X.Org/FreeDesktop.org Is Looking For Sponsors Or May Have To Cut Continuous Integration Hosting

[starshipeleven]

Seems like capitalism to me.

Contrary to popular opinion, capitalism does not imply scamming customers.
Scamming people out of their money is a completely parallel thing to the economic structure and is prevalent in communism and socialism too.

[Anty]

GCloud/AWS/Azure are all scams, unfortunately. They promise you the world, give you estimated bills, then you realise they lied with a straight face, and hide behind enough technicalities and small print that it's still cheaper to just be extorted.

Dude - don't spread FUD.

First you can use cost calculator like this one in AWS https://calculator.s3.amazonaws.com/index.html or GCP https://cloud.google.com/products/calculator.

It is well known that you pay for egress in cloud. It happens for both AWS and GCP cost is similar - $89 and $85 per TB/month.

So in case of GCP $75000/year gives 882 TB/year -> 73 GB/month -> 2.4 TB/day -> 10 GB/hour

And out of $75000 only $30000 is egress cost (remaining is running other stuff).
So final number is just 4 GB/hour.

Seems high? 4 GB/hour is just 1.1 MB/s which is not impressive at all...

[kpedersen]

A solution could be for Xorg to separate from the extra weight like FreeDesktop / Wayland. That will grealy reduce maintenance burden. Then Xorg can live on just like before.

Let those cool kids fend for themselves.

[intelfx]

I expect the general population to screw up when it comes to understanding the costs of 'Putting It In The Cloud!' vs. just owning your own equipment. I didn't expect it here, especially as multiple people must have known about this.

The costs of 'Putting It In The Cloud!' is paying more money in exchange for fewer problems.

Besides, owning your own equipment doesn't seem all that much cheaper[1], considering that you also need dedicated admins:

I took a quick look at HPE, which we previously used for bare metal, and it looks like we'd be spending $25-50k (depending on how much storage you want to provision, how much room you want to leave to provision more storage later, how much you care about backups) to run a similar level of service so that'd put a bit of a dint in your year-one budget. The bare-metal hosting providers also add up to more expensive than you might think, again especially if you want either redundancy or just backups.

[1]: https://lists.freedesktop.org/archiv...ry/041245.html

[Guest]

The costs of 'Putting It In The Cloud!' is paying more money in exchange for fewer problems.

Besides, owning your own equipment doesn't seem all that much cheaper[1], considering that you also need dedicated admins:

[1]: https://lists.freedesktop.org/archiv...ry/041245.html

That plus you don't have to worry about where you keep the servers. Rent/ownership of physical space is expensive. Whereas data centers are usually built on cheap land, and the cost is spread out among a huge number of consumers.

[Naquatis]

freedesktop.org is a completely volunteer organisation with no corporate backing or funding stream.

I mean even if I like to spend something for my desktop .. where is the donation button?

Give us some code projects like amdgpu mesa opencl 1.2 to pay for.

[starshipeleven]

A solution could be for Xorg to separate from the extra weight like FreeDesktop / Wayland. That will grealy reduce maintenance burden. Then Xorg can live on just like before.

Let those cool kids fend for themselves.

This article isn't about Xorg maintenance burden at all, but about the organization that maintains the infrastructure used by all projects in FreeDesktop.org umbrella.

go trolling somehwere else

[starshipeleven]

Dude - don't spread FUD.

First you can use cost calculator like this one in AWS https://calculator.s3.amazonaws.com/index.html or GCP https://cloud.google.com/products/calculator.

It is well known that you pay for egress in cloud. It happens for both AWS and GCP cost is similar - $89 and $85 per TB/month.

So in case of GCP $75000/year gives 882 TB/year -> 73 GB/month -> 2.4 TB/day -> 10 GB/hour

And out of $75000 only $30000 is egress cost (remaining is running other stuff).
So final number is just 4 GB/hour.

Seems high? 4 GB/hour is just 1.1 MB/s which is not impressive at all...

this is just ONE of the possible parameters. CPU power, RAM and availability in datacenters spread around the world are other parameters.

I'd like to point out that he mentioned Heitzner, and that's the provider of OpenWrt build servers. Download servers are cheap and you don't need the cloud in most cases.

[allquixotic]

The cost of the big cloud providers is really high and unnecessary, especially with bandwidth costs. They should build their own cluster on OVH infrastructure. Bandwidth is unmetered and billed by how much capacity you want.

I actually struggled to comprehend how you could spend $70k in one year on GCP just on bandwidth egress. It looks like you have to transfer 100 TB per month (!!!) to reach that level of expense. I priced out 10 N2 standard instances and 100 TB per month and it turned out around $7k USD per month.

Here is what they could get for $2500 per month ($30k USD per year; less than half their bandwidth bill) on OVH:

20 servers (each $120/month) each with:

• 2 x Intel Xeon E5-2660 (8-core)
• 256 GB DDR3
• 2 x 240 GB SSD + 4 x 2 TB HDD
• 500 Mbps unmetered Internet traffic

That's a total of 320 Sandy Bridge cores, 5.1 TB of RAM, 9.6 TB of SSD space, 160 TB of HDD space, and 10 Gbps unmetered bandwidth. If they need fast additional disk space, they can go with Wasabi object storage for $6 per TB-month, with no fees for API calls or data transfer. If you fully saturated the bandwidth, you could transfer 3.2 PB per month, which is 32 times more than the 100 TB ballpark they seem to be at now. So, even though the uplink is shared, I'm certain they'll be able to exceed 100 TB per month with this infrastructure. They could hit 1 PB (1000 TB) per month, spread across these servers, with no problem.

$70k for data transfer? Are you out of your mind?

[starshipeleven]

$70k for data transfer? Are you out of your mind?

it's not bandwith but CPU power and system resources to run the Continuous Integration build bots and automated testing infrastructure