Troy Brooks

Scalar Decisions Senior Product Manager Troy Brooks on Leveraging Subscription Licenses and Cloud Computing

Few pundits in the business know more about VFX pipelines than Troy Brooks. A former head of studio at Digital Domain in Vancouver, Brooks was for more than seven years the CEO of render-farm management software maker Pipeline FX as well as a CG supervisor at Electronic Arts and a software developer at Mainframe Entertainment (now Rainmaker). Today, he is senior product manager at StudioCloud – Scalar Decisions in Vancouver, which provides private cloud services for digital media. He advocates for VFX studios to take control of their own destinies by analyzing data provided by software like PipelineFX's Qube! and using it to make better, more efficient use of cloud services.

Brooks told us that the arrival of subscription-based software licensing models, when used with public and private cloud infrastructure, can give savvy facilities an advantage when it comes to managing cash flow and bidding on projects.

Here, presented in Brooks' own words based on an edited conversation with StudioDaily, are 10 ways you can use your own data to plot out a more efficient and sustaining course for your VFX business.


Count your rocks, your pebbles, and your sand.
You need to take the data you've collected over time in your render-farm management software like Qube!, look at the historical patterns, and make real decisions. I call it the rocks, pebbles, sand analogy. First you have a jar filled with rocks, and then you put the pebbles in, and the pebbles fit in around the rocks, and then the sand fits in around the rocks and pebbles. The rocks are your base-level utilization that will be consistent, even as your production peaks and valleys go up and down. The pebbles are capacity you can get through something like [Scalar] StudioCloud. It's private, closely held and secure, but can handle production peaks for up to three to six months at a reasonable price. And the sand is the public cloud, the Google model. You may need to go very wide over the short term — maybe to 500 machines for a trailer delivery over the weekend. It would be prohibitively expensive to do that for more than a couple of days, but that's what you need to turn the job around.

Keep base costs down for hardware.
Any studio has a baseline level of activity. Even when you've ramped down between shows, you're still holding on to key supervisors and senior staff. If you look through your historical data, you can figure out your baseline of this many licenses and this many workstations. There is still no more cost-effective way to get those than to buy them, or get them on a long-term capital lease. Those are our rocks, and then you figure out your levels for pebbles and sand. The problem is studios are always looking to get bigger, and when they get offered a big show, they think, "Oh, man, we can't afford to turn this down." They become a lot bigger, and their base level becomes bigger. But if they can do that as burstable capacity, with shorter leases or rentals, it's a safer bet than making the capital purchase. It helps them avoid falling into the trap — you buy gear, you can't keep your gear busy, you can't keep up on your payments, and you run into cash-flow issues.

Leverage subscription models for software.
The data you collect will allow you to make educated projections about what you're going to need to do through the course of a project and budget for those different levels of integration. You may need to burst up from a core staff of 50 people to 150 people for a show. You need to evaluate how many permanent licenses you want to own. Maybe you want to go all subscription, and be month by month on everything. If you can drop your costs to virtually zero when the show is over, that's a huge benefit to you. 

Allocate costs on a show-by-show basis.
This is one of the unique benefits of a render-farm management system that has a relational database at its heart. You can track how many hours of rendering you did on a show because that metadata is tracked. You can see which licenses you used. If you paid for them on an as-you-go basis — meaning you didn't buy them as a capital expense that you're responsible for depreciating — you can bill them out to that show, and you build your bids around that model. This is really what studios have been trying to do. It's a combination of having that data available in your render farm management software's relational database, having a variable cost model by going to subscription, burstable capacity and clouds and being able to bill those costs out specifically from show to show makes a huge difference. And there are tax implications from not having to make those capital acquisitions that vary a lot from location to location.

Know where you're spending your money. 
It's commonplace to see VFX studios go under. It's always because of cash flow. If you have too much capital acquisition, you can't pay your bills, and you can't be profitable. The model that the whole bidding process is built around is very aggressive. One studio executive facetiously said some years ago, "If I don't put at least one [VFX] studio out of business during a show, I'm not doing my job." That anonymous person probably wishes he hadn't said that. But there's a certain truth to it. Six big studios hold virtually all the work, making it an extremely competitive business. It's important to have a very fine-grained analysis of where your money is going on a particular show. If you see that you did 68 revisions of a single shot on a show in one department, you need to find out who was on those shots and why you did that. Maybe it's a hero shot or a tentpole shot and the VFX supe pixel-f—ed it to death. That happens. But you really need to know where you're spending your money.

Analyze your processes.
When you have a capital acquisition model where you bought the licenses and you're working on a show with a certain number of machines and a certain number of licenses, it leads to certain kinds of behavior. If the farm's not busy, why not throw another shot onto the farm? It's a sunk cost for the machine and the license. The problem is, it engenders bad habits and you get accustomed to it. But if every iteration you do is measured and counts against the budget for your show, or if you're literally paying money for it on Google Cloud or on a private cloud like Scalar, it engenders a different set of behavior. If you see where you're spending your money at the end of every show, it can lead to some very subtle analyses you do of your pipeline — what kind of renderer and storage you're using, and the loads on your storage as the farm draws off it. More immediately, you can say, "We're doing 50 to 60 iterations of every shot in lighting, and that costs us a lot of money. We need to look at why we can't do single-frame tests and wedge that out more carefully." Where is the money being spent, and what can we do to optimize it? It's like any kind of supply-chain analysis in what is essentially a manufacturing process.

Recover some of your capital costs.
I gave a talk at a VES meeting at the old ILM six or seven years ago about the notion of rendering in "the cloud." There were a lot of issues back then with the cost of bandwidth and security and lots of other things. My point at the time was that the business model of the various software publishers was the single biggest hurdle, because they did not allow you to do burstable rendering or use their licenses on remote locations. Built into the end user license agreement is a notation that you have to use those licenses, usually, within five kilometers or three miles of the physical address that is put in the deal. The notion of being able to use your licenses or rent them to someone else was unheard of. Fast-forward to today, and most of the publishers are getting on board with it. With the rise of the private cloud, some studios will buy or take out a long-term lease on 300 machines, put them in our Studio Cloud and, when they're not busy, they rent them out to other members of the Studio Cloud. You have a nice, diverse ecosystem of game studios, stereo conversion studios, animation studios, and television studios with different schedules and different peaks and valleys. Multiple studios here in Vancouver have recovered into the high six figures in their lease costs by renting out a bunch of their machines. It's real money. When you can do that, you're really improving your cash-flow position. But you need to have the historical data to know when you're going to need that gear back.

Demand even more flexible software payment models.
The big change over the last couple of years is the move toward subscription-based pricing. Adobe took the lead, but most of the publishers have followed suit. I'm trying to move software publishers toward metered, day-by-day plans. Right now you can do month by month or quarterly or yearly subscriptions, but if you can do different levels of analysis you can figure out that you need a certain number of licenses on a yearly basis, others on a quarterly basis, others month by month, and then even more on a metered basis.

Never assume you're too big to fail.
When I started at Digital Domain in 2009, I had a lot of experience dealing with different studios, how they ran their pipelines, and how they did resource management. I said, "OK, so you do a bid on a show, you do the show, and then you analyze what you did and feed that information back into your next bid." And their response was, "Yeah, that would be good if we did that." This is one of the top VFX studios in Hollywood and, at that time, they still didn't do that. There are a whole bunch of reasons why that might be, but that's a fundamental aspect of business: making sure your bids are accurate so that you're able to make money.

Remember that you get jobs based on your skills and your contacts, not just the amount of capital you can invest.
Right now, there is an opportunity for people to do work based on the merits of the people they have, rather than the capital budgets they can bring. The single biggest determinant of your success is whether studios have confidence that you'll be able to deliver. You need marquee players in your studio with a track record. I wouldn't say we're there yet, but I wouldn't be surprised if we saw a proliferation of smaller studios that can act like larger studios. You just need people and a space where you can put them. You would need a short-term lease, because five years is a substantial commitment. If you bring workstations into that space, that puts a significant load on the heating and power infrasturcture. And, on top of that, a machine room? It's an expensive proposition. But if you could move into a building already wired with dark fibre to a facility like StudioCloud, you could use it as a central point of presence in whatever tax-incentivized city you choose, with workstations, storage and rendering licenses all through remote facilities. All you need is thin clients and multiple 10 Gigabit connections to a facility that would rent you burstable capacity. That, to my mind, will be the game changer.

— Troy Brooks