New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
I see. Gotta stick with the "Let's try this provider --except blacklist.providers deadpool.providers" approach then.
Yes, however many would not go this way, you know there is a step in between, namely renting either some racks or a cage in an existing DC just like Amazon does....
It's funny how Amazon themselves do the exact opposite, always choosing owned in-house solutions over IaaS. Even AWS was made so they can profit from the idling hardware they normally use only on black friday.
I do, and it's more than floor space and empty racks.
There is certainly a point and/or situations where it makes sense to roll your own. Until you reach that point, or find your self in that situation, it inherently does not.
I'm in Virginia and using their Toronto location...I get around 32ms - I love Lunanode, so many features!
Tip for anyone who wants to use AWS:
Avoid it.
My friend runs a large based website, it used to cost him $200/month to run on AWS, which is fine, he woke up one morning it started taking off, one year later and $65,000 a month later, he's going broke. Only $1.2m left to spend, and since everything is so configured to use AWS, he doesn't want to migrate, so much for investors money.
AWS is nice if your looking for scale-ability.
Whereas, if he had been with a LET provider, his site would've continued to load fine for all those visitors for $7/mo.
If you spend all your budget on overpriced inferior services and get locked into them so hard that switching away can ruin your business, then you may never reach that point.
Nah, he could have started with five of these triple SSD 3.4 GHz E3 dedis with 32GB of RAM, and over time expanded to, say, 50 servers, and it'd still only cost about 2200 USD/month, 30 times less than AWS. And I don't know what kind of website you must have if 50 such servers wouldn't handle it (and how you don't earn those $2-3K/mo with that site).
Ya, LET is kinda crowd-funding while Amazon is just earning lots of money with it.
Correct. The biggest problem with their CloudFront/S3 is I can't set limits on traffic consumed.
So I can be easily drained for any sum by simply leeching the traffic off my resources. I have had such impact once, and had to terminate my CDN caching at AWS for that reason.
The feature is indeed much requested, but looks like AWS is quite reluctant at allowing suers to control their expenses easily.
Note: I terminate my constantly running VMs at them, butAWS is still a great platform offering many a tools in single place. Also, I appreciate their Glacier storage as well.
What website is your friend running...please specify the url.
Although these are only rough numbers and not indicative of any real world setup, spending $65,000 per month on EC2 would get some serious resources. In theory, the below would handle 10's of millions of unique visits per month. Note these are unreserved numbers and anyone running such a huge system would be taking advantage of reserved and probably spot instances pricing costs.
Amazon EC2 Instances:
25 x g2.8xlarge
**Per each instance: **(32 vCPU / 60 GB RAM / SSD 2 x 120)
Data Transfer:
20 TB transfer in per month (not highly likely)
Inter-Region Data Transfer Out: 60 TB
Data Transfer Out: 60 TB
Public IP/Elastic IP Data Transfer: 20 TB
Elastic Load Balancing:
Number of Elastic LBs: 10
Total Data Processed by all ELBs: 10TB
Amazon EBS Volumes:
25 x 100 GB EBS volumes May require some for backups. S3 is a better option.
RDS
10 x m3.large (10GB) MySQL databases
Cloudfront
1 x 20TB Cloudfront (CDN)
S3 Storage
1 x 2 TB
Inter-Region Data Transfer Out: 5
Data Transfer Out: 5
Data Transfer In: 5
I'll decline to post that without their permission.
The reality is it's about $55,000, but Incapsula to handle ddos attacks adds a few $, the paying himself, then paying his team every month, developers. It amounts to $65,000 after those additional costs.
I know for a fact he uses well over 5 TB on Cloudfront CDN, over 300 GB in SQL storage, and a ton of load balances, site averages over 2 mil request a day, amounts to about 83,000 visits an hour. That's the "low numbers." I'm not talking unique, at least 80% is recurring traffic.
So based on the above numbers you posted, it appears your friend runs a very public and active website but you cannot post the url without his permission?
Very public, yes, very active yes, and my personal choice not to post it is my personal choice, if you PM me I'll send you the URL, I just don't wish it to be indexed on here, attached with the URL address.
If you know what is good for you and your friend don't post it or PM it ... lest you want that bill to double overnight; with useless LowEndTraffic, LowEndDDoS, etc.
AWS is not that overpriced. Same amount of SSD is 15x at AWS but comes with more CPU power.
This is something I applaud, and recently ranted how few hosts do this. It is a delibrate decision to not use Route53 for Amazon.
If there's ever a global and cross tld problem with Route53 (however remote the chances) they will still be able to communicate with their customers and status page can remain up (subject to other dependencies).
Never had a problem with Route53, granted it can take upto 15minutes for each ns to serve the update (regardles of ttl - e.g. nslookup change.fqdn nsfaf.awsdns.org etc.. ), but after that you have pretty dane resilient DNS
Smart thinking imho
This is exactly how banks in Japan tried to stay afloat during the recession. There were no investment possibilities that made the interest rate they promised, so they'd bundle up all the deposits made by their customers and dump them in an account a competing bank.
That's exactly why I chose the Quesarito today at Taco Bell, because in world war 2 the majority of soldiers didn't wear sundresses.
Realistically no one really uses that, the *.2xlarge and *.large are far more popular iirc.
Same here, I've been using their DNS for years and haven't had a problem
I remember there was a billing glitch for me. And I took advantage of it by running up the gpu series VMs for a whole month. Then my account magically disappeared...
As I mentioned in my post: These are only rough numbers and not indicative of any real world setup_.
It was a bit of work to configure a hypothetical $65,000 per month EC2 based website usage cost that was mentioned by another poster and I don't imagine that anyone would take those rough numbers and configuration I created as a basis for a real world EC2 based website.
Also consider that much of AWS costs for large corps is likely RDS, Elastic Beanstalk, DDB, S3 and so on and not just "bare xen" EC2.
Well, not Elastic Beanstalk in any case, as it's free.
But yes, the business model is they have a whole bunch of stuff you don't have to configure and just works, so it's easy to launch there. Then you're locked in because your data is there and porting elsewhere will not be realistic.
No, if they caught you using prepaid cards, they will disable some services (e.g. Route 53 and Cloudfront)
And some regions of EC2.
You sure as hell had some interesting history teachers in school.