POSETTE CFP is closed. Thanks to all submitters. Remember to save the date for POSETTE: An Event for Postgres 2024 📌
Matthew Kelly, TripAdvisor's in-house PostgreSQL guru, spoke to us recently about PGConf Silicon Valley which is November 17-18 at the South San Francisco Conference Center. Early Bird pricing ends October 4th.
Matthew's talk at PGConf Silicon Valley 2015 is At the Heart of a Giant: Postgres at TripAdvisor. Matthew said the PostgreSQL database server has been at the core of the company's infrastructure since the "early days as a scrappy start up" 15 years ago.
PostgreSQL is the database powering TripAdvisor’s 250 million reviews from inquisitive travellers around the world. With more than 375 million unique monthly visitors, TripAdvisor is the world's largest travel site and has been helping travelers plan and book the perfect trip since 2000.
Matthew joined TripAdvisor first as an intern during his computer science studies at the University of Massachusetts Amherst and was hired full-time following his graduation in 2012. I sat down with him virtually recently and here's what he had to say…
Terry: How is Postgres used at TripAdvisor today?
Matthew: Postgres is the backbone of our entire core production infrastructure here at TripAdvisor. On any given day, our Postgres infrastructure throughout the company might have serviced more than 1 billion queries. These will range from the OLTP sort of operations you would expect from a high end website to analytics to real time anomaly processing to fraud detection. We even use it to hold the photos that our in our upload queue. We run numerous different hardware configurations and workloads. Sure we have Hadoop for some back office processing but Postgres really is the primary workhorse of our data infrastructure.
Terry: TripAdvisor gets a massive amount of traffic. Similar companies operating at that scale are using or considering WebscaleSQL – which as you know is a collaboration among engineers from several companies facing similar challenges in running MySQL at scale. How does PostgreSQL stack up against it in terms of forming the backbone of a popular site like yours?
Matthew: As I mentioned previously, we are able to leverage PostgreSQL to build at really massive scales. And I will be the first to admit that we haven't really leveraged the server to its full potential.
I was just browsing through WebscaleSQL's commit log and many of those commits look like they belong in MySQL core. It seems very telling to me that companies with such strong commitment to open source find it necessary to maintain their own branch.
Conversely, I think one of the best reasons to choose PostgreSQL is the vibrancy of the open source community built around it. I've only been involved in the community for a short time now but even so I've seen significant growth in the community and in conference attendance. And one only needs to look at the line count in the diff between the 9.4 and 9.5 branches (~435,000 lines) to realize that PostgreSQL is under very active development. I've dabbled involvement in a few open source communities, and its fairly clear to me that the PostgreSQL community is one of the strongest and best lead communities and that it has some of the most discipline development practices.
Terry: Your session is titled, "At The Heart Of A Giant: Postgres At TripAdvisor." What’s the most exciting part of your talk to you? And who should attend?
Matthew: My talk is mostly tailored at people who are skeptical of how well Postgres is capable of scaling up to the level of load which is experienced by a high end consumer website. However, the part I'm most excited about for this conference is something we were just beginning to explore last time I gave this talk. We've recently built the infrastructure to use PostgreSQL as a multi-mastered high performance key value store and are currently using that in production to help back our online user personalization and machine learning.
Terry: Aside from your own session, what other sessions/speakers are you most looking forward to attending?
Matthew: I think the talk I'm most interested in hearing is Peter Geoghegan talk on advanced usages of Postgres's new "UPSERT" functionality. The syntax that was chosen "INSERT [...] ON CONFLICT DO {NOTHING|UPDATE}' actually strikes me as more powerful than necessary so I'm really interested in the sorts of things Peter is imagining as use cases.
Terry: Thank you for speaking with us, Matthew.
Register for PGConf Silicon Valley by October 4, 2015 and enjoy Early Bird Rates. Use discount code CitusData20 for an additional 20% savings off the current prices. We hope to see you at PGConf SV!
