the data being written is cached on the node that fails? Is this also true with raid-1 (mirror) target?

3) Is there any way to tell starwind to use all RAM available on the system and use it for write-back cache as it wishes among all the target configured?

4) Is there any ways to modify cache settign on a target without re-creating that same target (or modifying .cfg + service restart)

5) LUN ID in vmware, I just read the post where anton mention we should be able to configure this in v5.4 but i am not sure where to configure this in the console

Thanks in advance!
Seb

[Max (staff)]

1) If i understood correctly, with write-back cache enabled, if the SAN fail (hardware failure , etc) there is good chance (or its 100% sure) that the data in the .img file or the physical disk if using disk bridge, will be corrupted?
>You have understood correctly.

2) But using cache is appealing for performances, and i recall having read on this forum that with HA configuration, with write-back cache, if one node fail, the data is still "clean" on the other node? What happen if the data being written is cached on the node that fails? Is this also true with raid-1 (mirror) target?
>In case of HA when one node fails, the other node writes all the cache to the hard drive and switches it off to insure maximum reliability. In case of mirror (raid1) the situation will be a bit worse if the server is not protected against power faults.

3) Is there any way to tell starwind to use all RAM available on the system and use it for write-back cache as it wishes among all the target configured?
> Just specify the desired amount in the device creation wizard, however I do not recommend to set the cache size more than 70% of total RAM available.

4) Is there any ways to modify cache settign on a target without re-creating that same target (or modifying .cfg + service restart)
> The only way to modify the caching operation mode in the current version is to edit the .cfg file and restart the service.

5) LUN ID in vmware, I just read the post where anton mention we should be able to configure this in v5.4 but i am not sure where to configure this in the console
> This feature has not been added to 5.4, you can join the beta program on our website to find out more.

[sjutras]

Thanks alot, clear answers!

[anton (staff)]

Back to Multi-LUN support. For now we have it in experimental mode and image file only (private featue) but performance and compatibility don't shine that's why we've delayed public access to it until we'll not have 6.xx branch with Multi-LUN support in core (fall 2010), together with VMware ESX/ESXi certification

Thanks alot, clear answers!

[sjutras]

Great news!

One more question, is there a performance difference between .img file and direct disk bridge?

[anton (staff)]

With caching enabled - NO.

Great news!

One more question, is there a performance difference between .img file and direct disk bridge?