Virtual Iron

» see virtualiron.com as it lived, on the wayback machine

~ the eulogy ~

Oracle's press release is dated 13 May 2009 and still sits in the archive under the headline "Oracle Buys Virtual Iron." It says that Oracle has agreed to acquire a provider of server virtualization management software, that customers can expect better capacity utilization and improved control of their enterprise software, and that the transaction is subject to customary closing conditions and expected to close this summer. Financial details were not disclosed. Until the deal closes, it says, each company will continue to operate independently.

By the middle of June the partners who sold Virtual Iron had letters telling them that development of the existing products was suspended and that orders from new customers would not be taken. The technology was to be folded into Oracle VM, and the last day a reseller could sell a new license was 30 June 2009. Support for version 4.4 ran out on 3 September 2009 and for 4.5 on 14 January 2010, and Oracle offered the installed base what it called lifetime support, a phrase one partner described at the time as a joke.

The numbers behind the decision were not a secret. Virtual Iron had taken in $3.4 million of revenue in the year before the sale against $17.7 million of spending, and its real asset was the management layer that Oracle wanted bolted onto its own hypervisor. Roughly ten to fifteen people were kept. The domain now redirects to Oracle's virtualization page.

~ the dossier ~

~ before the name ~

It started in March 2003 as Katana Technology, founded by Scott Davis and Alex Vasilevsky, and what it built first was a much stranger idea than the one it died selling. VFe took a rack of ordinary x86 servers and presented them to an operating system as a single large machine, gluing the memory together over InfiniBand across as many as sixteen sockets, with its own hypervisor underneath and up to 128 partitions per cluster. The goal was to make many servers look like one, which is the reverse of the thing virtualization came to mean.

InfiniBand never arrived in the volumes that bet required, and a hypervisor of your own invention is a hard thing to sell to buyers who have standardized on somebody else's. The company renamed itself Virtual Iron in January 2005 and abandoned the clustering product. What it kept was the management software it had written for that architecture, which it rebuilt on top of Xen. The second plan is the one that got funded: $8.5 million in September 2005 in a round led by Intel Capital, on top of about $20 million already in from 2003 and 2004.

~ thirty seven days ~

Count from the press release to the partner letters and you get thirty seven days. Oracle told the market the deal would close over the summer and had stopped the product before the summer began. The shape of this is routine enough: Oracle's own announcement named capacity management and power management as the things it wanted for Oracle VM, and those were features rather than a business.

The speed is worth a page because of the size of what got switched off. Something like fifteen hundred resellers were carrying the product, and a former employee's estimate of the installed base ran into the low thousands of companies, many of which had bought licenses in blocks of five hundred or a thousand. Those companies woke up inside an Oracle migration they had not chosen, on a product line that had been alive and selling when they signed.

~ the cheap seat ~

The pitch was price. Enterprise virtualization in 2007 and 2008 meant VMware, and VMware charged accordingly; Virtual Iron came in at a few hundred dollars per socket against several thousand, with the Xen hypervisor underneath doing the part nobody was charging for. That was a real argument, and the people making it understood the market they were undercutting. Alex Vasilevsky, who co-founded the company and served as its chief technology officer, is remembered by the Xen project itself as one of the people who got the hypervisor into commercial use.

Enterprise software of this kind never reached me and was never sold in my direction. You did not download Virtual Iron; a reseller in your city quoted it to your IT manager, and the IT manager either believed that a startup in Lowell, Massachusetts would still be there in three years or did not. The ones who believed it were correct for about four years, which in this business is not actually a bad return.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »