I think the way to go is a mix of the two. It is covenient for companies to get guaranteed money in advance. As for customer it's way cheaper to start with usage based model. This is because a lot of users are underusing their services.
This at least is the case in hosting industry where overscheduling is a big thing.
So best scenario for customer would be a pay per usage for low usage and subscription for moderate to high usage. At least this is what AWS is doing and I believe this is fair for both bussiness and customer
Here in Lithuania we have like 90% of plastic bottles have a deposit logo on them, pretty much all of the soda, beer, water, milk bottles are recycled. Most of the non recycleable bottles are from outside of EU brands, and those are generally more expensive. Collection points are also automated, where you put them on conveyer kind of belt ant it uses OCR to detect if the bottle can be recycled. Those collection points are attached to a store and you get a credit in associated store which can be used to either buy groceries or jus get your money in cash.
But if your service was down for more than what it takes to downscale to minimum scaling back up is not that big of an issue. It was down anyways. Also 24/7 instances exist for a reason, autoscaling is there to handle spikes, not a normal traffic.
Pay attention, and don't confuse intention with effect.
What she's saying is that if you configure scaling such that it'lll scale down when demand is unusually low, and then demand returns, the spike may be a difficult one to handle, particularly if your services depend on each other but each scales only based on its own history.
If A needs B needs C, and demand suddenly returns to A, does that cause C to scale up? Or will A scale up first, and but C stay low for another half-hour because it recently scaled down?
Having C stay under demand for a half-hour after an outage ends wasn't anyone's intention when the autoscaling was configured. But as I wrote, don't confuse intention with effect.
Caveat emptor, you get what you pay for as the Linux support for the Allwinner H2/H3’s that power the OrangePi and NanoPi boards are hit and miss at best especially for the WiFi chips. Up until recently you had to either use a somewhat sketchy ancient fork of a Linux 3.x branch that wasn’t maintained or patch the mainline kernel.
The folks over linux-sunxi.org have been doing good work bringing various Allwinner SoC support to mainline Linux. Particularly the EMAC (ethernet) driver looks to be merged soon [1]. The driver for the XR819 WiFi chipset still appears to be up in the air, though good patchsets can be found for it [2].
The last I used the WiFi driver for the Allwinner H2+ SoC earlier this year on the (equivalent) NanoPi Neo boards it was regularly dropping packets. I suspected it was causing issues with other WiFi clients as well but couldn’t figure out how to test that. Still it’s a decent little SoC and both the OrangePi’s or NanoPi’s are handy little SBC’s that I’d use for a project that needed basic WiFi support. If anyone’s good with WiFi drivers and wants to help make it more solid the XR819 driver could use some TLC. It’s outside my expertise though.
Be advised that Orange Pi have many different boards with similar names, but vastly different specs. The $6.99 model is for 256MB RAM and a Allwinner H2 chipset.
Also a lot of the boards can only be powered by a DC jack, not the onboard MicroUSB.
Yes, the Raspberry Pi models are a little confusing, especially with the '+' versions and PCB-revisions. But I don't think the purpose of the parent comment was to say that it was worse than others, more of a heads-up to someone looking to order a model to take a second look and verify the specifications.
You can have a look at xeon e5-2670 v2 cpus, they're relatively cheap, around $200 and have 8c/16t so in a double cpu configuration you could have 64 cores with only 4 machines.
Judging from the photos, I'm not sure how well off these patients are. Bargaining from $25k to $8k was impressive but if you need to bargain that hard for a lift machine, how much is it to have a battery backup? and how difficult is it to have a proper hook up? or find someone who can do it properly?
Why would you really need physical machines for linux and windows, you can always virtualize clients and run dual boot mac/windows or single thinkpad with win/linux dual boot.
If budget is tight used servers are great value, for example del R710 is cheap with plenty of horsepower. Downside being space and sound.
Most of the designed emails have images which will be blocked by the client(thunderbird), which looks really horrible. So I will have to be very interested to actually allow images to be downloaded.
As I understand it he disabled asserts in places which would certainly fail, like doing system stuff, modifying system state and such, but not actual relevant parts.
You can always have multiple VM's with gpu passthrough, so it's almost like dual boot kind of setup, but you can still run other things on the host. It works pretty well for me, only downside being that you're limited to single VM running at the time.
You can run multiple VM's at the same time as long as you have a separate GPU to assign for each VM. There are videos on Youtube of people demoing this.
This at least is the case in hosting industry where overscheduling is a big thing.
So best scenario for customer would be a pay per usage for low usage and subscription for moderate to high usage. At least this is what AWS is doing and I believe this is fair for both bussiness and customer