Public cloud prices to surge in US and Europe next year Public cloud prices are forecast to jump by almost a third in Europe next year as the cost of borrowing and energy squeeze providers. For the US, the expectation is for the price list to jump by a fifth. The market exploded during the early days of the pandemic as much of the world switched to remote working, and continues to …
It's always been the case - e.g. Oracle vs SQL - as soon as you start using each platforms unique functionality, you're hostage to the "cost of change", if you don't use the unique functionality, you're paying for something you don't use, and maybe even buy another product to provide it - which you then add the "pain of integration" to the mix.
[The big hyper-scalers have already bought into the whole eco-power thing, so don't expect their prices to hike as much as the smaller providers running off the "normal" grid?]
Yes, but unless they're generating their own renewable energy, the prices for renewable sources have also come up because people want energy from wherever they can get it. Also, in many cases, the sustainability numbers from clouds mean that they used gas because they have to have the reliability, but they offset emissions in some way to cancel out the carbon they calculated they released. In both cases, cloud providers are still subject to fluctuations in energy prices and probably only a little less than other companies due to their cash cushions.
And the most amazing (not!) is that the seasoned tech staff generally warned about putting your eggs in the cloud basket and predicted very bad weather to emerge from the clouds when they eventually coalesce.
But the MBAs knew better because short term profits always trump long term gains. Who do you think got fired, the techies or the MBAs? And who will be fired when the cloud costs become unbearable?
This is the difference. I can control our costs if the infrastructure is on site - we can consolidate VM's/containers onto fewer physical machines, turn up the air conditioning by a degree, get rid of the last older devices that are less power efficient than newer ones...
If everything was hosted by Azure or AWS we'd be beholden to them and any price increase they put our way, with no quick way to move away to save costs.
There's always a point when something that was previously acceptable becomes not longer acceptable. One might have delayed some consolidation because the costs and maybe downtime didn't justified it before - but now it became justified. People who never review the situation and adapt thinking they did already everything are those who become extinct.
Hosting yourself is also going to be costing more due to the increases in energy prices, but I guess some people don't book that against "IT".
Yes, and no.
The waste product of running your own servers is heat. If you can use that in winter (and it takes a particularly stupid person to fail to utilise this) then this comes off the gas heating costs. As gas is currently 50% more expensive than electricity it actually ends up not looking too bad.
"As gas is currently 50% more expensive than electricity it actually ends up not looking too bad"
Where does that 50% figure come from? Although the OFGEM price cap distorts these, typically prices for domestic are around 10p/KWh for gas and 34p/KWh for electricity - although commercial rates will be significantly different to these it seems unlikely that the ratios will be massively different; indeed since around 40% of UK electricity is generated from burning gas it's hard to see how gas could be more expensive.
"The waste product of running your own servers is heat."
While there are kitchen appliances that use excess heat from the refridgerators compressor to pre-warm water for the washing machine or even the shower, it is intended to build data centers that effectively use their excess heat for heating of residential areas and similar purposes.
Kitchen appliance:
https://efahrer.chip.de/news/pfiffige-warmwasser-waermepumpe-von-bosch-sie-nutzt-die-abwaerme-anderer-geraete_109532
Data center:
https://stockholmdataparks.com/2017/08/31/neues-rechenzentrum-stockholm-beheizt-10-000-wohnungen/
energy is a factor with on prem of course but I think not a huge one. Take my org for example, when Covid hit they wanted to cut costs. My infrastructure runs super reliable, hardly any failures, so I was comfortable cutting pretty much all VMware and premium HPE hardware support for our servers. That took the annual cost down anywhere from $3-5000/year/server (depending on server type) to about $200/year/server. The costs of server support alone(yet alone support costs for other things) exceeded the cost of power/data center space for the environment. It's not something I would do for every situation of course but the option is there. I averaged less than 1 vmware support case per year prior. My config is simple, conservative and has been solid as a rock for a decade across vSphere 4.1, then 5.5, then 6.5. (skipped versions not mentioned company started by moving out of cloud to vSphere 4.1 in 2012).
I took some of our storage and put it on 3rd party hardware only support as well saving a whole lot as well($85k/year -> $12k). I felt comfortable doing this because of the operating track record and my experience and low amount of change in the environment. Network equipment remained on vendor support.
With full public cloud you are hit with all of those extra support fees (included in the cost of the product) whether you need them(with cloud you probably do since they are always fucking with it) or not. Of course most big IaaS clouds aren't paying vmware fees, but they have their own support costs, development costs operations costs managing and updating their systems. Managed hosting has even more costs since they are even more hands on, vs Colo (which I have been working with since early 2003).
You also have the ability to run on hardware for longer periods of time, I have data center switches in mission critical roles that are literally 11 years old next month(planning on replacing soon, my 11 year old 10G switches still not EOL..), other switches are 6 to 9 years old. Load balancers 8 and a half years old(not EOL yet). I have servers still in use - 8 years old.
All of that is tons of cost savings. Things you don't have flexibility with in cloud (or managed hosting) because you can't control those things. Of course everything has to be retired at some point. Our mission critical all flash storage array celebrates it's 8th birthday in less than 2 weeks (0 hardware or software failures during that time). My hypervisor (ESXi 6.5) just went EOL this month so will be looking to update compatible systems to 7 soon(already have the license upgrades from a couple of years ago but postponed the project due to various reasons). I went past EOL on vSphere 4.1 and 5.5 as well before upgrading(which for me means fresh installs I'm not going to upgrade 6.5 to 7, going to start fresh with a clean config, prefer things to be clean even if it's more work up front).
Add to that the ability to over provision, something you cannot do in almost all IaaS clouds(certainly all of the biggest ones).
Add to that all of the biggest clouds they always say "pay for what you use" but really it's "pay for what you provision". If you provision a 8CPU VM with 32GB of ram you are paying for that VM regardless of how utilized it may or may not be. vs on prem you can have many such VMs if you know that they all won't be busy at the same time you can over provision safely.
Under the guise of inflation and energy price increases this is the bait and switch in action. Too late and too expensive to revert to on-prem so you have to suck it up.
On-prem energy costs can be mitigated by installing solar panels on the roof, or even the car park, which is half empty with a lot of folks working from home.
Some places have substantially cheaper electricity than others. Like Quebec vs London.
Will AWS or IBM or Vultr or lesser vendors offer a discount on places with cheaper electrons?
Or perhaps it will be labelled a "fuel/electricity" surcharge if you rent a VM located in Amsterdam over one located in Ohio.
They have that, although they don't explain why the prices are different in each region. See, for example, this system that compares pricing per region for Azure VMs:
With Amazon and Microsoft (and to a lesser extent Google) so dominant in the public cloud that there isn't enough competition to stop them from raising prices. The minor players will go along with it because they can't compete with Amazon and Microsoft's cost structure so any excuse to raise rates is fine with them!
It is stunning to think about this, given how fast technology improves and reduces cost. Cloud has always gone down in price, never up. CPUs may not get faster at the rate they did, but they are still adding cores at quite a clip and that's what matters for the way cloud is charged. Storage always gets cheaper per unit, and while memory prices aren't falling too quickly they are always falling in the long run.
Given the increase in energy prices in the EU I could see a temporary surcharge for that but a permanent 33% rise? If the Ukraine situation was resolved and next spring energy prices returned to more normal levels, there's no way they are going to roll back that 33% increase. Maybe they might make a token 10% drop but they will be used to the bigger cloud revenue juicing their earnings and won't have to give it back. Microsoft (who are seeing falling revenue from PCs due to the PC market returning to its pre-pandemic mean) and Amazon (who are seeing falling revenue from the pandemic order from home peak) like having an excuse to avoid having to report falling revenue overall.
Our CIO suggested that part of the 'benefit' of moving to 'The Cloud' was predictable regular OpEx rather than bursty CapEx. To be fair, he never claimed it would be cheaper - just the costs more evenly spread over time. The techs didn't believe him at the time.
Of course, it depends what you mean by 'predictable'.
One advantage of 'The Cloud' was the techs not getting suckered in to just 'squeezing one little experimental system in' onto existing hardware 'on a strictly temporary basis, of course'. You want to run something experimental? You apply for the budget to rent the cloud resources too.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
