Not sure if this is a “wife” joke, or legit. Co-location does not require a 10 year lease and electric and cooling, etc etc, are inclusive of whatever fee you have.
Not joking, I thought you were describing your on prem shiz.
I wish we could co-locate. But Teradici just doesn’t perform well enough to be a transparent experience, in addition to the low latency SDI signals we need for grading in the theater and HDR bay. We do have our “proprietary” SDI streaming service down to 14-16 frames, and that is good enough for the in-frequent WFH, but not low enough for day-to-day bulk grading, especially when we have film directors in house.
With our current equipment, most of which is not rack space optimized, we’d need 3-4 full racks.
If planned for space optimization from the beginning, could probably get that down to 2.
I’m really interested to see more all-in price data from cloud though. Apart from the 1 stat that Tom gave, ~9k/month for ~60TB usable of fast/safe storage, no one ever gives their cost.
@Josh_Laurence Willing to share any info?
Cloud pricing is not trivial, there are too many data points to consider to get a realistic projection and AWS cost calculator requires a PhD to operate.
We did hire an AWS engineer to help us with those projections when we started, then we did walk the walk and have refined our operations and cost ever since, changing software even that seemed to bandwidth hungry for example, etc…
In hindsight, all the optimisations are something we should apply to an on-prem solution but I guess the lack of visibility leads to a lack of interest and ultimately we just bundle things up and correct pricing, etc…
Sorry I can’t give you a number but, it is not that easy… check the calculator and see for yourself.
1 Like
I don’t get it. You can’t look at your yearly spend and give us your number? Over and over, no one is ever willing to reveal their spend. Why is it such a secret?
1 Like
The best ‘pricing calculator’ is to collect the bills per day, per week or per hour that other users are paying.
I’m sure it’s possible to be thrifty, but it difficult to calculate the costs of time lost or quality sacrificed by doing so.
Whether you own and operate a computer or Amazon does, the costs of doing so are similar.
Amazon must mark up their operating and storage and internet charges to make a profit.
Further, an internet outage or billing or credential snafu might put you out of commission, which could be a disaster.
The ability to scale up quickly from 5 users to 50 if you are that type of company I guess might be the only advantage I can see over on premises, or not have a ‘premises’ anymore.
Roger Berk
President
Creative Technology
Voice 330 388 6178
That seems oversimplified.
Amazon data centers operate at a different scale and can share a lot of infrastructure. It’s not a 1:1 cost comparison. Amazon can put comparable compute power in a rack for a lot less than any smaller shop can.
Amazon data centers can provide significant multi-site redundancy, well beyond what is reasonable for smaller shops. You are a much more likely to have a major outage on-prem than in the cloud depending on how it is setup.
Lastly, if there’s a global AWS outage, you are more than likely taking a coffee break because many other services you may be reliant on will be impacted and stop you anyway.
That is not to say that cloud infrastructure comes with significant cost, and it’s easy to have a run away meter you didn’t catch. So it’s easier to make expensive mistakes that wouldn’t be possible on-prem.
One thing I’m not seeing here, and I may just be missing something since none of this discussion has any bearing on me, but nowhere have I seen anyone mention that in order to have an on-premises setup like the ones described, you need to employ someone who knows how to set it up, maintain and trouble shoot it. And those people can cost a fair chunk of change. Without that, one tiny crisis can throw all your cost savings out the window.
1 Like
Logik’ers
I have been using a Linux Flame on [www.Simple Cloud.io](http://www.Simple Cloud.io)
It is a flat monthly rate.
Bring your own Flame licence and plugins solution.
It works well and can scale.
I am a freelancer and not working for Simple Cloud.io
The IMB data center is located in Dallas and they have other data centers all over the world.
It is a flat monthly rate with no surprises.
You can also use windows VM’s.
Many awesome features for sharing and colaboration.
email me at jeff.olm@gmail.com and I will connect you.
I am also available for freelance Flame work if you need another artist with a kit.
I have been a Flame artist since 1995. OG an “Original Discreet Logik Gangster”
best,
Jeff Olm
2 Likes
Exactly. You have two competing forces - until you’re a really big shop, the load factor on this person doesn’t justify an FTE. But without that response time and availability suffers, likely extending any issue into hours or days, when cloud outages are often an hour or less.
Back in the 00s when I was at Amazon, the MTR on any tier one outages was 15min for the retail websites.
Of course some of us do our own IT, but that pulls us away from more profitable billable time and hinders customer deadlines.
Two years ago it took an entire week to fix a Teradici outage because one of their techs gave us wrong instructions that made something small worse, not better.
An on-site linux admin is going to be way cheaper than a cloud infra specialist. For on-prem, often once you set it up, things run un-attended for quite some time. As said above, make one mistake with cloud, forget to destroy on VM, and your costs spiral. With on-prem, you know you have fixed costs and capabilities. And let’s just say power or fiber goes out at office, you can’t work anyway. WFH is a whole other issue I don’t believe in too.
1 Like
Does that cloud infra specialist assume you rig your own vs. signing up with any of the resellers that pre-build Flame solutions? Totally makes sense that a custom cloud solution is most expensive. But that can be leveraged over many studios and is more reasonable then. Anytime you can share low load factor resources, you’re better off.
Not really. I learned 10Gb networking, storage, Synology NAS, Linux, Teradici, and Ansible to run 3 Linux Flames by learning from peers, YouTube, and a couple hours of paid support of a freelance sysadmin in a couple of years.
3 Likes
Oversimplified? I don’t think so.
The complexity of running all that Amazon runs is many times more expensive than a Lenovo threadripper, network card, a few nvme drives and an Nvidia card.
They have site security, power, networks and backups and redundancy, as well as scheduling and multitasking over thousands of machines. It is VERY expensive. That’s why they charge many, many times more than it costs you to run your own desktop machine.
Besides the fact that you still need some hardware to access this cloud machine.
The internet outage I was referencing is the internet to YOUR location. A local machine keeps running, maybe with the need to use a slow connection like a cell phone to allow your system to license itself possibly.
The use of Linux systems has always been a mixed blessing, and after much of a professional lifetime wrestling with the vagaries, I would love to have them just appear and just work. But the cost and risks of outsourcing my tools would never appeal to me, even with the supposed ease.
Further, bugs crop up in the Autodesk software all the time. Not having hands on the hardware in times of deadlines and problems would scare me to death.
Roger Berk
President
Creative Technology
Voice 330 388 6178
Back in the day I did an AWS Nimble Studio setup for mainly 3d rendering, the occasional workstation, and about 1.5TB of FSx storage. Nimble was setup mostly default for instance sizing, 2x availability zones, Windows AD services, their standard Deadline config and what not.
The monthly cost to keep just infra and FSx storage, no rendering, no egress, was about 1.2k/month in Aug '22 with the bulk of it being broken down to:
AWS Directory Service 97
Elastic Compute Cloud 182
Amazon FSx 305
Amazon DocumentDB 243
Amazon Virtual Private Cloud 358
You’re not the norm, and “in a couple of years . . . .” doesn’t help the company that needs it now.
No offense, but such specs could be met with a few thousand dollar PC, even 1.5 years ago. You own it, and just turn it off when you don’t need it, or use it as much as you want. Fixed cost, and with such a little amount of gear, HVAC and electric really isn’t an issue.
Also what was the all in cost, not just baseline infrastructure?
None taken - I am 100% in the same camp as you in that the bang/buck ratio is better building out and owning in the colo vs the cloud. I just wanted to put actual numbers out there as to what the AWS ‘out of the box’ offering costs. I only have receipts from the last months before I pulled the plug, but I think in a busy month, all in was 1.2k-ish baseline, and another 2k or so in compute for rendering.
I have a feeling they give you this for free in exchange for shilling, but do you care to share costs and specs?
I tend never to consider any company that doesn’t make pricing public. I also think this company makes ZERO profit and at some point will fold. There was lots of hype of few years ago for cloud studios and VC money was being thrown around with abondon.
Happy to book a demo and have the sales guys give you the fixed pricing.