#RIP FAS9500
1 messages · Page 1 of 1 (latest)
Why not? Same drive count. Same nvram. Sure the 9500 has more ram but the 90 is a significantly newer cpu and chipset that can do more with less ram
Flexgroups require RAM. The FAS90 has a quarter per node. That means I should plan on at least twice as many nodes as the FAS9500. More Flash Cache, etc. Hopefully they have dropped the GB licensing to compensate
why would you need to license more gigabytes storage licensed if you spread the same data across more nodes?
Also, from what I've seen, even with FlexGroups the FAS90 blows the FAS9500 out of the water.... but if you still need the double RAM (not 4 times btw) , you can still buy the FAS9500 for the forseeable future
I mean I know that more RAM helps with speed, especially in spinning-rust systems where cache is essential, but I guess in such a case it would probably be a good idea to look into things like FabricPool which can provide even more performance with smaller controller models
FabricPool is highly overrated
in your opinion maybe 🙂
Yes, in my opinion'
for me it's FlexGroups that are overrated
Good for you man. Let's agree to disagree and keep it civil, shall we?
in any case, since there are no official numbers for the FAS9500 performance (sadly they're missing from the performance charts) there's no absolute numbers to compare
With FAS and spinning discs, I don't expect AFF or CFF performance. But with RAM, constituents, and FlashPools, I am able to provide an HPC environment with the performance and CAPACITY they require.
I do have an AFF for really critical runs, but most of my HPC runs use spining disc
so, do you have any hard numbers that the FAS9500 can reach but the FAS90 cannot? Like IOPS or throughput or latency?
No, but I do have 8700 numbers. I have 8700's and they die
well, sadly I don't have first-hand experience with the FAS90 yet (our customers prefer to buy C Series for some reason 😉 ) but I guess that if you share your concerns with your NetApp contacts they can surely provide you with an NDA session and hard performance numbers for the new models
It would be nice if the numbers were published
the FAS8700 is in the perf doc with a comparison to the FAS90. Only for IOPS though, and with 5 ONTAP versions difference
the number is around 3 times that of the FAS8700
Well, I'm not ready to replace the 9500's yet. Again, my concern is the RAM used for FG's.
search for "performance tech spec" in the field portal.
yeah, again, you don't have to replace them as they are not even EOA yet..
Yep. FG's have been a lifesaver, but they are RAM and Flash dependent for performance
waiting for a bit before buying the new model is not entirely bad... iron out the last few hardware issues etc., wait for the performance numbers from other users, etc.
during announcement NetApp said the FAS90 is "35% faster than FAS9500" but no real numbers have been released so far
Additionally another two nice differences:
- FAS70/90 has QAT support which FAS9500 has not, but needs to be enabled manually (bottleneck will be HDDs of course)
- FAS70/90 supports NS224 shelves which FAS9500 does not
also less rackspace
Fabricpool is awesome and saved us a shitload of money.... A-Series Metrocluster with 65 weeks of snapshots... You know, regulations
Yes, I am sure. However, we are all FSAS, and on-prem. I need to do the reverse of FabricPool, aka FlexCache and automatically move data to high perrformance storage, and back to FSAS when done. My problem is that I have huge repositories, and they are getting bigger, not smaller.
For me to use FabricPool, I would need to move 9PB of data to high performance storage, so I can then migrate it back.
In the meantime, I use FlashPools, which cannot be used with FabricPools in a StorageGrid config. So it would be very expensive for me to implement.
FlashPool running fine in your environment? Because we learned our lesson with FlashPool when it came out, and stopped implementing/recommending it after. That feature was so buggy in the first years, I think almost every FP installation had some issues at some point (the SSD part running full even though the HDD part has plenty space, snapshots not rotating properly, various panics related to FlashPool, etc.)
Glad to see that these issues seem to be fixed by now!