* The Business of Software

A former community discussing the business of software, from the smallest shareware operation to Microsoft. A part of Joel on Software.

We're closed, folks!


» Business of Software FAQ
» The Business of Software Conference (held every fall, usually in Boston)
» Forum guidelines (Please read before posting!)


Andy Brice
Successful Software

Doug Nebeker ("Doug")

Jonathan Matthews
Creator of DeepTrawl, CloudTrawl, and LeapDoc

Nicholas Hebb
BreezeTree Software

Bob Walsh
host, Startup Success Podcast author of The Web Startup Success Guide and Micro-ISV: From Vision To Reality

Patrick McKenzie
Bingo Card Creator

Hosting software distributives


I want my customers to download my app as quickly as possible. Currently, I host my installers on S3 and serve them through CloudFront. From what I see, there is a high cache miss rate. For every 100Gb of traffic from CloudFront there is 48 Gb of traffic from S3. So, just about 1/2 of all requests are served quickly from the nearby CloudFront node.

As far as I understand this happens because my installers get replaced by "more popular" content (http://docs.aws.amazon.com/AmazonCloudFront/latest/DeveloperGuide/Expiration.html).

Is there any service providers I can use to host my 100Mb of installers that won't remove them from their edge servers? I'll appreciate any other solution as well.

Ivan Nikitin Send private email
Friday, July 25, 2014
I am using Google Drive (as a Google Apps customer - I am afraid to use it for free) and its pretty fast.
Atanas Krachev Send private email
Friday, July 25, 2014
I don't have an answer for you other than trying one of the other CDN providers (http://www.thetoptens.com/best-content-delivery-network-cdn-providers/) but it's nearly impossible to benchmark them in a reasonable way against each other.

However, I wouldn't equate "not from CloudFront" with "slow".

I host my (much smaller, ~5MB) installers from S3 and I think it's plenty fast.

Now, maybe you have a concrete data that shows that downloads from CloudFront complete faster than those from S3 for your users, but as I understand it, CDN are most useful for low latency so it matters much more for lots of small files.

Amazon has big pipes. Once your 100MB from S3 gets going, the bottleneck will most likely be your user's "last mile" connection, not the fiber between Amazon's data center and your user's interconnect point.
Krzysztof Kowalczyk Send private email
Friday, July 25, 2014
I use Rackspace CloudFiles which automatically uses Akamai as the CDN: http://www.rackspace.com/cloud/files/

It's also easy to host a static website there, which turns out to be very cheap and very fast.
Gareth Hayter Send private email
Sunday, July 27, 2014
Does the download speed increase have a positive impact on your bottom line? How did you measure?
Dmitry Leskov Send private email
Monday, July 28, 2014
I'd have a problem with the download speed of my app, which is currently 140M, and quite slow from a webserver.

I've not been able to find a faster service that has a limit on how much you can be charged. I don't want to expose my business to an unlimited liability that's outside of my control (unlikely, but what if someone used a botnet to download my app from Amazon).

I appreciate that not many people seem to care about this, but any ideas? I'm perfectly prepared to pay a higher rate than normal to get this.
Robert Muir Send private email
Monday, July 28, 2014
@Krzysztof, my app is big and 150kb/s I see from S3 is far from enough in my experience (see below).

@Gareth, tried them and, unfortunately, got banned by their security once I entered my location (I'm not from US).

@Dmitry, it's my hypothesis. It's based on my feelings. I try and purchase apps for my needs too and really hate when something is getting on my computer very slow. I don't want to wait for 20 mins. Basically, if I can try the second app when the first one is still downloading, I'll most likely buy the second one.

@Robert, you can specify an email alert with AWS when spendings are going to exceed  a specific amount. However, isn't it as likely as an ordinary DDoS attach on your server? I mean do you have enough of bandwidth and computing power to cope with a 1-2Gbps attack?
Ivan Nikitin Send private email
Tuesday, July 29, 2014
@Dmitry, you released your product for Mac! I LOVE you!:)
Ivan Nikitin Send private email
Tuesday, July 29, 2014
We are serving our downloads from a Linode. We are on a legacy plan that includes 6TB/month transfer, which is sufficient for all our needs.

At the moment, you can get a Linode with 2TB of transfer for $10/month. If you need much more than that, Hetzner auctions dedicated servers with traffic capped at 10TB/month. Overage is free but the speed beyond the cap is limited; you have an option to lift the cap at 2 Euro/TB.
Dmitry Leskov Send private email
Tuesday, July 29, 2014
For the record, I get 2MB/sec between our Linode (US East datacenter) and a server hosted by Hetzner in Germany.
Dmitry Leskov Send private email
Tuesday, July 29, 2014
> @Dmitry, you released your product for Mac! I LOVE you!:)

Thanks :) It has quite a few rough edges, but the core is working.
Dmitry Leskov Send private email
Tuesday, July 29, 2014
@Gareth, It turned out that Rackspace CloudFiles doesn't send files over 10Mb in size over CDN:

So, no acceleration for Australian or European customers.
Ivan Nikitin Send private email
Tuesday, July 29, 2014
Ivan, its not my site going down that I'm concerned with, but how much money an attack would cost me before I could find out and stop it. There is no upper limit to the cost, apart from whatever Amazon's hardware is capable of supporting. I probably get a few G per month normally, and I'd hate to see a bill one day for $1M.

I never got past that point, so I didn't realize there were all these other issues.
Robert Muir Send private email
Tuesday, July 29, 2014
I made an experiment after discussing hosting options here. It turned out that download speed is actually an important factor and CloudFront is better than S3 for that purpose. I made a blog post with experiment details and conversion chart: http://nikitin.io/cloundfront-vs-s3/
Ivan Nikitin Send private email
Sunday, August 17, 2014

This topic is archived. No further replies will be accepted.

Other recent topics Other recent topics
Powered by FogBugz