5 ms·
Amazon's Cloud Is One of the Fastest-Growing Software Businesses in History
- beachstartup 12y agoa couple of points about aws: 1. it's not really a software business. their marginal cost of products sold is nowhere near that of software. the software they develop facilitiates the sales/leasing of hardware and networks. 2. it's a division of a much larger parent company that poured billions into it. it's a "business" but not a company, which is what they're comparing it against.
- dba7dba 12y agoI agree with the department versus a business. No other startup could've done what they did, providing the tech knowhow, logistics (always a big issue at that scale), and the brand. EVERYONE willingly gives their credit card info to Amazon.
- Chronic29 12y agoCategorizing AWS as a software business is incorrect. That is like saying Amazon.com is a software business (which is incorrect).
- dictum 12y agoAmazon's cloud services are a software business in the same way a utility company that wrote software to automate some management tasks is a software business.
- teej 12y agoI had to call a human being and use a fax machine to get my power turned on with the local utility. I can spin up an EC2 cluster with an API call.
- opendais 12y agoHeh. I signed up on a website and put in a requests to turn my utilities on.
- Mandelbug 12y agoHave you actually used and looked into AWS? There are dozens of services that do a lot more than just what people generally think of as "the cloud".
- flyt 12y agoOne of those rare examples where the best product is winning in an entire category.
- meddlepal 12y agoI don't know if they're the best at everything. I've been using Google Compute Engine extensively lately and I think it is a lot more polished feeling at times than EC2.
- tszming 12y agoI am interested to know which part of the GCE is a lot more polished than EC2.
- nulltype 12y agoWell the TCP load balancer is pretty nice. Our EC2 load balancer (HTTP, through heroku) seemed to die all the time.
- crb 12y agoOne data point: On AWS, a hardware outage takes your machine down (with notification). GCE offers live migration, which you generally don't even know has happened. See http://www.rightscale.com/blog/cloud-industry-insights/google-compute-engine-live-migration-passes-test http://www.rightscale.com/blog/cloud-industry-insights/googl...
- toomuchtodo 12y agoYou should be building your AWS infrastructure around failure scenarios, and while you should be notified of a failure, your app should continue to chug along. /DevOps
- crazypyro 12y agoIsn't he just saying that Google does this automatically for you?
- kimonos 12y agoTwo thumbs up!
- padobson 12y agoI'm not surprised at all. When I design a web app, I'll design it around AWS infrastructure so that they're the "Ops" in my "DevOps" title. They handle everything about web app development that I don't care to.
- tluyben2 12y agoI wanted to come and type that up here; when we do apps, web, scaling, cdn, file storage, trivial SSL etc we just prep for AWS. All the scripts are there as most people ask for it and it always works. With their solid offerings and constant price drops I would see no reason to ever move. It's all the stuff so painful once and now is all such a warm bath.
- yry4345 12y ago> "I would see no reason to ever move." (Obligatory) Apart from the single political point of control. AWS is to web apps what GoDaddy is to DNS and Verisign is to SSL.
- justicezyx 12y agoCould you justify your relating? Your relating basically suggest any one's father is any father to his son/daughter. I could not agree that for any child, they are fathers treat them equal good/bad... In summary, it's find to be alert, but be sure to not over-generalize...
- iancarroll 12y agoYou're right, actually. I've dealt with all three, AWS is definitely more friendly (if we're talking about support here) but they all have the monopoly. AWS, honestly, is the only one that actually has an advantage. Verisign is overpriced, GoDaddy is just a typical domain registrar.
- ChuckMcM 12y agoIt isn't particularly surprising, but it would be much more interesting to look at margins vs revenue. What is the margin on that $5B vs the $3.1B? I'm guessing that given the recent price cuts in between Amazon, Google, Softlayer, and others that the margins are starting to get squeezed down a bit. I find that particularly interesting when looking at the relative success of Docker. The reason being that IBM's Z1 mainframe seems to have better operational economics near full utilization than rack and stack servers do. You can see how that works by looking at where the OpenCompute stuff takes out cost, no more sheet metal around servers, shares power supplies etc. Makes me think that 'blade' computing should have been 'building sized' from the beginning :-)
- opendais 12y agoAren't those like $75,000? I'm curious what you are basing the operational economics statement on.
- alttab 12y agoProbably the lifetime of the mainframe versus the long running cost of hosting it on AWS? 8 years, you know what you need it for, might be worth it? Not if you don't have very large POs in hand, however. This includes most start ups.
- ChuckMcM 12y agoI currently have about a thousand servers that do Blekko's web crawl in a colo facility, they are pretty modernish (westmere vs Ivy bridge but similar configuration), the operation economics for running those is a couple million a year fully loaded. My particular application is a 'bad fit' for services like AWS due to its large storage foot print and the fact that it constantly churns through petabytes of data but were we to host web servers w/ Docker or the equivalent we could probably provide something like a 100K web hosts. Doing that $5 'droplet' style with Digital ocean would be say a half a million a month so pretty good margins (caveat all the things like racks and drive replacement and floor space) but when those margins got squeezed something like an equivalently sized Z10 could shave operational costs from both a operator perspective and a power/cooling perspective. (total watts per delivered core-gigabyte-megabit of compute delivered to the Internet) At web scale, costs shift in interesting ways, operator time is much more valuable than say the cost of hardware power, and cooling overwhelm floorspace costs etc. Luiz Barrosa is an engineer who gave a talk titled Data Center sized computers which alluded to some of those shifts, Urs and Luiz updated it a couple of times. As I've gone from NetApp to Google and now to Blekko it has been interesting to me to watch how the multi-core, large memory, SSD changes have pushed around system design like this. And I was thinking that back in the day the "big problem" with blade computing was that using 20 blade chassis an install with 20 blades was cost effective but one with 21 blades was not. The virtualization wave made that moot since you only install full racks/chassis as demand calls for it, and current container wave adds to that single OS type installs for even better cost savings. Fun times.
- deeths 12y agoAWS is doing some great stuff, but this is a really misleading article. Ignoring the obvious issue that AWS is a service and not a software business (and they have very different growth characteristics), the data on the graph is incorrect. Perhaps calling it a "tech" company would have been more clear. AWS's $3.1B revenue in 2013 would be year 3 of the graph. Everything after that is speculation, but is compared to actual historic data from other companies (without explanation). Several of the other companies lines seem significantly off also. For instance MSFT's first billion dollar year was 1990. By year 3 they were making $2.76B, not the just under $2B shown. By year 4 they were showing $3.75B, not the under $3B shown. By year 6, MSFT was making $5.94B, not the $4.5B shown. Similarly, the lines for CRM and VMW are showing lower revenue numbers than the actual ones for the later years. For instance CRM should be $3.72B in year 4, not the $1.75B shown. If you stop the graph at year 3, where actual AWS data stops, AWS has grown only a little bit faster than MSFT. Also, the periods graphed for MSFT, CRM, and VMW all include recessions, whereas AWS is growing during a boom.
- eevilspock 12y agoFurthermore, if you inflation adjust all of the numbers, AWS's graph will deflate further.