The Nirvanix Failure: Villains, Heroes and Lessons

It's been instructive to watch Nirvanix's hasty -- and to my mind unnecessarily messy -- exit from the cloud storage market. As with any other saga, the Nirvanix story has villains, heroes, innocent (and not-so-innocent) victims, and lessons learned.

I saw more than one commentator on Twitter ask what the big deal was -- after all, vendors go out of business all the time. The difference is that when your array vendor goes belly up, you still have the array; when your public cloud vendor closes, you lose access to your data.

Vendors such as Iron Mountain have closed their cloud storage services, but they've given customers plenty of notice so they could find replacement services and move their data. Nirvanix gave customers less than a month's notice before shutting down.

While some observers have vilified the venture capitalists who backed, and then abandoned, Nirvanix, I don't see them as the ultimate villains in the story. When a VC gives an entrepreneur five million dollars for 30% of his new company and first-born child, the VC isn't committed to buy into future rounds or to swoop in and save a struggling business. Venture capitalists are gamblers, and Nirvanix turned out to be a bad bet.

[How are your cloud services performing? One startup aims to tell you. Find out in "ThousandEyes Peers Into Cloud Performance." ]

This is not to say the VCs are blameless. Nirvanix went through five CEOs in six years; the revolving door in the corner office was due in no small part to the VCs, as changes in top management were reportedly a condition of some firms' participation in a funding round.

The real villain is Nirvanix's final CEO, Debra Chrapaty. Nirvanix had many customers with over a petabyte of data. Those customers deserved more notice than they received. They, and the rest of the cloud computing community, also deserved better communications.

Sure, customers got a direct notice, but Nirvanix made no public announcement for approximately ten days after rumors of its demise appeared in the Wall Street Journal. I find it hard to believe that a competent management team wouldn't have seen the end coming more than 90 days out. Chrapaty and company either didn't know, or decided to keep the end a secret.

Minor villains include the opportunists who flooded my inbox with releases announcing they were offering a month's free service or unspecified migration assistance to Nirvanix customers, or who bought the Google Adword 'Nirvanix' so anyone searching for that word would see their "free month" offer.

Heroes include Aorta Cloud CEO Steve Ampleford and the remaining engineering staff at Nirvanix who managed to get the data for every customer that contacted them migrated.

The migration happened over a high-speed link in the shared datacenters where Nirvanix housed its gear, or via physical media to a new OpenStack Swift-based infrastructure. Even better, Steve and his team worked with all the major cloud gateway vendors to allow users to use the new infrastructure without reloading their appliances' disk caches.

Lessons Learned

The primary lesson to learn is that you, as your organization's data custodian, can outsource the actual data to a cloud provider. But you, not the cloud provider, are ultimately responsible for your organization's data. Sure you can have a contract that commits the provider to all sorts of SLAs, but that contract isn't worth the paper it's written on if the provider goes belly up.

All a contract does is give you the grounds to sue the other party if they don't meet their obligations. When the other party can't meet their obligations and has no money to sue for, the contract is worthless.

Another lesson is that it may be safest to store your data with the big boys. Google, Amazon, Microsoft and even telco-affiliated cloud providers like Savvis and Terremark have other businesses to not only support their cloud storage offerings, but also provide the resources to make your SLAs actually enforceable.

Vendors that close down individual services have to protect their reputations for their other offerings, and have historically provided enough wind-down time to avoid the mad scramble Nirvanix customers are only now recovering from.

Whether you store your data with a big boy or a startup, the final lesson is to keep an emergency backup copy of all your data under your control. This is where tape can play a role. For example, some vendors combine a tape system with an S3-like API. Applications, or gateways, could send a second copy to tape as they write data to the cloud.

It's never fun to have an important partner disappear. If you plan ahead you can make an event like the Nirvanix collapse an annoyance rather than a business-disrupting (or career-ending) saga.