Any one that knows DeepStorage knows that we’re opinionated and that we’re always willing to have a reasonable debate about our opinions. Unfortunately, as we’ve discovered in politics recently, some people take anything you say about them that isn’t a straight up compliment as an attack that must be responded to with as much force as possible.
Again, as in politics today, these people also have “Alternative Facts” insisting that their opinions must be accepted as facts, not the opinions they clearly are. A reasoned debate, however, has to have participants that can agree, as the great Daniel Patrick “Pat” Moynihan famously said, “’Everyone is entitled to his own opinion, but not to his own facts.”
A reasonable debate also requires the parties to avoid ad hominem attacks on the other party personally. We can debate positions all day and disagree without being disagreeable, but once you make it personal, there’s little chance of actually shedding light, rather than heat on the matter at hand.
So after writing what we thought was a sober comparison of the acquisition costs of comparable Nutanix clusters and Pure FlashStack converged infrastructure systems, Nutanix Blogger Josh Odgers wrote a 7000+ word screed attacking not only the content of the paper but our motivations and integrity.
I generally choose to take the high road on these things. I didn’t respond publically when Mr. Odgers insisted that my thoughts on data locality were a direct attack on Nutanix because any reference to locality in HCI MUST be about Nutanix.
We’re Not Perfect
Mr. Odgers did point out one factual error in our report. We misread the Nutanix spec sheet and used the power consumption of a block (4 nodes) in our calculations as a per node value. This does mean the Nutanix solution uses less power than the Flashstack.
As this was intended as just additional information for the cost of acquisition study, we did not include the cost of power in the main charts that compare the two systems’ cost.
After correcting our spreadsheet, power savings from the Nutanix solution amount to about 6% of the purchase price of the Nutanix cluster over 3 years assuming power costs at 20¢/KWH and a PUE of 2.0.
We regret the error, and we’ll be posting a revised version of the paper next week (week of Aug 21, 2017)
Facts Opinions and Assumptions
The Nutanix folks argue that any price comparison should be based exclusively on facts and be free of opinion. That would be nice, but there are always underlying assumptions, every TCO study assumes some number of person-hours at some hourly rate for the cost of labor. If the analyst assumes a $150,000 Ph.D. has to perform some task that will have a different effect on the bottom line than if he thinks a $50,000 tech can handle it.
I’m not going to present a full defense of the paper here. Frankly, a company that’s declared we should just “hang it up” doesn’t deserve the courtesy. I will point out two of our underlying assumptions, which like all assumptions are based on experience and opinion, that the Nutanix folks claim are factually wrong:
The Nutanix folks argue that our choice of vSphere was a factual error and that we should have instead chosen Nutanix’s own AHV hypervisor. After all, they argue, 23% of their customers now use AHV.
This places Nutanix in the curious position of arguing that over three-quarters of their customers are also making this factual error. We decided a cost analysis should follow the other 77% of Nutanix customers who use vSphere or Hyper-V instead of AHV.
Since Nutanix phrases it as 23% of customers, not 23% of workloads, nodes or clusters we assume that customers with mixed configurations like vSphere in production and AHV in test or even vSphere as their only hypervisor but Nutanix storage only nodes, which run AHV, are included.
It is my firmly held opinion that important workloads must be protected by at least a N+2 redundancy scheme. I’ve seen too many double drive failures and read errors on rebuilds to trust N+1. A few years ago at VMworld John Troyer had a game show format session. One question was “What’s the default FTT for VSAN”? The answer was “FTT=1, but @DeepStorageNet says it should be FTT=2”.
The paper lays out our logic and clearly identifies it as an opinion: “While each administrator must determine what level of resiliency he or she requires for their workloads, as relatively paranoid types our own standard for mission-critical workloads is that the storage system is able to survive…”
Again the Nutanix folks are presenting their opinion as fact by saying that RF2 (their N+1 2-way replication or single parity erasure coding method) is better than Pure’s N+2 RAID6ish approach and that our choosing RF3 could only be “to try and make the article sponsor’s product look more favorable.” not a long held opinion that’s common among storage professionals.
The response from Nutanix, a publically traded company, seemed a bit out of scale to me. Mr. Odgers can be a bit sensitive and I don’t view him as an official spokesperson for the company. I’m a big boy and I can take a bit of misdirected animus,
but then Nutanix’s President decided to chime in and tell the world DeepStorage should “hang it up”.
Is Nutanix’s official position, expressed publically by their President, and retweeted by a minimum of 10 Nutanix employees, that DeepStorage should “hang it up”?
I don’t really understand what he meant by “hang it up.” It could mean he would prefer that we stop writing about Nutanix, or he could mean that he would prefer that DeepStorage simply go out of business. Either way, it’s a curious thing for the President of a publically listed company with a multi-billion dollar market cap to say about a journalist/analyst in their field.
I’d appreciate some clarification.
Clearing up a True, not alternate, fact
Mr. Odgers’s summary concludes:
It’s important to note that even after DeepStorage has been made aware of many issues with the three reports, not a single item has been revised (at the time of writing).
This statement is simply untrue, and Mr. Odgers knows it. We did revise the Atlantis report that Mr. Odgers decided was an attack on Nutanix because we referred to Nutanix obliquely with “2.5 times the mailboxes of a leading HCI provider’s ESRP report” and “Five times the IOPS of that HCI provider’s system.”
My response to Mr. Odgers at the time, which we know he read because he quoted it, said:
“We don’t pretend to be perfect, and after Josh’s post led us to take one more careful read of the report, we did identify some -things that needed correcting:”
The revised report was posted less than a week later. While we didn’t make all the changes Mr. Odgers wanted, to say we made no changes is untrue.