Like so much it takes me a bit to go through all the possible meanings. This is especially true as it is almost always spelled “baited” breath. I tend to think of a person with breath that smells like dead fish.
Is anyone else not liking the sound of the “Fast Lane”? I don’t find it plausible that “supercomputers” were purchased and are being reserved for people willing to pay more. Does this mean they’ll remain idle when premium-paying users don’t need them? That would be very surprising from a systems administration standpoint.
Maybe premium users will be serviced using the same server-side hardware and algorithms as the rest of us, but their jobs will just get pushed to the front of the queue.
A more welcome announcement would be, “Hey we just added some faster computers to our processing pool, and now everyone’s prints will go 20% faster.”
How I interpret this announcement is, “Hey, we’ve decided to save money by only upgrading our processing capacity for a very limited set of users willing to pay more. Everyone else will just have to suffer with the same server technology we’ve been using for the past three years.”
Glowforge print processing is handled by Google Cloud. Glowforge pays for compute services based on performance like any other customer. They can choose to forward Premium customers’ jobs to resources with higher performance, for a higher fee.
My understanding is that there is a separate section of Google cloud services with faster servers. They cost more to run, so that’s part of the subscription price we’ll be paying.
Shouldn’t impact anyone who chooses not to subscribe. It’s just a faster option for those doing calculation heavy prints. My prints generally don’t need it. For someone running intricate prints with a lot of nodes, it will save a lot of waiting time while the job is processed.
Since they’re using Google Cloud offerings I don’t expect they own any servers. They’ve probably contracted for a certain machine profile and as demand ramps up, Google spins up a new virtual machine matching that profile. They likely have throughout parameters that trigger automatic server provisioning and when demand drops, Google drops instances.
The new “supercomputer” capacity is likely just a new higher performance server profile that will get instances spun up or down based on demand.
With typical cloud implementations the customer only pays for compute used.
Extraordinarily unlikely. Why would they lie about that?
Here’s the wholesale pricing, you can see for yourself: https://cloud.google.com/compute/all-pricing. Note the higher-end machines are orders of magnitude more expensive. And Glowforge probably uses some specialized GPU processing, which gets even more costly.
Part of the whole point of cloud services is that you (as in Glowforge) can spin up and spin down servers very quickly to accommodate demand. So most likely they’re not paying for idle machines, and it’s not a big deal from a system administration standpoint, because that’s exactly how modern software is deployed.