I'd characterize it more as putting some 'spin' on the story.
I find it interesting that the one important thing
he continues to refuse to comment on is the number of users
(average or simultaneous) OnLive actually has
hosted. Which is disingenuous at best considering that a promised zero-downtime and the virtually unlimited number of users they can support
is pretty much the raison d'être
for OnLive as a business. Otherwise, it's no different than any other MMORG host.
As was noted by the interviewer:
if the load balancing is such a huge task, how many concurrent users is OnLive equipped to serve? Perlman wouldn't comment on the number of simultaneous users OnLive has seen thus far (neither the most nor an average),
All Perlman offers in response is what amounts to a marketing pitch:
[OnLive] was architected from the ground-up to be completely real-time and adaptive, no matter what it is hosting, no matter what network it is using, and no matter what device it is connected to. This is largely implemented in real-time Linux code on specialized, ultra-low latency processors. The number of Linux sessions varies greatly depending on what is needed to support the experience being delivered, depending on the network, device, location, application, etc.
The article goes on:
Perlman said this type of setup doesn't compare to any existing online system: "There is no simple way to characterize all of the computing resources needed to manage this," he wrote. He noted a time when one of the company's ColoSpaces had a two-day power outage due to generator failure. Other customers of that ColoSpace were completely shut out, but OnLive rerouted its users to sister data centers without interruption so that they could resume their activity seamlessly.
"In fact, people spectating them around the world were able to resume spectating within seconds," Perlman noted. "We did not receive a single customer service ticket (e.g. our compression algorithms have evolved to the point where the latency to a distant data center is less than it was to a nearby data center at launch). There is no simple way to compare this to any existing online system and no simple way to characterize all of the computing resources needed to manage this."
Nothing of substance there. All it says is that their infrastructure is amazing. No metrics to support that assertion. Just an assurance (followed by some war stories) typical of what might be given to institutional investors.
Dunno. Makes me wonder when a tech company like this suddenly starts hand-waving and refusing to talk tech.
Maybe I'm disappointed because I was interested in OnLive, not so much as a gamer, but more as a Linux tech observer and a network integrator. And so far I haven't heard anything that tells me much other than somebody somewhere got blindsided by something. But exactly what that was Perlman has yet to come forward with. I don't buy the load-balancing bit. That's a common element in network design. You build and test for it. It's something that should have been anticipated and planned for.
Maybe it's me, but I still feel something important is deliberately being left out of what's being said.