An overview of today's performance boundaries, and current and upcoming technologies that are breaking them.

Maria is a new storage engine for MySQL. Monty will explain Maria's special features targeted towards performance, scalability and reliability, including a new type of indexes. He will give you the technical foundation to understand its current and upcoming advantages over other storage engines.

Gearman is back, now written in C with high performance, persistence, replication, new client APIs, and features for building anything from websites to map/reduce clusters.

How do you scale CSS for millions of visitors or thousands of pages? Object Oriented CSS will allow you to write fast, maintainable, standards-based front end code. It adds much needed predictability to CSS so that even beginners can participate in writing beautiful, standards-compliant, fast websites.

Slave lag is the typical problem that limits scalability of MySQL Replication. Learn best practices for scalable, lag-free MySQL Replication.

Learn how to build for large scale without throwing complex solutions at simple problems. Learn monitoring, trending, replication, federation, and tools for Postgres.

Learn the pros and cons of Patricia Tries in disk-based index storage, with benchmarks from ScaleDB showing results for multi-table indexing.

Understanding the theoretical IO performance of disk arrays, the actual IO performance of your disk array, and how to narrow the gap. The typical default controller settings are counterproductive in an OLTP environment, but the arrival at that conclusion is very counter intuitive. I'll offer a few silver bullet tuning imperatives for any controller, with examples on a few real life controllers.

How to use the new diagnostic tools released in Percona and OurDelta builds, including the improvements to the slow query log and index_statistics.

SQL proxies keep routing and failover out of the application, making them useful for scale-out designs. Learn standard proxy designs including pushing proxy functions back into client libraries with a discussion of “proxy math” for performance. We'll illustrate the trade-offs with performance numbers from benchmarks on Amazon showing that bad proxy strategies can *quadruple* the number of CPUs required in deployed systems.

How to design covering indexes for various schemas (star, snowflake) that can speed up queries by orders of magnitude. Learn performance models including point-query and range-query. B-Trees and Fractal Trees, and benchmarks.

QA cycles sound great, but they introduce a problem. Avoid the cyclical performance panic and debugging of a month's worth of commits when your new features get through QA and hit production with severe performance regressions.

EMT ties together information from your existing monitoring and status tools. It lets you store information locally or send it off to centralized monitoring servers for aggregation, graphing and other uses. Eric shows how EMT works and what it can do for you.

Galera is a multi-master synchronous replication system based on Database State Machine research. Except for the commit stage, it looks to the application just like a single MySQL server. Seppo shows results for sysbench, DBT2 and Drupal benchmarks on MySQL/Galera cluster, write-intensive applications, conflict resolution, EC2 tests, and scalability when adding nodes.

Social Applications can generate a significant amount of traffic and revenue, but to scale applications on both platforms requires very different approaches. This talk describes the differences, what resources are located where, certain pitfalls, and how to protect the datalayer from explosive growth.

New servers, old servers. What should you be doing on both? Learn things that should be set up and forgotten, things that need to be done routinely to get the best performance from your servers, and performance reviews.

Still waiting for online, non-blocking InnoDB backups? The wait is over. Vadim discusses the usage and architecture of the new open-source InnoDB backup tool.

Sphinx's strengths complement where MySQL is weak: advanced search, fixed-memory GROUP BY and sorting, sequential I/O, running multiple queries in a single execution pass, seamless distribution across servers, and scalability on multiple CPUs. Learn all about it.

Lessons learned from benchmarking, performance analysis, and interpreting filesystem performance data.

Application architectures have not kept up with the growth in memory size. This talk provides an overview of different application classes and how they can be re-architected to achieve better performance with new storage-class memory subsystems.

Calpont is an open-source columnar storage engine designed to scale horizontally across many machines. Learn how it performs distributed
scan, distributed aggregation, and distributed join operations.

Databases always grow. With partitions, MySQL 5.1 becomes an efficient and easy to use data warehouse. This session explains with practical examples how to create and use MySQL partitions.

Apache CouchDB is a distributed, fault-tolerant, schema-free, document-oriented database with bi-directional replication. It is unlike traditional databases, yet it can answer many questions that are asked in web applications. Learn what is behind the buzz.

Gil has built large systems that use every Amazon service there is. Explore practical scaling solutions with EC2, S3, SimpleDB, and SQS.

Learn about ext2, ext3, ext4, reiserfs, XFS, JFS, and BTRFS. Learn how each works, performance in different workloads, reliability, support, and available tools.

Today's Web 2.0 and cloud computing datacenters severely underutilize the tremendous technology advances of recent years. Much potential is untapped, and scaling, performance, power, space, and complexity issues remain. This presentation examines these leading edge technologies, and discuss the challenges and opportunities in taking advantage of their promise.

Ask performance questions to the experts.

The now-classic book The Innovator's Dilemma introduced the term “disruptive innovation.” Baron discusses where disruptive innovations are happening today in the MySQL products and services arena, and where they are likely to happen in the future.

The interaction between the application and the database is a vulnerable place where even small mistakes can amplify into business-crushing performance problems. Most such problems are simpler to solve than people make them. What information should you ignore? What must you insist on having? How can you get it? Learn from the legendary Oracle performance expert.

Don MakAskill founded and built SmugMug, a popular photo sharing site. He is known for demanding very high performance and exploring the edges of what's possible. Learn what happened when he pushed the application, database, hardware and OS beyond its limits, and how he solved those problems.

Mark Callaghan is Google's resident MySQL expert. He also has many years of background in database internals at Oracle and Informix. In this talk he explains how InnoDB does disk I/O and what can be done to improve it.

Understanding storage engine internals is the key to performance tuning. Learn the components of PBXT and how to monitor them with the xtstat profiling tool. Learn how recent performance improvements in the code base tripled performance on memory-bound DBT2 tests.

Hive is a sub project of Apache Hadoop that builds data warehousing on a distributed file system and a map/reduce engine, with a SQL front end to query it. It is used at Facebook for reporting, data mining and analytics. It stores over 300TB of compressed data and loads more than 1TB of compressed data daily. Learn Hive internals, how it is used at Facebook, and how it compares to systems that solve similar problems.

Theo built a multi-terabyte instance of Postgres in a high volume, mission-critical operational datastore that replaced Oracle. Learn about solving real-life problems such as a near-catastrophic hardware failure at terabyte level.

Pagination is common pattern for most web based applications, most often developers use LIMIT OFFSET, NUMBER mysql specific sql to paginate and get very slow response as user paginate to deep pages. In this talk Surat will provide solutions to write efficient MySQL query to paginate through large data set.

The usual solution for complex queries on large databases is big SMP machines or large MPP clusters with expensive storage systems. This talk explains Kickfire's efforts to reverse this trend. We will discuss specific hardware and software optimizations in our storage engine, data structure, memory, hardware accelerator module, and disk I/O subsystems. We will also share our experiences with the TPC-H benchmark.

Everyone loves to use sexy buzzwords like sharding. This talk is about how to get 10x more performance on budget hardware through smart and utterly non-groundbreaking design.

A whirlwind tour of how to get Sphinx up and running, and get your application to query it.

A walk through of how oProfile works, when you might want to use it as opposed to other available tools, and then a look at specific examples of problems diagnosed and fixed using oProfile.

LucidDB is an open-source column-store database built for analytics from the beginning, not as an afterthought. John will show recent Pentaho Kettle and Mondrian components that can extract and preaggregate data.

Hypertable is an open source, high performance, distributed database modeled after Google's Bigtable. The design emphasis is on scalability on commodity hardware, and it supports massive sparse tables. This presentation will include an architectural overview, a demo of the system, performance studies, and some details of Baidu's deployment.

PostgreSQL 8.4 will have SQL:2008 Common Table Expressions. You can use them to wrangle trees and other recursive structures directly in your database management system, saving lots of time and aggravation.

MySQL and Search at Craigslist

Jay will describe scalability and performance improvements to the Drizzle server. He'll highlight refactoring efforts around the core kernel, and discuss the Drizzle community's transparent development practices and open benchmarks of the Drizzle server.

Physical disk drives and RAID controllers are widely studied and well understood. What about I/O performance in a virtual environment? Florian shows you what you don't know about block device performance.

Learn the most important concepts for tuning your application and database server to maximize InnoDB performance. Peter's practical approach shows you which 20% to focus on to get the most performance with the least effort.

You've optimised everything. Your app responds in a few milliseconds, but users still see slow pages. You forgot that your web app has a client side component that runs in the user's browser. Learn rules to improve end-user perceived performance of your web application. We'll look at tunable parameters, ways to optimise front end code, and how to find out what needs fixing with YSLow.

Pathtraq is one of the largest web access statistics services in Japan. Its core database runs on a single MySQL server and simultaneously provides analysis information to the users semi-realtime. Learn how to effectively compress URLs, use UDFs to load, store, and aggregate compound variables, schedule periodical tasks with message queues, cache control, and more.

Learn how to avoid common mistakes, drill down to understand the true performance impact, and measure performance gains from corrections. Practical examples on configuration, schema, application code changes. Real-life case studies show disk space reduction, increases in throughput, and reduced query execution times.

In this session we'll look at how to securely backup your data without stopping the world, and then store it away and hope you won't need to use it.
We'll walk you through the use of LVM, how to set it up, LVM snapshots offloading the data and encrypting it.

Telecoms are usually built on low-level technologies for performance. Erlang is a high-level distributed programming language. This talk presents a systematic, research-based comparison of Erlang to other telecom technologies, demonstrating that it is high-performance, resilient, fault-tolerant, scalable and dynamically reconfigurable. But most importantly, it is highly productive to work with Erlang.

Whether you have one MySQL slave server or 50, it's a challenge to use them efficiently in your application. Learn how to serve as much load from the slaves as possible, while preserving a consistent user experience and avoiding serving stale data.

Lightning Talks will be scheduled during the conference.

Come talk to the open-source leaders who are building the next generation of software. Let us hear the features you've always wanted but you've been afraid to ask for.