QoS minimum / floor implementation

[ibkl4]

Does anyone knows why the Adaptive QoS floor feature is implemented/available in Ontap 9 but the QoS floor feature is only available for AFF?

[AlexDawson]

At a guess, the latency of rotational media makes it difficult to implement in a way that SSD's near-instant and consistent access time do not. If you require a definative answer, please send me a message or email and I will connect you to our TME team for AFF/FAS

[Ontapforrum]

@Alex: Do you mind answering this query: It says 'cluster must have at least one AFF node*' , What does it mean by one AFF, ususally we add nodes in pair. Is it one Pair or One node with only Flash Aggregate in the Cluster ?

*Min-throughput options are available only on All Flash FAS (AFF) Platforms running ONTAP 9.2 software, where the cluster must have atleast one AFF node.

FAQ: Storage QoS policy in System Manager 9.2:
https://kb.netapp.com/app/answers/answer_view/a_id/1001095

Sorry for chipping in.

[ibkl4]

@Ontapforrum: We are using a 4 Node cluster with 2 FAS and 2 AFF using Ontap 9.5. The problems remains:

- Configuring a floor/minimum AQoS rule on a  volume on rotating spindles (FAS) works perfectly.

- Configuring a floor/minimum QoS rule on the same volume cannot be configured.

- Configuring a floor/minimum QoS rule on the same cluster on AFF works perfectly.

Question: Is there a technical reason behind?

[Ontapforrum]

@ibkl4  :I am sure engineering must have given enough thought and testing to state that 'QoS' minimum only applies to AFFs. I haven't configured or worked with AdaptiveQoS so can't really comment on that. However, I think the whole game of QoS minimum or maximum consists of complex algorithms and logic. How quickly system responds is probably one of the key factors to how QoS ratebucket rebalancing works. With differnet workloads and rate of client traffic it makes it difficult to compute and distribute the credits. B'cos AFFs are all flash, that means they are quick in their response time and to adjust their IOs, which is probably not realiably possible with spinning disks. Having said that, this is just my interperation and I may be completely wrong on this. I am sure NetApp will come back to your query in due course. Thanks!

[ibkl4]

---

@AlexDawson wrote:

At a guess, the latency of rotational media makes it difficult to implement in a way that SSD's near-instant and consistent access time do not. If you require a definative answer, please send me a message or email and I will connect you to our TME team for AFF/FAS

---

Fully understand, but it's implemented for Adaptive QoS on FAS (rotational medias). That's why I'm asking, because I have a customer which has AQoS implemented and wants to change to QoS. I haven't got an explanation why the floor QoS suddenly should not work anymore.