Re: Performance tuning

From: Jens-S. Voeckler <voeckler@dont-contact.us>
Date: Tue, 23 Nov 1999 09:06:30 +0100

On Mon, 22 Nov 1999, Henrik Nordstrom wrote:

]Normal rules apply. Calculated straight left to right.
]
]> cache_dir_size * 2 := 134217728 [Byte]
]> store_avg_object_size := 13312 [Byte]
]> L2 := 256
]>
]> L1 = 1.1538 ???
]
]Yes, or actually 1 since only integers are allowed. The +1 is to guard
]the calculation from any rouning/truncation error (will be enought
]irregardless if the result is rounded or truncated).
]
]Put another way:
]
]Squid will use one L1 directory for each 416 MB of cache_dir space with
]the default store_avg_object_size setting of 13 KB. More L1 directories
]will only waste inodes, less will create an inbalance in how the L1
]directories are populated by Squid.

You got me. I was off by 2**10 in my calculations. So now, with your
explaination of 416 MB per CD I got to revisit my numbers. Thank you:

137438953472 / 13312 / 256 / 256 + 1
158.53846153846153846153

Using L1=128 as my first estimation - *before* seeing your formula - was
not too far off the mark. Would 192 be a safe bet, or should I go straight
for 256? In other words, is squid using steps in powers of 2 and some
(faster) and/or/shift algebra, or are numbers like 160 legal because squid
uses modulus algebra?

Le deagh dhùrachd,
Dipl.-Ing. Jens-S. Vöckler (voeckler@rvs.uni-hannover.de)
Institute for Computer Networks and Distributed Systems
University of Hanover, Germany; +49 511 762 4726
Received on Tue Nov 23 1999 - 01:17:25 MST

This archive was generated by hypermail pre-2.1.9 : Wed Apr 09 2008 - 11:57:32 MDT