About us
Our Products: Personalize and Connect
The web future is bright and sunny. We enjoy developing technology and products that try to help push things forward.
Fanflow
Connect with Fans on a premium levelFanflow helps artists afford to create new content that otherwise wouldn't exist. If you have fans who crave more from you, then premium content Fanflows may be a great addition to your free work.
Liveflows
Liveflows is a free blog personalization tool we're working on that will show each reader the most interesting, engaging and satisfying posts to read when they visit your blog. In progress.
Our Technology:
We traded batch processing strength for greater flexibility, event-driven processing and real-time results.
We've spent about three years rethinking file systems, databases, web application serving, load balancing, health checking and failover, recovery, replication, security and DR. The results so far are encouraging, with a collection of new technologies in a highly scalable and flexible system architecture.
Our design differs sharply from other horizontally scalable, non SQL solutions like BigTable/GFS and clones like Hadoop or Cassandra. BigTable/GFS is a super smart design for storing a web's worth of data several times over and using map/reduce to batch process long-running jobs of computations across Petabytes. For large jobs which process enormous amounts of data, you need to be very efficient at shipping around TB and PB of data. The key to performance is not lots of small reads and writes, or flexible ad hoc processing, but huge internal bandwidth (custom 10 GbE switches) and optimization for large sequential disk operations. Big jobs entail copying 10s or 1000s of MBs across the wire, processing on remote hosts, and returning the result to the disk clusters. Rinse and Repeat till the Job is Complete.
Our workload is vastly different, so our distributed database design is unlike others. Our system can still scale out to N-nodes, and contain PB's of data. But our workload demands we excel at small reads and writes and be able to perform ad hoc processing. Our design isn't optimized for large sequential reads/writes, but on random reads, lots and lots of relatively small writes and flexible, real-time calculations across tables that might live on mulitiple hosts. Data is expected to be personalized, and often includes complex permissions and ACLs to support varied security and commerce requirements. The structure is not one big table, but many "tables", denormalized, with the luxury of joins. File-level, instead of block level, granularity allows data to be efficiently de-normalized across multiple spindles, handle surges in popularity and return just the dataset that is needed.
This works for what we do, and it's very easy to manage.
Our Team:
tl;dr Some Stanford dudes who were in the networking space are now having
fun building
---
Ben D'Ewart, Philipp Kehler, Israel L'Heureux, Sarah L'Heureux and Chris Peiffer have been plugging away at this project for some time now.
In 2000, Israel, Sarah and Chris founded Redline Networks, which was acquired by Juniper Networks in 2005. At Redline, we were awarded a bunch of patents for novel web and internet networking technologies such as TCP connection management, our scalable web server architecture, as well as being the first to do HTTP multiplexing, and combining compression, SSL and multiplexing.
Redline's product line implemented the HTTP protocol at switch speeds. It increased capacity, speed and security for customers such as Disney, ESPN, Orbitz, Travelocity, PBS, Chase Financial, Louis Vuitton, Microsoft and hundreds more. The product worked, too: The Redline earned a a 9.5 out of 10, the highest score among 229 enterprise products reviewed by InfoWorld Magazine's Test Center during 2003.
Israel never finished his engineering degree at Stanford, but Sarah ('96) Chris ('98) and Ben ('00) got theirs done. Ben also invented base-6 beverage counting at Oktoberfest 2007 . For the win. Not making it to Octoberfest 2008 is a fail, mitigated only by a company trip to the south of France that overlapped. Philipp attended the Technical University of Ilmenau (Germany) where he studied media economics, but had more fun running underground clubs.
We're having a lot of fun on this project. It's already satisfying to make new technology and help artists earn a living online. Next, if we can give blogs and publishers sweet personalization tools, it will be better for the web, too.