Category Archives: performance

Resource reservations and local DOS

An interesting post on the stable@ mailing list by Matt Dillon illustrates why are jail resource limits and other such measures important on todays machines. Actually, this also illustrates one of the "pro" arguments for virtualization.

Resource reservations and local DOS

UFS read-ahead

After 10 years of it being conservatively tuned, I've recently increased the default read-ahead (vfs.read_max) in FreeBSD from 128 KiB to a whopping 512 KiB. And of course, I have received an e-mail from a concerned developer asking if that is perhaps too high :)

How much impact can read_max have I'll try to illustrate on this excellent example on a machine I'm currently configuring.

vfs.hirunningspace and disk write latency performance

A while ago I increased the default value for the vfs.hirunningspace tunable - which greatly helps with performance when the disk system supports tagged queueing (e.g. NCQ), allowing many more requests to be offloaded into the controller and/or the drive(s). But deep queues bring their own problems, especially in pathological cases.

OpenJDK vs Diablo JDK performance

"DIablo JDK" is the "official" FreeBSD-foundation-distributed version of Java, presumably tested with Sun's compliance tests. OpenJDK 6 is the new-ish open source Java base, which will if recent events are any indication, soon die under Oracle. OpenJDK is a much newer "minor" version of Java and as such offers some performance gains:

OpenJDK vs Diablo JDK performance