in-memory HA storage (diskless vdi)

[anton (staff)]

You pay for StarWind-managed and StarWind-licensed capacity. So with a 3-way replication you'll have 1TB resulting "virtual" LUN. That's what you'll pay for. Dedupe will reduce amount of the on-disk data so with 1:5 ratio you'll have 200GB used on the disk on each node but still 1TB "virtual" LUN you'll "see" and pay for.

As I understand right now; 1tb 3 way license would give me 800gb SSD 200gb ram drive for 3 servers, so 1 TB HA served. If dedupe would increase my efficiency 5 times, I wont have to buy extra 4TB for that, Am I right?

Please clarify this for me.

[bacterialbag]

We license served capacity. So dedupe would decrease space on the underlying disk and you'll still pay for 1TB.

The 1tb license HA is BEFORE dedupe right?

I have some more questions regarding the HA ram setup for VM's;

I have a 3-node setup, is there anyway to get HA ramdisk, without using the size of the ramdisk on all 3 hosts for triplication? (or duplication in 2 node setup?) something like; each node has its own ramdisk, which is synced to the shared SSD storage. Something like that? Can you please give me some more explanation?

Atlantis is able to use 75gb of ram and serve 50 vm's on this, and actually use the 75gb as a 1.3tb virtual NFS disk, how do they do this?! Is this possible with starwind? Is your dedupe really better? because when I look at these numbers i'm a bit confused!

Also;
How far are you guys on shared cache? When will it be released?

Thanks

[anton (staff)]

1) We don't have any functionality for a RAM disk to load / store data on start / shutdown. Not planned as it's a cannibalizing feature to our in-memory scenario: it's very dumb to develop two branches of code allowing to do similar things @ the end of the day.

2) Having SSDs as a back-end storage to RAM is kind of weird setup: SSDs can technically give all IOPS for your VMs. So you just increase installation cost by having RAM oversized. What for? Please clarify maybe I just misunderstand what you want to do...

3) Deduplication. That's the way to store 1+ TB in 64+ GB. No magic. Atlantis uses licensed dedupe engine. We have own one. Our one is better (faster and better saving numbers).

4) Depends on what you call a "shared" cache. If it's a dedupe cache then it's shared by design and we do this from the very beginning.

[bacterialbag]

I want user experience to beat laptop with ssd performance, that is why I run on SSD and hope your solution can get me the same in-ram performance as Atlantis. (those vm's using atlantis are crazy fast, I want to achieve the same using starwind!)

Hope you would have some advice for me on how to best setup/configure things to achieve that?

Thanks!

[anton (staff)]

StarWind In-Memory scenario is not released to public yet.

I want user experience to beat laptop with ssd performance, that is why I run on SSD and hope your solution can get me the same in-ram performance as Atlantis. (those vm's using atlantis are crazy fast, I want to achieve the same using starwind!)

Hope you would have some advice for me on how to best setup/configure things to achieve that?

Thanks!

[bacterialbag]

StarWind In-Memory scenario is not released to public yet.

I want user experience to beat laptop with ssd performance, that is why I run on SSD and hope your solution can get me the same in-ram performance as Atlantis. (those vm's using atlantis are crazy fast, I want to achieve the same using starwind!)

Hope you would have some advice for me on how to best setup/configure things to achieve that?

Thanks!

When is release?
When can I get beta?

In what time-period will you finish implementing most in-ram features, and deliver a stable, in-ram capable product?

[anton (staff)]

We'll have big version update in mid-January (mostly LSFS and some other things) and after that I'll be able to tell accurately.

[bacterialbag]

We'll have big version update in mid-January (mostly LSFS and some other things) and after that I'll be able to tell accurately.

Okay so what about beta? You said give us a few weeks a few weeks ago, but if I understand you correctly now, it will take you guys longer?

I thought the update that was coming was the update with the ram upgrades?

Could you clarify this for me please?

[anton (staff)]

We're busy with the other things and shifting and adjusting priorities the way we decide In-memory thing did not turn into public beta state yet. Sorry about that...

We'll have big version update in mid-January (mostly LSFS and some other things) and after that I'll be able to tell accurately.

Okay so what about beta? You said give us a few weeks a few weeks ago, but if I understand you correctly now, it will take you guys longer?

I thought the update that was coming was the update with the ram upgrades?

Could you clarify this for me please?

[bacterialbag]

So, hows it going with the in memory stuff?

Hope to see beta soon.

We're busy with the other things and shifting and adjusting priorities the way we decide In-memory thing did not turn into public beta state yet. Sorry about that...

We'll have big version update in mid-January (mostly LSFS and some other things) and after that I'll be able to tell accurately.

Okay so what about beta? You said give us a few weeks a few weeks ago, but if I understand you correctly now, it will take you guys longer?

I thought the update that was coming was the update with the ram upgrades?

Could you clarify this for me please?

[anton (staff)]

Goes well. We currently deliver 3x-10x times more IOPS then competitive in-memory solutions using the same hardware. In-memory storage is not highest priority now so time frames for beta would be clear closer after next StarWind update (touches LSFS, cache, cache pre-heat and other things related to in-memory storage).