Sorry Google, it's boring old workloads that are pumping up AWS and Azure, not sexy AI Google Cloud Platforms's chief thinks the service will surpass AWS by 2022. Speaking at Forbes CIO Summit in Half Moon Bay, California, last month, Diane Greene claimed Google has "a huge advantage in our data centers, in our infrastructure, availability, security and how we automate things. We just haven't packaged it up …
Went to the Google Next event a couple of weeks ago and Google really are no where in the enterprise environment. LOTS of focus on ML but that would involve putting your core data in Googles hands and as far as I can see most enterprises are not interested in doing that. Spoke to a couple their guys and they are not looking at providing their tools for on-prem hosting whereas MS are quite happy to sell you a on-prem Azure, Dynamics and other tools even if it affects their cloud hosted profits.
Google, AWS and Azure are all investing in a data centre infrastructure encompassing bandwidth/land/physical buildings/hardware that will allow them to deliver services to any point on the globe as economically as possible.
While Googles "boring" services may not match those offered by Azure and AWS they're still in the game because of that they already have and the amount of money they bring in.
The Oracle's, SAP's, IBM's, SalesForce etc hoping that their existing applications can be delivered by much smaller cloud infrastructures will either lose out to new solutions from the big three or become a customer of one of the other large infrastructure investors (Facebook/Alibaba/Baidu/Tencent/AT&T based on Intel's top 7+1 customers) and move their solutions to one of those providers.
There will always be solutions where cloud doesn't fit the requirement, but for applications covering large geographic areas, the big three will be hard to beat.
It's not AI that companies want but (in no particular order):
* Security
* Privacy
* Uptime
* Backups
* Clear costs
* Blocking DDOS traffic
* Automatic increase in capacity at peak times (or else why have cloud?)
Curiously so called Cloud [Hosting] providers are cagey about such details. Priorities vary.
Google's VM tech is substantially different to AWS's. Amongst other things it lets them spin up VMs more quickly than AWS and lets them offer customer selected memory/CPU options (as opposed to the families that Azure and AWS offer).
That, in turn, let's them offer things like BigTable and BigQuery. AWS can match, but not as efficiently because they have to keep more spinning reserve.
And then there's stuff like Spanner.
Look at something like AWS Athena. A Google idea copied by some Hadoop folk at Facebook, commercialised by AWS and released 7 years after Google published.
"Google's VM tech is substantially different to AWS's. Amongst other things it lets them spin up VMs more quickly than AWS"
No one really gives a crap how long it takes to spin up a VM within reason. But for the record Azure is faster. Just like Hyper-V is fastest under Open Stack...
to the company that can figure out how to move enterprise workloads into the cloud.""
Well in Amazon's case actually paying a dividend would be a start. 20+ years old and never paid a dividend with P/E ratio of 112. That would be 112 years to double your investment, if it distributed its profits this year.
Here's the thing. In theory "cloud" offers 24/7/365 availability, migration to other data centres in the event of failure, smooth recovery when a (in theory rare) failure does occur, capacity on demand.
And then in February the AWS East-1 data centre went titsup.
Which showed most of this theory to be complete BS. The biggest selling point for "cloud" seems to be bean counters, where it offers the illusion you can replace CAPEX (on new servers, broadband, data base licenses) with an operating cost of X $/month.
Companies want security and consistency and sysadmins want a relatively easy life
IRL all "clouds" run somewhere and a lot of those places seem to have governments that like slurping anyone's (and anythings) data. Let me suggest the long term winner in this game will
1)Supply tools that completely map a businesses networks and data bases systems, identifying security weaknesses, violations from best practice, data migration issues etc. Include a cost calculator function so sysadmins can see what the impact of changes would be.
2) Let you use those tools to configure the necessary "cloud" VM's and storage either to mimic the original structure, or with some (or all) of the shortcomings fixed
3)Handle the licensing admin (of which I can bet there will be a lot)
4)Specify any limits on where the instances can be located.
5)Set up a suitably configured big pipe to handle the Extract/translate/Load process to the cloud.
6)Allow easy monitoring of and raising/lowering of processing and storage capacity on demand.
7)Make it as easy to migrate off your cloud, either to another or back to their own servers.
1 & 2 are tech heavy and likely to need deep insight. Better to do very well on a few environments than badly on a lot.
Note this process gets customers onto your cloud and makes them feel they don't have to stay. paradoxically the ability to leave easily is more likely to encourage them to stay, since they don't feel locked in to begin with.
As for data centers might I suggest in England the North East around Barrow? Lots of smartish people (they can build a nuclear submarine) and a chilly climate to dump heat into. Most parts of Scotland would be OK but I'm wondering if the Sturgeon is planning a migration of her own.
"P/E ratio of 112. That would be 112 years to double your investment, if it distributed its profits this year."
And your point is? Clearly they have no need to pay a large dividend to maintain perceived shareholder value. Unlike, for example, Railtrack, which had a P/E of 12 just before all the investors lost everything.
"migration to other data centres in the event of failure ... then in February the AWS East-1 data centre went titsup."
And again your point is? us-east-1 had an outage on one service, and people migrated elsewhere just like the promise.
"And again your point is? us-east-1 had an outage on one service, and people migrated elsewhere just like the promise."
I think part of the problem was that many of them couldn't migrate...because all the other data centers were too full. It's like trying to move 10 eggs from a broken carton only to find the only other carton only has room for two more. It's a very hard problem regarding load: nobody really likes surge capacity because you have to pay for it even when it's doing nothing; ideally, you keep as much of it running for as long as you can to maximize your return on investment, but then you end up with no surge capacity for when crap happens. It's true of iron; it's true of hospitals. It's a case of you can't please everyone.
Yeah, that's definitely not true. If by cloud you mean Google or AWS, it is way less costly. People who make those claims are just calculating part of their costs. For instance, they take the cost of a low end Dell x86 server and then compare it to the same amount of compute on the cloud. Ignoring the cost of management software, clustering software, support agreements, the data center itself (real estate, generators, cooling systems, cabling), electricity, the interconnect network, added management and maintenance staffing, etc... and they also don't factor in that, if on demand, you have way greater flexibility to just turn the server off if no longer required. Scale to actual utilization vs peak with on prem.
It essentially like saying "this house costs $400,000 but I can buy as much wood as that house is made up of for way less than $400,000."
"Given that CIOs trust Microsoft and already give a large percentage of their workloads to them"
Well, they are winning the MSFT stack workload bc they control all the pricing. Azure is tiny outside of MSFT stack. If CIOs were in love with MSFT, they would be the market leader... many would like to be rid of them.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
