Skip to main content

HZ divider effect on timer interrupt overhead

Red Hat and related distro's (like CentOS) use 1000 timer interrupts per second, per CPU core or thread (this is called the "HZ value" inside the kernel). Because this causes a lot of extra work in case of virtualization, and caused (past tense since RHEL 5.4!) problems with timekeeping, the "divider" kernel parameter has been introduced. For example, by booting with "divider=10", the kernel uses 100 timer interrupts instead of 1000, and "divider=25" means 40 timer interrupts per second.
I did a little test today to see what difference that makes when running CentOS5.4 on vSphere. Tests were done with the current 2.6.18-164.11.1.el5 x86_64 kernel in a single vCPU VM. These are the results from my test environment:
  • standard settings: HZ=1000: 60 MHz cycles used when idle.
  • divider=2: HZ=500: 36 MHz cycles used when idle.
  • divider=5: HZ=200: 25 MHz cycles used when idle.
  • divider=10: HZ=100: 20 MHz cycles used when idle.
  • divider=25: HZ=40: around 10-15 MHz cycles used when idle (results slightly variable, also for higher dividers)
  • a nohz "tickless" kernel (I used 2.6.24.7-146.ay, not currently available in RHEL/CentOS by default) used 70 MHz cycles when idle in my test environment. This is a surprise, and I don't have a good explanation (yet) !
Note that 40 is the maximum divider setting. Using 41 will show "tick_divider: 41 is out of range" in dmesg, and the parameter will be ignored.
On your own machines, you can easily check:
  • the current divider setting by looking at the currently active kernel parameters: cat /proc/cmdline . No divider parameter means divider=1, the default.
  • the measured timer interrupt rate (close approximation of HZ value): dstat -i -I timer
Gaining 20-40 MHz of CPU power on your virtualization host might not seem important, but this is a per-guest win ! Do this on all your VMs, and you could be freeing several hundreds of MHz per host, and get more VMs, better consolidation ratio's, better performance, or lower power usage.

Summary: before RHEL 5.4, divider=10 was recommended for timer accuracy. This is no longer true, but as I've shown, it still helps lowering the timer interrupt overhead. Don't forget that the ideal divider setting depends on your application: thread wake-up delays can occur in high divider scenario's, and responsiveness could potentially suffer because of that.
Labels:

Comments

Post a Comment

Popular posts from this blog

Volkswagen UHV bluetooth touch adapter & its problems

My Volkswagen car has the "universal cellphone preparation" UHV built-in. This is the main part of a car kit, but requires an additional adapter for connecting to a cellphone. At first, I was using an adapter for my good old Nokia 6310, even after I changed to the Nokia E71. Connecting was easy: pair the phone with the "VW UHV" bluetooth entity, and done. This has the phone connected to the car kit at all times, so even non-call-related functions use the car audio system (e.g. voice recognition). But progress will have its way, no matter what happens. So in comes the "bluetooth touch adapter". Instead of a phone-specific adapter, this is a small touchscreen device that slots into the UHV dashboard mount. Connecting a phone is very different now: the Bluetooth Touch Adapter connects to the "VW UHV" device via bluetooth the phone connects to "Touch Adapter" device, also via bluetooth The device doesn't allow step 2 if step 1 didn'...
12 comments
Read more

Reset lost root password on vSphere ESXi 6.7

VMware's solution to a lost or forgotten root password for ESXi is simple: go to  https://kb.vmware.com/s/article/1317898?lang=en_US  and you'll find that "Reinstalling the ESXi host is the only supported way to reset a password on ESXi". If your host is still connected to vCenter, you may be able to use Host Profiles to reset the root password, or alternatively you can join ESXi in Active Directory via vCenter, and log in with a user in the "ESX Admins" AD group. If your host is no longer connected to vCenter, those options are closed. Can you avoid reinstallation? Fortunately, you can. You will need to reset and reboot your ESXi though. If you're ready for an unsupported deep dive into the bowels of ESXi, follow these steps: Create a bootable Linux USB-drive (or something else you can boot your server with). I used a CentOS 7 installation USB-drive that I could use to boot into rescue mode. Reset your ESXi and boot from the Linux medium. Ident...
Post a Comment
Read more

GEM WS2 MIDI System Exclusive structure and checksums

MIDI is the standard for communication between electronic music instruments like keyboards and synthesizers. And computers! While tinkering with an old floppy-less GEM WS2 keyboard, I wanted to figure out the structure of their System Exclusive memory dumps. SysEx is the vendor-specific (and non-standard) part of MIDI. Vendors can use it for real-time instructions (changing a sound parameter in real-time) and for non-real-time instructions (sending or loading a configuration, sample set, etc.). In the GEM WS2, there's two ways of saving the memory (voices, globals, styles and songs): in .ALL files on floppy, and via MIDI SysEx. The .ALL files are binary files, 60415 bytes long. The only recognizable parts are the ASCII encoded voice and global names. The SysEx dumps are 73691 bytes long. As always in MIDI, only command start (and end) bytes have MSB 1, and all data bytes have MSB 0. The data is spread out over 576 SysEx packets, preceded by one SysEx packet with header informat...
23 comments
Read more