Approach problems by accepting the inherent characteristics of distributed systems and build those into the abstractions they provide as tools. Oracle, is a highly available database that letвЂ™s you pretend the world is different than it actually it is.Д«oth solve real problems, but both are expensive, so do something different. NetApp, for example, creates the illusion of an infinite disk. Create abstractions that let you pretend things werenвЂ™t actually distributed.

By definition if you have independently operating systems in different regions they have to be distributed. To achieve certain levels of availability you must have geographic diversity. The end of the talk is cut off, but itвЂ™s mentioned bitly uses quite a few different databases.ДЄnything that is truly highly available will be inherently distributed. Д«itly uses hostpool to manage a pool of hosts. The real time distributed messaging system bitly uses is Nsq. Note, that these were just what was mentioned in the talk, itвЂ™s not a comprehensive list.

LESSONS LEARNED BUILDING DISTRIBUTED SYSTEMS AT BITLY The rest of the 400 are dedicated to various services either storing and organizing user data or providing various forms of processing and analysis (databases, metrics, web crawling, text analysis, etc).Д¬rawls a hundred million web pages a month. ~30 servers handle all incoming traffic from the outside world including shortens, redirects, api requests, web ui, etc. Not all 400 servers are handling redirects. StatsДҐ0 people in the company and approximately 20 engineersД¤00 servers. So letвЂ™s see what Sean has to say about distributed systems. If you are trying to make a jump from a single box mindset to a multibox way of thinking, this talk is well worth watching. Sean talks from a place of authentic experience. I also really like his explanation for why event style messages are better than command style messages. This approach isolates components, lets work happen concurrently, lets boxes fail independently, while still having components be easy to reason about. SOA + queues + async messaging is really powerful. Perhaps my favorite lesson from his talk is this one (my gloss): While this talk is not on bitlyвЂ™s architecture, it is a thoughtful exploration on the nature of distributed systems and how you can solve bigger than one box problems with them. The idea of making links go to a different domain than your own so that a 3rd party can do the analytics is a scary proposition, but itвЂ™s also kind of genius. But the links all went back to the domain of the web site. Logs contained data from annotated links along with cookie data to indicate where on a page a link was clicked, who clicked it, what the link was, etc. Writing a URL shortner that works is easy, says Sean, writing one that scales and is highly available, is not so easy.Д«itly doesnвЂ™t make money with a Shortening as a Service service, bitly makes money on an analytics product that mashes URL click data with with data they crawl from the web to help customers understand what people are paying attention to on the web.ДЄnalytics products began as a backend service that crawled web server logs. All new request would be taken from the cache.Have you ever wondered how bitly makes money? A URL shortener canвЂ™t be that hard to write, right? Sean O'Connor, Lead Application Developer at bitly, answers the how can bitly possibly make money question immediately in a talk he gave on bitly at the Bacon conference.Once the token is generated it would be stored in the cache (redis) and database server.For security reason we should also ad some randomness in the generated url.When the token are exhausted we need to have new set of of tokens and range also should be changed.To avoid collision the token generation should be distributed with specific range to each server.When the requests per sec are in thousands we cannot rely on a single server generating the token.Token generation should be distributed across different servers.We need to use token generation technique for generating tiny url.Load balancer sends the request to the zookeeper instance.When a long url is generated it is sent to the load balancer.ex 41 bitsДў) Allowed characters on the tiny url could be caps and smaller alphanumeric characters The basic rule of thumb is that we need to convert the long url into tiny url and store it in database or cache for future retrieval.ДЎ) Sometimes we would have restrictions on the tiny url. We have different options to generate tiny url.

Tinyurl system design