Who has 2,000 tickets to the gun show? Cisco's HCIA platform HyperFlex After 18 months of selling, some 2,000 customers are travelling along the Cisco HCIA highway. HyperFlex is Cisco's hyperconverged infrastructure appliance (HCIA), marrying acquired Springpath's software with Cisco's UCS servers. Back in February, there were more than 1,000 customers. Seven months later, in September, Cisco …
Some background info, AC for obvious reasons: That customer is a Fortune20 company which is replacing their entire HPE infrastructure (servers and storage) with UCS HyperFlex. This is obviously a massive project and after the initial 800 are deployed and first batch of workloads migrated into that private cloud, additional 2000 (and more) nodes will be added. This was a highly competitive deal against Nutanix and VxRail who were disqualified early in the process due to lackluster storage performance.
Btw, the 2000 customers numbers is not correct, it is significantly higher. It is however the number floating in the currently available internal marketing deck which hasn't been updated in the last two or three months. Also, several members of the Fortune10 are deploying HyperFlex.
Nutanix has stopped badmouthing HyperFlex (they still do of course but no longer with outright lies and dismissing it) and is now offering discounts of up to 70% (usually 25-35% off internet list price) to match HX prices.
@chris -
If this really is a Fortune 20 customer replacing all their HPE infrastructure for HyperFlex it shouldn't be to difficult for a hotshot reporter to find out who the customer actually is. Which means it shouldn't be to hard to find out who the Cisco partner actually is.
@AC for obvious reasons -
"disqualified early in the process due to lackluster storage performance" I'd like to hear the details. How did this customer determine storage performance on HyperFlex was better than Nutanix and VxRail? What toolset, datset, tests, etc. did they run to make this decision?
Nice try. What you call "thrown in" is called "seeding" where a base cluster with only 12 months of support and software subscription (and implementation) is added to a larger deal. Once the customer tests real workloads on these clusters and decides to move production to them, they have to buy additional nodes, support, and software subscription.
Those don't count as HyperFlex customers but are PoV/PoC. Cisco did this successfully with UCS many years back and it's been wildly successful. Dell, HPE, and others do that all the time.
My guess is you work for one of those competitors and aren't too familiar how accounting, profitability, and commission based compensation work. Cisco has multiple business entities and HX is one of dozens of products within the data center business entity. There might a few people whose salary depends on HyperFlex doing well but tens of thousands who do not. And since Cisco is a Fortune100 company, they are definitely not in the business of giving away stuff for free and lowering margins just to get some bragging rights.
What you are implying is that Cisco is sacrificing core business revenue and margins to boost HyperFlex numbers during the largest core business refresh in the past 7-8 years and Wall Street scrutiny of said core product refresh. You are also implying that a large number of sales representatives and their organizations are voluntarily giving up commission so that the HyperFlex guys get paid. That's a ridiculous statement on so many levels.
HCI is supposedly the integration of compute, storage, networking, and hypervisor. Boundaries are a bit fuzzy, some claim that orchestration is also included. I am not sure about that but unified management is definitely a major aspect of HCI as is streamlined patching and upgrades.
VSAN is by definition only software defined storage. VxRail (a VCE product, not Vmware) mandates specific hardware and includes VxRail Manager. For both scenarios, managing the network is a bit of a PITA; the VSAN network implementation guide is quite a read.
VxRail with NSX, vRealize, and some sort of streamlined network integration (I think VCE will sell Brocade IP switches if you ask but they might not be preconfigured) would be a true HCIA.
Vendors will always discount a big client. Vendors will always discount new products. When I was at Unilever and they were migrating their entire SAP estate off Oracle DB onto IBM DB2, which grapevine said was pretty much free in comparison, there was an IBM account manager onsite as his full time job.
You don't get momentum without selling it at a loss til you get established and benefit from economies of scale. If anyone can afford to take short term losses then Cisco can.
I got a couple Hyperflex for free with an order, some older rack mounted servers. I’m guessing they are counting us as a customer, even though we barely use it. I appreciate the gift, esp the included ESXi licenses. The hardware supports Nutanix, which the outgoing Cisco rep slyly mentioned. Still need to do a proper evaluation of HCI before jumping in.
Here’s a tip:
Get six new 32GB SD cards and save the ones with installed ESXi with all the plugins and the HX controller VMs. Save the service profiles in your fabric. Install Nutanix on the new SD cards.
That way you can go back and forth fairly quickly. You’ll have to nuke the filesystems of both vendors and revert to blank drives in between switches but you’ll save yourself configuring the same thing over and over.
There’s a new HCI benchmark available for free and supported by all major vendors incl Nuanix, Cisco, HPE, and Vmware (there’s a recent article here). Run the stress tests, load some databases on it and stress those two clusters to the breaking point and beyond. Crash them and watch how they behave when they crash, and how they behave when and after they recover. Run tests for at least 72 hours or much longer if you can (your environment will be up for way longer than that, you will want to know how the caching behaves after it’s been up for a while).
Then decide for yourself which one is better for YOU, dude.
As to your statement about older servers. Those can only be UCS M4, which is Intel Haswell (E5-26xx v4). There are no UCS M3 HX. Nutanix, VxRail only ship Haswell, no Skylake yet. UCS HX ships as M4 (Haswell) and M5 (Skylake).
Believe what you want, but that "2,000+" customer comment most CERTAINLY includes the "for free HyperFlex nodes thrown in with switching deals"! No doubt about it!
Oh, and if you work for a Fortune 20 company and you have convinced your boss to purchase HyperFlex and move your workload to it...both you AND he need to be fired!
It isn't that HyperFlex isn't good kit, but it most certainly ISN'T good enough kit to be moving a lot of production workloads to! Great for the lab/dev workloads, but mission critical ready it most certainly should NOT be picked for!
Oh, and I have customers that are happy to show me their HyperFlex nodes running VMware in their labs...or even as production ESXi nodes in their DC...but not a single one is running HyperFlex!!
Oh, and there are a few customers that really get a positive vibe about their Cisco investment these days when their CEO/etc. discuss how much they are looking forward to being a SOFTWARE ONLY company...warm fuzzies abound!
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
