On 2020-10-27 14:44, Jeffrey Steiner wrote:
I'm glad it was helpful! I could have gone on for two or three hours on the topic.
You probably should, record it, and send out the link to us :-) BTW I haven't watched you yet, thought I'd do that today. Hope I have ample time this evening. I'm in GMT+1 so...
I really can't emphasize enough how often we see situations devolve into solving nonexistent problems. I think it's mostly because all-flash changed things so much.
Hear, hear! PoC Purgatory. Yuk. PoC's with micro benchmarks for Storage systems. *sigh* Tells you nothing new you didn't already know, thing you don't really need to know for the real life use case, or could find out just as well (risk mitigation) in other ways than "testing" it yourself. Useless waste of time in 95% of all the cases people put $$$, time, effort into it
Back in the days of spinning drives, storage performance was almost always a problem, and there were almost always real benefits from trying to tune things a little better.
That's occasionally true for all-flash, but on the whole the problem shifted from "how can I make storage go faster?" to "what exactly do I need in the first place?"
I agree w this in principle. That said: there are lots of processing intricacies (I know, not really tunable...) that can go wrong and bottleneck inside a complex storage system like ONTAP. Yeah, I know... I know more about how WAFL works inside than I really should. :-\
Cheers, /M