Where are ROW modes in V8?

digitalis99 · Fri Oct 24, 2014 10:27 pm

Subject line pretty much says it all. I'm trying to figure out how you can configure a disk in redirect-on-write mode like existed in V6. Does V8 add the ability to store such a disk in RAM? I talked a bit about ROW functionality related to V6 here, and was told to wait. V8 seems to not include it at all. ???

Sat Oct 25, 2014 8:41 am

LSFS does Redirect-On-Write. LSFS does In-Memory operations (currently locked from user but we'll unlock it very soon). LSFS has deduplicated L1 cache paired non-deduplicated L1.

digitalis99 wrote:Subject line pretty much says it all. I'm trying to figure out how you can configure a disk in redirect-on-write mode like existed in V6. Does V8 add the ability to store such a disk in RAM? I talked a bit about ROW functionality related to V6 here, and was told to wait. V8 seems to not include it at all. ???

digitalis99 · Sat Oct 25, 2014 3:08 pm

anton (staff) wrote:LSFS does Redirect-On-Write...

OK, then I guess my question is still valid. Where do you configure it? When I configure a thin-provisioned LSFS device, the only option in the "Advanced" options is a checkbox for deduplication. I don't see any option that would allow changing or reconfiguring an existing LSFS device. This is in the free version for Hyper-V, in case that makes a difference.

Sat Oct 25, 2014 6:33 pm

You cannot control Redirect-on-Write as there are no options of features you enable or disable or tweak. It just always works. Like gravity on Earth. Or do you want to "play" with snapshots?

digitalis99 wrote:
anton (staff) wrote:LSFS does Redirect-On-Write...
OK, then I guess my question is still valid. Where do you configure it? When I configure a thin-provisioned LSFS device, the only option in the "Advanced" options is a checkbox for deduplication. I don't see any option that would allow changing or reconfiguring an existing LSFS device. This is in the free version for Hyper-V, in case that makes a difference.

digitalis99 · Sat Oct 25, 2014 7:05 pm

anton (staff) wrote:You cannot control Redirect-on-Write as there are no options of features you enable or disable or tweak. It just always works. Like gravity on Earth. Or do you want to "play" with snapshots?

I'm confused. Let me expand on what I do in V6 to hopefully learn what the parallel is in V8.

In V6, I use redirect on write modes (with discard) of .IBV's in order to boot multiple hardware instances off of the same image without write collides. You're saying I can do the same thing with an LSFS device without telling it a bunch of systems are going to want to write to it? That just feels weird...seems like I should be configuring the device to support/expect that explicitly.

If there's no configuration necessary, how do I tell the device to discard the writes of closed sessions? Whether or not to discard must be a setting somewhere.

Sat Oct 25, 2014 7:09 pm

I see now... That's not ROW it's just a feature based on ROW. OK, let me talk to engineers on Monday as I'm not sure this functionality is preserved. Hold on, please! Thank you for patience and sorry for confusion

digitalis99 wrote:
anton (staff) wrote:You cannot control Redirect-on-Write as there are no options of features you enable or disable or tweak. It just always works. Like gravity on Earth. Or do you want to "play" with snapshots?
I'm confused. Let me expand on what I do in V6 to hopefully learn what the parallel is in V8.

In V6, I use redirect on write modes (with discard) of .IBV's in order to boot multiple hardware instances off of the same image without write collides. You're saying I can do the same thing with an LSFS device without telling it a bunch of systems are going to want to write to it? That just feels weird...seems like I should be configuring the device to support/expect that explicitly.

If there's no configuration necessary, how do I tell the device to discard the writes of closed sessions? Whether or not to discard must be a setting somewhere.

digitalis99 · Sat Oct 25, 2014 7:22 pm

Phew! I thought I was going crazy there for a minute.

Sat Oct 25, 2014 8:18 pm

https://www.youtube.com/watch?v=rS_S1bd-SjQ

digitalis99 wrote:Phew! I thought I was going crazy there for a minute.

Tue Oct 28, 2014 10:09 am

In V6, I use redirect on write modes (with discard) of .IBV's in order to boot multiple hardware instances off of the same image without write collides. You're saying I can do the same thing with an LSFS device without telling it a bunch of systems are going to want to write to it? That just feels weird...seems like I should be configuring the device to support/expect that explicitly.

If there's no configuration necessary, how do I tell the device to discard the writes of closed sessions? Whether or not to discard must be a setting somewhere.

We are adding new features to LSFS to support scenario like this:
- ability to create several LUNs on common LSFS container;
- ability to create new LUN based on existing snapshot;
- large deduplicated cache to store common active data set in memory.
I think, we'll be able to give some beta version for tests at the start of December.

Until then, I would suggest using V6 to implement the scenario with ROW and discard.

digitalis99 · Tue Oct 28, 2014 5:31 pm

Cool. Is there a place I can download the last of the V6 series? I have version 6.0.5228, but I don't know if that was the last in the series.

Tue Oct 28, 2014 5:43 pm

I've sent a PM with link to the latest V6 build.

digitalis99 · Thu Mar 12, 2015 2:04 am

Alex (staff) wrote: We are adding new features to LSFS to support scenario like this:
- ability to create several LUNs on common LSFS container;
- ability to create new LUN based on existing snapshot;
- large deduplicated cache to store common active data set in memory.
I think, we'll be able to give some beta version for tests at the start of December.

Until then, I would suggest using V6 to implement the scenario with ROW and discard.

Sorry to dredge up an old thread, but I was curious if there was any development/progress on this? Sounds like there are some issues with RAM caching an LSFS at the moment, but it sounds like that will be fixed soon. No rush on this, just curious.

It would be really nice to move to v8 so we can RAM cache the "golden image" and/or the ROW with discard devices. v6 is, ahem, not very optimal since it has to go straight to disk for EVERYTHING....

Thu Mar 12, 2015 4:39 pm

We're currently on a verge of a incremental update which includes caching improvements.
I would recommend starting the testing as soon as it is released
ETA is end of next week

digitalis99 · Thu Mar 12, 2015 5:37 pm

Anatoly (staff) wrote:We're currently on a verge of a incremental update which includes caching improvements.
I would recommend starting the testing as soon as it is released
ETA is end of next week

Cool. Have you re-implemented the ROW with discard functions of v6 into v8 yet?

Tue Mar 17, 2015 7:10 pm

Not at the moment.
But we are actually planning to implement something alike.
I don`t have ETA on this unfortunatelly.