Fri

Apr 27
2007

Brady Forrest

Brady Forrest

Jamglue's Growth and Scaling

jamglue
Jamglue is a Seattle-based, YCombinator company that has grown quickly since its launch in December 2006. They are an audio remixing site that allows for simple sharing of audio samples across its members through a searchable database and social network. Remixes are made via a slick and fun Flash app.

Jamglue is currently getting 110K visitors for 1.1M pageviews per month. They've been using S3 since early development. Jamglue only has two servers of their own (and they were bought from the Reddit guys for a couple hundred dollars and a few beers). At an Amazon Web Services event yesterday co-founders Gautam Jayaraman and Casey Muller revealed how they had increased their usage and what their bills looked like at each stage.

In their presentation the Jamglue co-founders broke their usage into four phases. In phase one they used S3 to store audio files. Users upload audio files to JamGlue's servers and the files would be sent to S3 at a later time. They didn't want unrestricted access to the audio files so authenticated, auto-expiring redirects would be sent to users via the Jamglue servers. In phase two they started using S3 to store their user avatars. The avatars are never on JamGlue's servers.

Jamglue's confidence in S3 grew and they now store all static content there (phase 3). This includes their CSS, JavaScript, and SWFs. They didn't get into specifics, but they called out that they have had issues when S3 is slow and these files fail to load. The final phase moved the mixing process to EC2. The audio files are still on S3. EC2 instances grab the files from their remix them into a new audio file and then push the updated files back to S3. Planned future work includes having the audio files go directly to S3 and to move more computational processes to EC2. They may also need to buy more of their own hardware to speed up the site.

Close to the end of the presentation Gautam and Casey showed their approximate monthly AWS bills for four milestones.

  • Pre-Launch - $1.50/month
  • Beta - $15/month
  • Launch - $150/month
  • Now - $600/month

Those are pretty amazing numbers -- especially when you consider Jamglue's traffic numbers. Jamglue is a young company and not having to make big investments in infrastructure will help them while they sort out their revenue model (currently looking like sponsored contests), but they do not have many paying customers who will complain if the site is down (though certainly their non-paying users will). They can afford (for now) to rely solely on a third-party vendor for their infrastructure.

Don MacAskill's SmugMug is in a different boat and has been quite public with both the savings from using S3 and some of the heartaches. Read his posts for how they have developed an architecture that utlilizes S3 while keeping them from being fully dependent on a third-party (also see his slides from his ETech talk on this topic).


tags:   | comments: 1   | Sphere It
submit:

 
Previous  |  Next

0 TrackBacks

TrackBack URL for this entry: http://blogs.oreilly.com/cgi-bin/mt/mt-t.cgi/5453

Comments: 1

  Greg Linden [05.01.07 02:12 PM]

Hi, Brady. Interesting post, but I have a hard time figuring out if those cost numbers are all that impressive.

Findory.com, for example, got over 10M page views and had over 400k unique visitors in April 2007, substantially more than Jamglue, and Findory runs on three online servers (six servers total, including crawl and development servers) for a cost of under $300/month (under $600/month for all six servers).

There may be an argument that Findory needs fewer servers than Jamglue, though, since Findory is mainly compute-heavy (personalized pages generated in real-time) and Jamglue probably is data-heavy (audio streams). It would be interesting to have an estimate how many servers and how much bandwidth Jamglue would need to be serving all their traffic and data themselves.

It is true that the SmugMug numbers are compelling. Part of that is because SmugMug is primarily using S3 as secondary online backup for their image data. Other startups that have similar needs might want to consider that model.

Post A Comment:

 (please be patient, comments may take awhile to post)






Type the characters you see in the picture above.