Lazy Reading for 2011/05/01

There hasn’t been much to nab for Lazy Reading, lately.  Oh well.  The last few weeks were good so it has to even out sometime.

  • Did you know GBC stands for Great Ball Contraption, a Lego device designed to move little plastic balls?  Here’s 20 of them chained together.  (via b3ta)
  • The original University POSTGRES.   (thanks, Jan)  This is a source for PostgreSQL, as far as I can tell, which makes it in some ways contemporary to BSD’s origins.  I am not surprised.  PostgreSQL seems to be the thinking person’s alternative to MySQL like BSD is the thinking person’s alternative to Linux.
  • Do you have a pf.conf?  The people behind fwbuilder can use it for examples, so they can support pf in their config builder.  (via)
GSoC: dsched BFQ, virtio, LVM mirror

Yay acronyms! Brills Peng was accepted for the Summer of Code project “Improve dsched interfaces and implement BFQ disk scheduling policy” – and now there’s a nice writeup describing what’s planned. Also, Stéphanie Ouillon did the same thing for the virtio drivers project.  Adam Hoka also joined in with a summary for his LVM mirror project.  Please keep this up, students.

dragonflybsd.it available

If you are a European resident, Federico Biancuzzi has the DRAGONFLYBSD.IT domain name available to donate. He doesn’t want to let it go and have it taken by a domain squatter. Talk to him at sale@securitydaemon.com if you want to hold on to it for some unspecified time.

Sysbench and DragonFly releases

I did some comparative benchmarking between the 2.6, 2.8, and upcoming 2.10 release for DragonFly.   As several people have guessed, performance has improved significantly, and the difference would probably be even more pronounced if I was using more modern hardware,  e.g. swapcache or a system with AHCI.  I have a mailing list post with details, and here’s the graph that sums it up:

Shorter bars are better(Sorry, no Lazy Reading this week.  Life didn’t co-operate.  At least there’s a pretty graph!)

 

Zombie shirts

Not shirts with zombies on them, but DragonFly shirts that don’t have a seller.  I had a random Google search turn up a store selling DragonFly T-shirts, among other things.  It was essentially a spam store.  The seller wasn’t producing anything but instead reselling other people’s material for a commission, similar to the splogs out there that recopy material from other blogs or Wikipedia and slap ads on it.  (I’ve seen Digest material pop up that way.)

Following the link back shows that the shirt is sold through a Cafepress store called ossgear.  It looks like the original store owner asked permission to use the logo back in 2006.  ossgear.org is no longer a functioning domain, and I can’t find any other reference to this seller; they appear to have stopped doing business 5 years ago.

The moral of this story: Sites like Cafepress will try to profit from your work long after you’ve stopped using them.  The frustrating part is that the logo isn’t even right!

RAM vs. deduplication

Tomas Bodzar asked about RAM usage with Hammer and deduplication, pointing at this example that shows ZFS requiring…  I’m not sure.  Lots?  Anyway, Matthew Dillon noted that offline deduplication in Hammer would use available RAM/swap for CRCs on all files, but only a limited subset for ‘live’ dedup.  For a real-world example, Venkatesh Srinivas described deduplicating about 600G down to 400G, with a machine having only 256M of RAM. Yes, only 256M.

How much swapcache can help (with graphs)

I love graphs.  Jan Lentfer made some! Both of these show recent speed improvements in DragonFly – especially some spectacular results from swapcache(8) and the recent NCQ tagging improvements.  (Note that only the third graph represents the NCQ improvements; the first two graphs were done before.)

The first one is a comparison of pgbench running on the same hardware twice – once with the 2.8 release of DragonFly, and once with a recent 2.9 version.  2.9 is definitely looking to be faster than 2.8.

Next up is a 2.9 system run with and without swapcache, showing an astounding difference between the two.  It’s pretty clear just how much performance improvement you can get from swapcache…  (see Jan’s notes on the setup after the graphic.)

 

Jan’s notes, from EFNet #dragonflybsd on IRC:

15:08 < lentferj> these are SELECT-Only tests
15:09 < lentferj> JustinS: it’S important to note, that the database is 2,5x
bigger than RAM on the swapchache test
15:11 < lentferj> JustinS: I did a Select-Only ramp-up of 30 minutes to get
caches and swapcache filled
15:12 < lentferj> JustinS: and then I ran
15:13 < lentferj> for i in 1 2 3 4 6 8 12 16 24 32; do
/usr/lib/postgresql/8.4/bin/pgbench -U pgsql -h atom -s 400
-S -c “$i” -T 600 pgbench; done
15:13 < lentferj> so, select only pgbench for 10 minutes each
15:13 < lentferj> with increasing numbers of client
15:14 < lentferj> pgbench on another box, 100MBit switched network
15:15 < lentferj> JustinS: the first graph (2.8.2 vs current) is the same w/ a
database that fits in RAM entirely
15:15 < lentferj> so measuring concurrency performance (w/o I/O)
15:17 < lentferj> the swapcache comparison was on a 2GB box with a 5GB database
and 16GB swapcache (INTEL) attached to a sili card
15:17 < lentferj> on a atom 330 :)

Now, here’s testing with the recent NCQ tagging update for AHCI:

These results are astonishing.  Please, someone compare with other operating systems!

Here’s the stats for this last test:

 

  • 5.6GB database, system w/ 2GB RAM –> io benchmark
  • pgbench with increasing no of client 1->32, SELECT-Only Mode
  • sili controller Dawicontrol DC-3410 SATA PCI controller which is using a Silicon Image 3124-2 chip
  • 2 Seagate Barracuda ES.2 250GB SATA II disks
  • lvm stripe over those disks
  • postgresql.conf is default, except shared_buffers set to 512MB and effective_cache_size to 1024MB
  • atom330 on a Foxconn mobo
  • SSD is SATA INTEL SSDSA2M040 2CV1