Exclusive + analysis Cisco has confirmed that the Invicta all-flash array line is no more, telling The Register: "As part of product lifecycle management, we withdraw technology from the marketplace when necessary to focus our efforts on what is critical for the future of our customers' business as well as our own."
The networking goliath added:
Cisco is prioritizing the elements of our portfolio to drive the most value for our customers both now and in the future, and today, we are announcing the End of Life (EoL) for the Invicta Appliance and Scaling System products. We will continue to support existing customers who have deployed Invicta products in accordance with our Products and Services End of Life Policy, which includes ongoing technical assistance, software support, and spare / replacement parts.
When Cisco entered the server business with its UCS line it had high hopes, and they came good. Paul Perez, Mr UCS inside Cisco, had driven the UCS business to near $3bn annual run rate but was hungry for more. We're told Cisco realized the UCS business could be boosted by adding storage to its serving capabilities.
It partnered with NetApp to develop an all-flash storage blade with full UCS M integration. The project ran into rough waters and Perez decided to buy in technology instead of developing it. Cisco evaluated several all-flash companies and decided on Whippany, New Jersey-based Whiptail as the optimum combination of performance and acquisition cost. It bought it in September, 2013.
More ReadingZorawar Biri Singh lands Cisco CTO post as firm shuffles its top brassSoftware-defined what? Look at our glorious ASICs says CiscoThe green salamander is OUT: Cisco gives up on Invicta flash arraysCisco quietly slips out new all-flash arrays – perhaps it doesn't want EMC, NetApp to noticeWatch out EMC and vendor pals, Cisco's set to flog you with Whiptail
They would get: Whiptail's two arrays, a single chassis Invicta appliance and a scalable Accela system; its engineering team; and its roadmap. It would also get the capability to finish development on the NetApp-Cisco storage blade hardware.
Perez had Whiptail focus on bringing its Fibre Channel-connected appliance and scaling systems to market using UCS hardware with, we're told, "a very aggressive roadmap." The Whiptail engineering team had to cancel several interesting projects after joining Cisco to concentrate on these new goals.
Culture-wise, there was a "them and us" attitude, with the Whiptailers not feeling appreciated or integrated right from the start of Cisco's ownership.
They arranged the new storage blade's specifications with their Cisco management and agreed that, because of the limitations of the inherited Cisco/NetApp hardware, it would not be able to support deduplication.
Invicta UCS field issues
Then an unwelcome eventuality came along: there were problems with Whiptail, now renamed Invicta, storage products when operating in the UCS environment. But Cisco was impatient to get the appliance and scalable arrays released. We have been told that the pressure from Perez "was quite enormous and eventually we became officially one of the fastest-ever products integrated by Cisco," with Invicta launching in January 2014.
It was all too quick, as the Invicta-UCS issues surfaced in the field, and Invicta's reputation with Cisco's UCS sales force deteriorated badly. As we now know, the scaling system was put on shipping hold while the problems were being resolved. We're told that Invicta appliance sales targets were deeply lowered as well.
It was a setback but only a temporary one. Spirits were high. New offices were being prepared for the Invicta engineers a few miles from their existing Whippany, New York, location. They were working on fixing the scaling system and continuing development of the storage blade, based on the UCS server hardware.
Candelaria calamity and Perez exit
Then Whiptail founder James Candelaria suffered a heart attack in October last year and had to step down. That departure was devastating for the firm, sources told The Register. He was regarded as a visionary and had done lots of research and development work. His loss hit the engineering team hard. They had to cope with an unwillingness by Cisco to add custom modifications to the storage blade hardware. It could have SSDs attached to it at a basic level but nothing beyond that.
Candelaria had tried to get the Cisco management to take a more creative approach in order for the developing product to remain competitive but it was, we're told, an uphill struggle. At least Perez had some technical vision and was committed to Invicta's success, having been instrumental in buying Whiptail, but when he defected to Dell in March this year it turned out to be the killer blow.
The team now had no advocate within Cisco. Its new manager was Satinder Sethi, a veep of product management and solutions, and the team did not get on with him. A source told The Register that they viewed him as their "worst enemy," Sethi having been opposed to the Whiptail acquisition in the first place. They began feeling they were in a bad place with a noose gradually tightening around their necks as:
- Construction of the new building was put on hold
- Conditions for meeting the shipping hold on the scaling system were tightened
- When these conditions were met, the shipping hold was still not lifted
- The storage blade project, 90 per cent ready, was put on hold because the missing deduplication was now deemed to be crucial
By May, the team felt they were on the way out, that all of Perez's traces in the UCS division were being eradicated by EVP and chief development officer Pankaj Patel, Satinder Sethi, and others. The storage blade had been effectively cancelled for political reasons, with the team arguing that Cisco could have sold truckloads of it even without deduplication. No one wanted to hear this.
The Register has asked Cisco for comment on these specific allegations and will update when we hear back from them.
This month the team were mostly laid off, apart from a 20-man skeleton support remnant at the Whippany office. They felt they had been given whipped tails and the manner of its doing left a sour taste in their mouths.
Lessons to learn? The storage vulture's musings
What can we glean from this sorry tale?
- Cisco's famed acquisition machine fell apart with Whiptail
- Cisco's UCS organization wasn't fully committed to the acquisition
- The Whiptail team wasn't integrated well enough into Cisco
- Hindsight says Cisco's technology due diligence on Whiptail's tech and its UCS suitability was incomplete
- Leaving out deduplication originally was a mistake
- Storage blade development was crippled by the NetApp/Cisco co-developed hardware and its inadequacies, and Cisco's unwillingness to agree on custom HW fixes
- The departures of Candelaria and Perez left the team without any advocate inside Cisco
- The Invicta/UCS problems gave ammunition to the anti-Invicta people
It seems to El Reg that Invicta's failure was fundamentally due to failures by Cisco management. These failures have left it without any presence in the flash storage market at a time when server-based hyper-converged systems and server SANs are of growing importance, with networked arrays facing sustained decline over the next five to ten years.
It has a converged system and hyper-converged system market presence through partnerships, but obviously leaves revenue on the table for these partners. It is at their mercy regarding technology developments and unable to have its own unique tech as its partners partner elsewhere.
It's now wasted two years of development time and, we're told, more than $415m on the engineering costs since the acquisition. Its competitors have gained a two year head start, and if Cisco wants to add storage heft to its UCS systems so it can succeed in the converged/hyper-converged systems market, then it has to buy its way in again.
That means the people who have killed Invicta need to see the light and persuade Chuck Robbins and his exec team to spend a billion or more buying Nutanix, Simplivity, Tegile, or some other firm which is in, or headed towards, the hyper-converged market. Pure Storage, anyone?
Alternatively, they could buy a tier-two or -three startup in the space more cheaply, but then they would have a longer development time. Pay top dollar for a top player, guys; you screwed it up before. Don't screw it up again. ®