X.Org says it's saving a packet with Packet after migrating freedesktop.org off Google Kubernetes Engine

The X.Org Foundation has successfully completed a migration from Google Kubernetes Engine (GKE) to Packet, which it reckoned "should save us around $30 per day." The X.Org Foundation manages a number of key open-source projects including the Wayland graphics protocol, the X.Org server, and the Mesa 3D graphics library. The …

[Peter2]

So...

The price given by the nice helpful salesman is actually well under the figure charged by the company when they have decommissioned their existing infrastructure and this price continues to continually rise despite drastic efforts to reduce the cost?

I'm shocked I tell you, shocked.

[Pascal Monett]

I think this is a very good cautionary tale, which I will point some of my over-enthusiastic customers to.

They will likely brush it under the rug, but in a year or two, I'll be able to remind them that they had been warned.

[b0llchit]

Unfortunately, the MBA-types you'll be telling this story to are notoriously impervious to any cautionary tale because those MBA-types cannot fathom that they might be wrong. Such a thing, being wrong, simply does not compute for this type of person.

And then, when they realize, that the costs are too high and you say "I told you so", you will become the guilty party because you should not have created this problem in the first place. Yes, there can only be others to take the fall.

The MBA-type cannot be wrong so you techie always take the blame. I told you so!

[martyn.hare]

Reminds me of someone...

Who “wanted to complain” because his ERP system would no longer spit out reports. Apparently it is the fault of IT that we didn’t inform him that Flash was going EOL despite Chrome, Edge and the Adobe Flash updater itself all popping up to inform him of this very real, looming issue on a very regular basis. When it was also pointed out that IT doesn’t support the ERP system components (the ISV does) and that IT had no reason to be aware of the external software dependencies, said person still wanted to complain about “poor service” from IT and we all couldn’t help but point him at someone with a fancier director title to moan at instead.

MBA types are cancer.

[Anonymous Coward]

MBA-types

Never a truer word spoken! This sounds remarkably like a friend of mine who is extremely proud of his MBA and will be sure to mention it at every opportunity

[Anonymous Coward]

Cloud Costs

I'm often asked why we use our own hardware/locations around the globe rather than AWS and other public clouds.

People just don't understand the bandwidth costs involved - Google's $0.12/GB is a bit higher than AWS (depending where in the world you're shifting bits from), but is much, much, much higher than the $0.001/GB we pay in some locations (let alone the locations where we've got free connectivity).

Given we're a bit-shifter at heart, that difference adds up quickly and it's a cost that would have to be passed onto customers (at which point, how do you compete with those cloud provider's own offerings on price?)

It's a particularly nasty cost too because it's unpredictable and driven by the behaviour of others. If you have 100 EC2 instances you know the compute costs, but your bandwidth costs are going to depend on user behaviour. Got hit by a whack of junk traffic? That's still GB's out.

You get similar with backing up to S3/Glacier. It's good you've got a backup, and definitely better than no backup, but the cost of pulling that backup down to restore it might be considerable (depending on size)

Public cloud, obviously, has it's place - there are workloads that it's well suited for, and businesses who's needs are a good fit, but it's very, very far from a panacea

[jeffty]

Re: Cloud Costs

Customers don't seem to get a lot of these hidden costs, because they've never had to think about them previously, especially from a network perspective. In the on-prem world, they'd pay for a nice shiny Cisco/Arista/HPE fabric in their DC (which hits them smack in the wallet for refreshes, support and maintenance), but they can hammer traffic across it all day long.

Noone cares about the utilisation as long as there's no congestion, no bottlenecks, and noone suffering slow responses to their application. The only monitoring is typically for links going down or physical failures. There's very rarely any sort of capacity/impact analysis on new infrastructure or apps - 25/40/100gb uplinks on the LAN and gb uplinks on the WAN mean they don't have to think about it and if it proves a problem, all they do is increase the bearer size, or add more uplinks or more LAN internally to meet the demand.

Those costs and considerations don't go away in a public cloud, the cloud builders just incorporate them into bandwidth charges as the cost of maintaining their internal network fabrics. All of those data transfers you took for granted (image pulls/pushes, updates, backups, migrations) now all have a dollar cost associated with the data you intend to shift about, even if it's just internal and not leaving your cloud instance. Before it was all you can eat, now it's definitely pay as you take a bite.

I've only come across one organisation that took into account the impact on WAN bandwidth usage (nationally) that changes to their applications in the DC would impose, and even then it wasn't to anywhere near the level of scrutiny to work out the increase in bytes/kb per transaction within the DC (and % or gb increases across the board) that would be needed to derive the costs of running this stuff in the cloud. If you've done the modelling to anticipate this kind of cost - awesome, but you're in the (sensible) minority.

[bombastic bob]

Re: Cloud Costs

I've been looking at another ISP for my company domain again. The accounts are being migrated and I would lose a couple of features I want to keep. Fortunately, a comparable service [to the existing capabilities] exists.

In the mean time I've had to look at a lot of hosting options and pricing.

In short, AWS, Azure, and Google are *NOT* the only game in town. Many lesser expensive alternatives exist. And it may benefit them to combine their cloudy services with a hybrid shared/dedicated hosting platform for admin things. Based on what I saw in the article it may be the "adinistrative" crunching that's doing them in.

(you know how managers are, "what's the average man hour for XXX type of YYY and can I get that in my e-mail every monday morning, summarized and totaled?" - multiply that by a dozen or so, and CLOUD SERVICE STICKER SHOCK ensues)

[Justthefacts]

Are the numbers correct?

At the beginning of the article it talks about saving $30 per day.

But later on it becomes over $3k per month, which is 3x. I think it’s the higher number that is correct.

[sreynolds]

Re: Are the numbers correct?

Yeah I laugh when meetings and meeting are held for trivial amounts, that were you to crunch the numbers you would find that the meeting are worth many more times than the actual expenditure.

In any case why not just mention the hourly charge as that is what the unholy trinity AWS GOOG and MS charge by?

[Robert Grant]

Re: Are the numbers correct?

I suspect if this weren't volunteerism, the $30/day would never be worth saving in exchange for all the k8s creation/maintenance etc etc work that they will now have to do.

[doublelayer]

Re: Are the numbers correct?

You might be correct in this case, but the same applies if you have an existing team and a much higher usage. A company serving a bunch of data may be at a threshold where it's cheaper to have someone do a bit more admin work rather than pay the bandwidth bill. It just has to be calculated before making the final decision, which a lot of finance departments don't bother doing.

[Tim Anderson]

Re: Are the numbers correct?

The quote is correct. Regarding the savings, my assumption is that this was the saving from the last stage of the migration process as otherwise you are right, it doesn't add up. Presumably had no action been taken at all, the $6K per month bills would be continuing, as that was the peak.

[Peter2]

Re: Are the numbers correct?

I'd bet that $6k p/m was the peak because they realised they were paying that and started going all out on reducing the spending.

If they hadn't have started taking drastic measures then I suspect that it wouldn't have peaked there.

[Anonymous Coward]

Re: Are the numbers correct?

My interpretation of the article, was $2,500/mo are the fixed hosting costs without usage/transfer costs.

They saved $30/day, approx $1,000/mo.

So now they use Packet and pay only $1,500/mo in hosting and much reduced usage charges due to everything being on the Packet infrastructure.

[ovation1357]

Are they going to call this the X11 Desktop Migration to Cloud Policy or XDMCP for short?

Or perhaps the X Reduce and Rethink or X RandR project?

I'll fetch my coat!