9 posts • joined 5 Jun 2017
Wow. What a shame. Why do you use wrong NetApp's numbers? There are two groups of graphics in NetApp's white paper — with image distortion enabled and disabled. And you took results for test with image distortion enabled. Pure tested FlashBlade only with image distortion disabled.
So right numbers for NetApp are:
ResNet-152: 138, 274, 547, 1092
ResNet-50: 341, 631, 1259, 2521.
Please update the article.
ORT of NetApp is 1.04 ms and WekaIO has 1.02 ms. Do you call it much higher latency?
NVMe FlashCache is read only cache. I hope SFS2014_swbuild has some amount of write operations and suppose there are huge amount of file metadata changes.
NetApp has 8 times more of usable space. I suppose it doesn't make a lot of sense for source code storing. But clients do not buy storage for only one type of workload.
Anyway my first point was that for me just 2 times more builds on system with NVMe SSDs versus NL-SAS drives with the same ORT is not very impressive.
Re: Safe enough - IF no third party code
What about Pure Storage Purity Run functionality? It gives you possibility to run VMs and containers on FlashArray controllers. So they have to patch FlashArray.
And if we are talking about some vendors who use Linux as base for their storage OS, for example EMC Unity is based on SuSe. Meltdown/Spectre patch is a kernel patch, so eventually EMC will have to update their OS to newer version of kernel and my hit performance degradation.
Re: HCI - the great distraction of our time.
SolidFire guarantees 50K/100K IOPS per node (depends on exact model) and there is no matter what kind of data do you store. No matter can this data be compressed or deduplicated, use you snapshots, clones, replication or not. The same thing with QoS. Minimum performance is always guaranteed.
I think there is less performance per HCI node as it uses just 6 SSDs, but anyway it is guaranteed and independent from compute nodes that gives you really predictable performance for VMs.