Parallelism Update

It's been over a year since I last blogged about parallelism, so I think I'm past due for an update, especially because some exciting things are happening.

First, Amit Kapila has published a draft patch for parallel sequential scan.  Many things remain to be improved about this patch, which is neither as robust as it needs to be nor as performant as we'd like it to be nor as well-modularized as it really should be.  But it exists, and it passes simple tests, and that is a big step forward.  Even better, on most of Amit's tests, it shows a very substantial speed-up over a non-parallel sequential scan.

Memory Matters

Database performance and hardware selection are complicated topics, and a great deal has been written on that topic over the years by many very smart people, like Greg Smith, who wrote a whole book about PostgreSQL performance.  In many cases, the answers to performance questions require deep understanding of software and hardware characteristics and careful study and planning.

But sometimes the explanation is something very simple, such as "you don't have enough memory".

Linux disables vm.zone_reclaim_mode by default

Last week, Linus Torvalds merged a Linux kernel commit from Mel Gorman disabling vm.zone_reclaim_mode by default.   I mentioned that this change might be in the works when I blogged about attending LSF/MM and again when I blogged about how the page cache may not behave quite the way we want even with vm.zone_reclaim_mode disabled.

For those who haven't read previous discussion on this topic, either on my blog, on pgsql-performance, or elsewhere around the Internet, enabling vm.zone_reclaim_mode can cause a lot of problems for applications, such as PostgreSQL, that make use of more page cache than will fit on a single NUMA node.  Pages may get evicted from memory in preference to using memory on other nodes, effectively resulting in a page cache that is much smaller than available free memory.  See the second of the two blog posts linked above for more details.

PostgreSQL isn't the only application that suffers from non-zero values of this setting, so I think a lot of people will be happy to see this change merged (like the guy who said that this setting is the essence of all evil).  It will doubtless take some time for this to make its way into mainstream Linux distributions, but getting the upstream change made is the first step.  Thanks to Mel Gorman for pursuing this.

Troubleshooting Database Corruption

When your database gets corrupted, one of the most important things to do is figure out why that happened, so that you can try to ensure that it doesn't happen again.  After all, there's little point in going to a lot of trouble to restore a corrupt database from backup, or in attempting to repair the damage, if it's just going to get corrupted again.  However, there are times when root cause analysis must take a back seat to getting your database back on line.

Why The Clock is Ticking for MongoDB

Last month, ZDNet published an interview with MongoDB CEO Max Schireson which took the position that the document databases, such as MongoDB, are better-suited to today's applications than traditional relational databases; the title of the article implies that the days of relational databases are numbered. But it is not, as Schireson would have us believe, that the relational database community is ignorant of or has not tried the design paradigms which he advocates, but that they have been tried and found, in many cases, to be anti-patterns. Certainly, there are some cases in which the schemaless design pattern that is perhaps MongoDB's most distinctive feature is just the right tool for the job, but it is also misleading to think that such designs must use a document store. Relational databases can also handle such workloads, and their capabilities in this area are improving rapidly.

Subtly Bad Things Linux May Be Doing To PostgreSQL

In addition to talking about PostgreSQL at LSF/MM and Collab, I also learned a few things about the Linux kernel that I had not known before, some of which could have implications for PostgreSQL performance.  These are issues which I haven't heard discussed before in the PostgreSQL community, and they are somewhat subtle, so I thought it would be worth writing about them.

Back from LSF/MM and Collab

Last week, I attended the Linux Storage, Filesystems, and Memory Management summit (LSF/MM) on Monday and Tuesday, and the Linux Collaboration Summit (aka Collab) from Wednesday through Friday.  Both events were held at the Meritage Resort in Napa, CA.  This was by invitation of some Linux developers who wanted to find out more about what PostgreSQL needs from the Linux kernel.  Andres Freund and I attended on behalf of the PostgreSQL community; Josh Berkus was present for part of the time as well.

My overall impression is that it was a good week, except that by Thursday the combination of 14 hour days and jet lag were catching up with me in a big way.  However, from the point of view of the PostgreSQL project, I think it was very positive.  On Monday, Andres and I had an hour-and-a-half slot; we used about an hour and fifteen minutes of that time.  Our big complaint was with the Linux kernel's fsync behavior, but we talked about some other issues as well, including double buffering, transparent huge pages, and zone reclaim mode.