Amazon has slurped cloud cash for years, but a recent analysis by El Reg shows that the mega bit-seller charges a surprising premium for some of its services, and is facing stiff competition on price from rivals. Vulture West's floating cloud bureau was recently roused from a beer-soaked sleep by a phone call from the denizens …
you can expose anything you like to me
What about EU hosted cloud?
I know that your MPs don't care, but I'd rather see a comparison of cloud providers that are able to respect EU privacy laws, leaving out any that have to bend over to the USA.
It would be great if you could expose the data; since you've elected to use log scales it's practically impossible to figure out what the data points actually represent at a glance.
Given the downtime some of the cloud providers have had this year, it would be nice to see a plot of who had the most downtime and how severe that outage was.
For instance, Microsoft's glitch with deployment last month would rank relatively low as nothing actually went down, you just couldn't deploy.
However the certificate expiry outage, the other one they had that I can't recall the cause of, and Amazon's various EBS clusterf**ks would rank higher as there was actual loss of availability.
It would be nice to see, once and for all, whose cloud provider has the worst actual availability.
Please can you publish your results in an Excel / Google Docs file. In addition would it be possible to disclose the methodology you've used? I recently attempted the same exercise, however I was unable to successfully disaggregate memory, vCPU and storage (and storage type): e.g. how is the price difference between an m1.small and m1.medium attributed between the difference in vCPU and memory? Thanks.
Open data format?
Why not also provide the data in format people can use easily, even if it's just csv.
could you rewrite without using the word cloud please
those graphs are pointless if you don't label your axes properly!
price ($) - is that per second/minute/hour/day/month? a one off cost?
Re: those graphs are pointless if you don't label your axes properly!
That doesn't matter for comparison or analysis --- 20% more expensive stays 20% more expensive.
For acquisition, looking at any specific online offer (vendor/RAM/cost) as you would anyway if interested, that would let you pinpoint the scaling since the graph wouldn't overlap (from 60x scaling between minute and hour, down to 7x between day and week) with such a ballpark figure.
A downloadable CSV would be fantastic as I could then use Pentaho to send it to me whenever something has changed.
Yes please, More data. Always more data.
Data please -- and live popups
Yes, please make the data available.
For those of us that are only curious, a version where the name of the particular offering would pop up as you pointed at each data point would be nice.
just the title of the article shows how clueless the author is (never said that about an el reg author before). amazon has been absurdly priced for MANY MANY YEARS now, again it comes down to the architectural flaws in how the system was designed(they couldn't do a better job at the time because their technology sucked, and they're so far down that suck road now that it's really really difficult to change gears at this stage).
But wait.. there's more. Google has the same model.. HP cloud is the same model.. Rack space Openstack cloud is the same model!! Openstack in general (last I was briefed on it anyway) adopts that same broken model (again because the technology isn't good enough to do it right). The likes of VMware, Hyper-V, RHEV etc got the model right, it's just complicated to scale to very high levels. So these public cloud players take the short cut, of massive compromise on features, availability, and capacity utilization in exchange for massive scale == large costs for the VAST VAST majority of applications.
You're basically reversing a efficiency trend in IT that has been going on for the past 8-10 years and going back to 90s era provisioning strategies, at least from an IaaS perspective. PaaS, and SaaS are provisioned and billed with different models and those can make sense at large scale operations for customers(obviously both of those have their own limitations). the IaaS technology the public cloud folks are making available is just absolutely terrible by contrast.
But this author doesn't know what he's talking about, I don't know why I bother to keep writing comments in his articles I suppose I should just stop.
(El Reg reader for the past ~14 years)
It's a Daily Mail style bait article. Just go with the flow, (i.e. ignore).
The posted data was interesting to see. Other things that would be interesting to see but may be more costly or difficult to compare are:
- Network infrastructure capabilities (VPCs, ELB's, etc, vs. cost)
- I/O capabilities vs. cost
- instance uptime vs cost. i.e. how much it costs, just to have an EC2 instance up for a month vs. equivalents on other offerings
After that it becomes interesting to compare some of the application support pieces, like DynamoDB and SQL offerings from the various vendors, and the relative prices.
Just some suggestiobns.
It's not just the CPU/RAM
For deploying a practical application, it's the databases, the bandwidth charges, the load balancing, the redundancy, the location of data centres, the backups, the long term storage...
It's the availability these that currently draw me to use Amazon for my next project: the cost for the CPU/RAM combination is only one factor.
A more useful comparison might be based on a set of sample applications with N web servers, a N gig database, 2 geographic/availability zone redundancy, and an assumed amount of traffic, regular backups, failover between servers, etc. and then estimate the running costs of the example setup.
This stuff is important but hard
Comparisons are difficult. And odious. And temporary. Sigh.
For me, this is a useful piece of work, which *I* would find hard to duplicate.
Please update it regularly, say twice a year?
twice a year, interactive graphs
And a link to the raw data so we can check/provide alternative transforms for our bosses if we're comparing amazon to google, and no-one else. i.e. We could have decided who we want to work with because of factors other than cost.
like to share is welcome, including spreadsheets :-)
What kind of instances?
Thanks for the great charts.
Just wondering if these prices are for Spot Instances, On-Demand Instances, Reserved Instances, Dedicated Instances, or their equivalents with the each provider? Any idea how the cloud hosting market share is split between pre-paid workloads (i.e. reserved or dedicated) and on-demand?
Please keep up the good work.
Re: What kind of instances?
Hullo - these are all on-demand workloads, which I believe are meant to be about 60/70% of overall market, though YMMV.
How do I get this spreadsheet?
I'm currently reworking the model to include further info, and will figure out a way to expose it in a more useful manner either via a live sheet or a data dump at some point soon. Cheers!
- iPad? More like iFAD: We reveal why Apple ran off to IBM
- +Analysis Microsoft: We're making ONE TRUE WINDOWS to rule us all
- Climate: 'An excuse for tax hikes', scientists 'don't know what they're talking about'
- Analysis Nadella: Apps must run on ALL WINDOWS – PCs, slabs and mobes
- Yorkshire cops fail to grasp principle behind BT Fon Wi-Fi network