Commit 22fbade5 authored by Mike Hibler's avatar Mike Hibler
Browse files

Updated numbers including the pc3000s

parent 61465167
Thoughts and initial work on using digest/hashing techniques to improve
frisbee performance.
Last updated 10/7/05 with pc3000 numbers.
Create a "signature" file for an image using a collision-resistant hash
like MD5 or SHA-1. When imageunzip (or frisbee) run, they first load
in the signature file and use that information to check the current
......@@ -19,12 +21,14 @@ we must meet two criteria:
processor, memory and disk speeds but is in general true. For our
machines, here are some numbers (all MB/sec):
type read(1/6) write(1/6) md5 inflate
type read(1/6) write(1/6) md5 inflate sha1
pc600 22.1/18.4 19.6/19.6 71.6 32.5
pc850 26.7/28.0 21.4/21.4 86.4 45.1
pc2000 43.4/43.4 38.9/38.8 242.1 63.2
The two numbers for read/write are numbers for reading 1GB of data
pc3000 80.0/75.5 76.3/73.6 (WCE) 360.7 90.1 125.5
The two numbers for read/write are numbers for reading/writing 1GB of data
starting at the beginning of the disk (i.e., the first GB) and starting
at 5GB (i.e., the 6th GB). They show a little of the slowdown as we
get out further on the disk (We mostly live in the first 6GB).
......@@ -34,6 +38,13 @@ we must meet two criteria:
latter involves not only a memory read for every byte, but multiple
writes as well.
Note that pc3000 reflects resonably current state of the art:
3Ghz processor, 2GB of 800MHz memory, 10000K RPM SCSI disk with the
write-cache enabled. Note also the relative rate of hashing vs.
decompression gets considerably less favorable when using SHA-1 hash
instead of MD5. Since MD5 has been "compromised" this may be a concern
at some point in the future.
I have already written an "imagehash" utility, that can create the
signature files (MD5 or SHA-1) and can be run to check the signature
vs. a disk. Currently it creates a 16 (or 20) byte hash for every 64KB
......@@ -48,14 +59,26 @@ time to load the disk (in seconds):
pc850 82.3 65.4 20.5% 87.3 (68.5 + 18.1)
pc2000 68.0 44.9 34.0% 55.8 (48.7 + 6.5)
pc3000 94.9 37.4 60.6% 37.5 (31.3 + 5.6)
"imagehash serial" shows a run without overlapping reading with hashing
along with the broken out time for those two phases. Note that there are
some bizarre effects here that I don't yet understand: 1) pc850s are slower
than pc600s to read the disk when serially imagehashing, yet the disks
are faster, 2) pc850s and pc2000s show extremely good overlap of IO with
hashing, which pc600s show almost no improvement (85.1s to 81.1).
hashing, which pc600s show almost no improvement (85.1s to 81.1). Note
that you cannot compare the pc3000 times with other node types since it
was run with a much different image. The 60% savings is probably exaggerated
since it appears our server cannot keep frisbee fed and it goes idle:
net thread idle/blocked: 0/0
decompress thread idle/blocked: 34810/0
disk thread idle: 738
Note that the net thread is never idle, so data is coming in, just not
fast enough. But that is another issue...
Anyway, we do see that imagehash is faster than frisbee by 13-34%.
Anyway, we do see that imagehash is faster than frisbee by 13-34+%.
However, in that saved time, we need to be able to transfer over, decompress,
and write any actual changes. How much of this standard frisbee action
we can overlap with the hashing is important. If we have to do the
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment