Add processors, reduce downtime

As part of my work on improving availability by reducing downtime, I'm doing a detailed characterization of the system recovery time using the System Recovery Benchmark for hardware and OSes (SRB-A)[1]. The Solaris 10 OS added the Service Management Framework (SMF) which will allow the startup of OS services to occur in parallel at boot time. In previous Solaris OS releases, the OS services were started serially.

In my testing, I see a 57% reduction in the Solaris service startup time by going from 1 to 2 processors or threads. I measure this by noting when the init process (process ID = 1) begins until I get a login prompt. Adding more processors or threads allows more work to be done in parallel during boot. Going from 1 to 4 processors or threads provided me a 65% reduction in service startup time. While this isn't as dramatic as going from 1 to 2 processors or threads, it does imply that a relatively generic Solaris installation can manage 4 concurrent services starting. There is no inherent per-process limit here, just the needs of the services are not completely parallel. SMF manages dependencies and some services depend on others, hence the serialization.

In short, you can add processors or threads to reduce downtime with the Solaris OS. Cool.

[1] James Mauro, Ji Zhu, and Ira Pramanick, “The System Recovery Benchmark,” IEEE Pacific Rim International Symposium on Dependable Computing (PRDC 04), 2004.

Comments:

Post a Comment:
Comments are closed for this entry.
About

relling

Search

Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today