Jeff Baxter

I remember they used to, if there was an hour long or two hour long line, they'd come down the line and say, does anyone have a flight leaving in the next 45 minutes?

Does anyone have a flight leaving in the next 30 minutes?

and pull someone to the front of the line.

That's basically what we're doing here, right?

All workloads continue to get processed.

Even if they don't have a QoS minimum attached or anything like that, they're going to continue to get processed.

But what's going to happen is as you queue up at any stage in ONTAP for next processing, if you have a deadline attached to you and if you're not going to make that deadline by the front of the queue, we will pull and reorder you to the front of the queue so that you do make your deadline and you do make your flight, or in this case, you meet your QoS minimum.

So it's not just something that we've glommed onto the top of it.

It's actually built into the entire ONTAP stack to ensure that people get the performance they've asked for when they set a KOS minimum.

Yeah, exactly right.

Exactly right.

It'll start to punish anything that doesn't have a QS minimum set, and you'll just start to see the latencies increase.

Yep, absolutely.

It's great because you can put maximums on top to stop runaway neighbors, to set the expectation of performance.

I was telling the story, and someone said, no, it's important to have the SLAs on top of the QS maximums, but the other important thing is

When people pay for 40,000 IOPS, I want to give them no more than 40,000 IOPS because when it's the first start of a system, it's not heavily loaded.

If they start getting 75,000 or 80,000 IOPS, that becomes their new normal.

And then when I do load the system up over the next six months or next year, all of a sudden they start getting the actual 40,000 IOPS.

All of a sudden it feels like they're getting less, you know, half the performance when in reality they were getting more than they were ever paying for.

So it's important for us to provide that maximum capability so that people are