Together with Geert and Andrew I’ll be teaching the MySQL Cluster tutrial at this year’s MySQL Cluster User Conference – Santa Clara, on April 12th. If you’re interested in using MySQL Cluster but aren’t sure how to get started (or you’ve used it but would like some tips) then this is a great opportunity. Check out the tutorial description.

If you register by 15 March then you get the early-bird price and if you use this ‘friend of a speaker’ code then you get an additional 25% off: …

A special extended edition of Tech Messages for 2010-02-20 through 2010-02-24:

• Gartner EXP Worldwide Survey of Nearly 1,600 CIOs Shows IT Budgets in 2010 to be at 2005 Levels
  Virtualization tops the technology priorities.
• Cisco Offers Three-Year, Zero-Percent Financing to U.S. Small and Medium-Sized Businesses -> Cisco News
• Marc Alff's blog: Performance schema overview
  Really interesting stuff for MySQL developers.
• …

With the introduction of each new platform, comes the opportunity for new thinking, new applications and new winners. DEC and Oracle were beneficiaries of the move to the minicomputer. Microsoft was the main beneficiary of the move to the PC. Sun rode the workstation to fame. Today’s exciting new platform is the cloud, and one of the upstart contenders is NoSQL.

One might argue that the cloud is merely the hosting of well established platforms such as the PC. Larry Ellison has made this very claim. However, the cloud is very different.

How is the cloud different? Sometimes when you combine things, the combination is very different than the components. For example, Salt (NaCl) is very different from its poisonous individual components. Cloud computing enjoys a similar combinatory effect. Sure it is merely a mixture of PC platforms, virtualization, lots of Linux and low-cost scalable disk arrays. But the combination is …

Hello and welcome to Log Buffer #180. Time’s a-wastin’, so let’s go!

Oracle

There was so much Oracle stuff this week that I’ve decided to cram a little more of it into Log Buffer by providing a little less context than usual.

Jonathan Lewis shares an explication of aliases: “I was asked the following question recently: ‘Does the use of table aliases affect performance?’ To which the best answer is probably ‘Yes, though in general you probably won’t notice the difference and there are reasons more imporant [sic] than performance for using table aliases.'”

Doug Burns continues his most recent series: Statistics on Partitioned Tables …

Martin Scholl (@zeit_geist) has started a new project based on the PBXT storage engine: EPBXT - Embedded PBXT! In his first blog he describes how you can easily build the latest version: Building Embedded PBXT from bzr.

The interesting thing about this project is that it exposes the "raw" power of the engine. Some basic performance tests show this really is the case.

At the lowest level, PBXT does not impose any format on the data stored in tables and indexes. When running as a MySQL storage engine it uses the MySQL native row and index formats. Theoretically it would be possible to expose this in an embedded API. The work Martin …

PMP on demand revealed one of reasons why we’ve been seeing ‘Opening tables’ during proper operations, not just during startup (see my previous post on this topic).

We had a thousand or so threads waiting on LOCK_open, and the only thread holding the mutex was this darling:

Thread 902 (Thread 1637624128 (LWP 22113)):
#3  mutex_spin_wait (mutex=0x2aaaac3232b8,
        file_name=0x8b3bf7 "trx0trx.c", line=220) at sync0sync.c:565
#4  trx_allocate_for_mysql
#5  ha_innobase::allocate_trx
#6  check_trx_exists
#7  ha_innobase::info
#8  ha_innobase::open
#9  handler::ha_open
#10 openfrm
#11 open_unireg_entry
#12 open_table
#13 open_tables
#14 open_and_lock_tables
#15 mysql_insert

So, kernel mutex, which is quite contended, will escalate to LOCK_open on table open, which will block all queries from …

Columnar DBMS tuning may require some new approaches versus traditional row-based tuning; gone are full/partitioned table scans, index lookups, and nested-loop join operations, instead replaced by full/partitioned column scans, access by RowID, and hash join operations.  This series of articles is intended to measure potential benefits of different approaches to jump-start your expertise with this new paradigm.  Today’s topic is dRead More...

We just noticed that the libmysql.dll shipped with the 4.0 and 5.0 beta releases does not work as expected with the above mentioned SSL keys. So, if you run into some access violation in such cases, you will need a fresh one, can by copied from a MySQL/Win32 server installation, located in the "bin" directory.

However, I will release a second Beta as soon as possible so you can forget my above notice soon.

(This post was written at my old location.  You’re now reading it at the new location. )

I’m in the process of moving my blog to hosted WordPress at WordPress.com. Much as I loved the flexibility of hosting my own blog (and the rest of the company), the upkeep required just isn’t worth the extra flexibility. So my customization will go away, which is sad, but I should have to spend less time managing all the blogs at SmugMug. Which might mean I actually blog more often again.

Here’s the new URL in ‘beta’ form: http://don.blogs.smugmug.com/

I should have redirects up shortly, so everything should work fine, but if your feed reader breaks or something, now you know why. Apologies in advance for any hiccups…

And, of course, I …

Running many benchmarks on fast storage (FusionIO, SSDs) and multi-cores CPUs system I constantly face two contention problems.

So I suspect it's going to be next biggest issues to make InnoDB scaling on high-end system.

This is also reason why in benchmarks I posted previously CPU usage is only about 50%, leaving other 50% in idle state.

First problem is index->lock mutex.
InnoDB uses single mutex per index, so when you run mixed read / write queries, InnoDB locks index for write operation and thus keeps all selects waiting when update/insert is done. This is implemented in this way because write operation may cause B-Tree page split, and InnoDB needs to move records between pages to finish operation. It is getting even worse when for write you need to perform some additional IO to bring page into buffer_pool.

What could be done there internally: there is B-Tree lock free or only page-level lock …