[69] in 6.033 discussion
Gray & Siewiorek typo
Saltzer@ATHENA.MIT.EDU (Saltzer@ATHENA.MIT.EDU)
Wed Apr 6 23:10:16 1994
If you were wondering how to make sense out of the equations given in Table 2 of the Gray and Siewiorek paper, the enclosed correspondence may be of interest. The book that Gray mentions is the authoritative treatment of failure recovery and coordination, strongly recommended if you ever find you need to do something serious in either area.
----------------------------
From: Saltzer@MIT.EDU "Jerome H Saltzer" 5-APR-1994 11:58
To: <sfbay::jimgray>
Subj: typo?
Jim,
We have our class reading your paper with Siewiorek, "High-Availabilitiy
Computer Systems," from Computer, September 1991. Some of the students
have asked about the equation column of table 2, especially for the Triplex
plus repair architecture. The formula given there doesn't follow from any
intuitive model of what is going on, and what is worse, it predicts an MTTF
of only about 750 years, which is not what one would expect, nor is it what
it says in the MTTF column.
So we are assuming there is a typographical error. But there is some
debate on just what equation you might have had in mind.
By chance could you enlighten us on what the original manuscript said,
before the printer had a chance to reset it?
Thanks.
Jerry
-----------------------------
Date: Tue, 5 Apr 94 14:03:55 PDT
From: <sfbay::jimgray>
To: saltzer@MIT.EDU
Subject: yes! Thanks! RE: typo?
Right, thats a typo I hoped we fixed but I guess not.
You can find a derivation on page 113 of "Transaction
Processing Concepts and Techniques", Gray & Reuter
Morgan Kaufmann, 1993. To save you time and money,
the triplex equation should have MTTR**2. This makes
the units work out (a simple check) and matches the
derivation. Using Figure 3.14 of the book, and extending
it with backarrows (to show repair) is a good way to
"prove" the result.
-------------------------------
