Steve Gillmor and I are here at the Sun Campus in Menlo Park, CA for a conference on Open Source Business Intelligence. Zack Urlocker is the VP of the Database group at Sun, he came to the company via MySQL.
12:40 – Zack came to Sun through MySQL. MySQL were planning to IPO, but they were interesting in what Sun were doing. Says that Sun has come a long way in the past 5 or 6 years as being a strong supporter of open source and a very strong advocate – which is what MySQL found attractive. Zack says that 5 years ago it was difficult to imagine Sun open sourcing Java or Solaris, but those moves sent a strong message that Sun were very serious about open source.
12:43 – The Sun acquisition of MySQL extends suns reach into the LAMP stack, better positions Sun as a web platform and Sun’s resources assists MySQL’s capacity to delivery solutions and support.
12:45 – Zack spent a lot of time recently tlaking to large enterprises, and has noticed that the trend towards open source within the Fortune 100 has shifted dramatically recently. MySQL refer to the trends as Enterprise 2.0, and it is a very real trend and it has driven a big increase in revenue at MySQL.
- 12 years old
- 400+ employees
- 750 partners
- 70,000 downloads per day
- Customers across every major OS, hardware vendor, geography, industry and app type.
- MySql were at $100M revenue when they were acquired by Sun
12:47 – MySQL were optimized as a web application database. They never set out to comepte head to head with Oracle or the large RDBMS providers. MySQL set out to be a disruptor, and they found their niche in the web area, just as web applications and the open source web stack was being formed and usage was accelerating.
12:49 – MySQL marketing has a very strong emphasis on case studies and references. A very strong tip for similar companies looking to sell their solutions into the enterprise. Rather than the fat Oracle approach of RFP’s and large-scale sales teams, have solid real-world and public examples, and easy entry point (Eg. the free download) and the product, if good, will sell itself. Deployment stats:
- Average deployment is 2-3TB
- Largest deployment is 10,000 servers (its not Google)
- The largest deployments employ sharding
- 35% of users of Oracle users are running MySQL as well side-by-side, the penetration is higher than Oracle Express
12:55 – Case studies!
- 4 years of historical ticket data
- 800 million records, half a terabyte of live OLAP data
- 4 server greenplum cluster
- Most queries under 8 seconds
- Self service portal for travel agents with integrated reporting
- 2500 users with contract renewal, ordering and reporting
- Using Redhat, MySQL, Pentaho
- $1M in TCO savings
- Embedded BI reporting
- Intrusion detection appliance used in F500 companies, governement
- Logging millions of events per day for 3000 customers
- Using debian mysql and jasperreports
- Requires highly scalable custom reporting
1:00pm – IDC is reporting that data wharehouses are growing, but not as large as the press reports. Gor eg. 60% of data warehouses are under a terabyte, and only 4% are larger than 25TB – so it is a classic ‘pyramid’ market, so it is definitely not a one-size-fits-all market, with most of the market being at the lower end (where MySQL and Postgres are used)
1:07 – The key architectural advantage of MySQL is that the storage engine is a plugin, with a common base, interpreter, optimizer etc. It means that on a case-by-case basis an architect can select the appropriate storage engine depending on their requirements. The built in engines cover most case, but there are numerous commercial and open source storage engines available varying from the simple such as CSV file support through to column-oriented datasets.
1:15 – Long conversation about NDBCLUSTER, which is the storage engine now known as MySQL cluster – which is a shared-nothing architecture that allows high availability and scalability. Lots of discussion about the difference between what MySQL are doing, and what companies like Facebook are doing with building a small-scale SQL engine running on top of Hadoop. Yahoo! are also running a very large-scale Hadoop instance, and are building a light-end SQL engine on top of it. NDBCLUSTER is currently running in numerous Telco datacenters – such as at Vodafone, France Telecom, Nokia etc. etc.
1:20 – MySQL and the open source business model – enterprise server, monthly rapid updates, hot fix programs, monitoring, web-based central console, expert advice, 24/365 support – this model is working really well for MySQL, it took a while for them to get there but with a $100M annual run-rate it is a real working open source based business model. MySQL is on average 90% cheaper than Oracle, Sybase etc.
1:25 – Oracle just announced a 15% across-the-board price increase. MySQL charge $40k per year, per customer – regardless of CPU, usage, size etc. etc. They are commiting to a flat-rate and that pricing structure. MySQL have a very capable product and a strong economic argument, and they are making very strong inroads into the large-scale enterprise space.