edj@aramus.ksr.com (E. Douglas Jensen) (02/03/88)
[ This is in response to a query that I sent to Doug about the state of the ] [ Alpha system. --DL ] We did not intend to support the first generation (existence proof) version of Alpha outside CMU. However, because Alpha is so capable and effective, we are under great pressure from our sponsors to begin transition of the technology into the real-time control application domain. That has the advantages to us of getting feedback from real industrial users, plus facilitating further funding for continued research and development of Alpha. But of course I am trying to very carefully control the export of CMU Alpha since it does require considerable time on our part to educate and support its users. The second-generation "industrial strength" Alpha has additional features (e.g., UNIX interoperability) and will be portable and supported by Kendall Square Research. Release 2 will occur in 9/89 and release 3 in 9/90; 3 will have very extensive and elaborate facilities for fault tolerance, an unusual object store, and other features not currently found even in research OS's. But again I am under great pressure to have a supportable interim version of CMU release 1 until 9/89's release 2 of the new version; I have not decided what to do about that yet, except that I expect to subcontract to someone to handle most of that work since I don't have the manpower at either CMU or KSR to do more than supervise it. You can read about the prototype of Alpha's kernel in the thesis of one of my PhD students: Northcutt, Mechanisms for Reliable Distributed Real-Time Operating Systems, Academic Press, 1987. Feel free to contact me for further information about Alpha. Doug
zenith@uunet.UU.NET (Steven Zenith) (11/20/88)
Re posting from from: dls@mace.cc.purdue.edu (David L Stevens) Date: 19 Nov 88 19:15:59 GMT > Suppose you have a network of workstations and you map the virtual >address space of each by page faults that result in network requests to one >or more servers. I'm looking for algorithms to manage the "global" virtual >address space of the network, and pieces in use by the individual machines. > > Applications: shared libraries, shared "in core" binaries, shared >"in core" filesystem data, inter-machine process migration, etc etc. This appraoch to global addressing seems like incredibly hard work and fraught with difficulties (maintaining integrity etc..). If you really need this kind of solution in the mentioned application areas, I would strongly suggest you look at the `Linda' work done at Yale. The first paper to read is: Generative Communication in Linda by David Gelernter. ACM Transactions on Programming Languages and Systems, Vol. 7 No. 1, January 1985, PAges 80-112. The Linda solution presents the application with a kind of `global associative memory' in which data resides. The guys at Yale have a number of `off the peg' implementation solutions (see Nich Cariero [not sure spelling is correct] thesis on Linda Implementation submitted this year [I'd love a copy of this please!]). From a systems point of view Linda provides powerful solutions to many of the other problems confronted in a distributed environment. But you didn't hear this from me (!)! The concept of generative communication in particular simplifies the problem of resource allocation and recovery (won't someone please do some formalisation of this work, then I might be able to justify using it!). If you haven't yet had a brief summary overview of Linda posted I'd be happy to oblige. * Steven Ericsson Zenith Snail mail: INMOS Limited, 1000 Aztec West, | zenith@inmos.uucp Almondsbury, Bristol BS12 4SQ. UK. zenith@inmos.co.uk Voice: (UK)0454 616616x513 ** All can know beauty as beauty only because there is ugliness **
zenith@uunet.UU.NET (Steven Zenith) (11/21/88)
Re posting from from: dls@mace.cc.purdue.edu (David L Stevens) Date: 19 Nov 88 19:15:59 GMT > Suppose you have a network of workstations and you map the virtual >address space of each by page faults that result in network requests to one >or more servers. I'm looking for algorithms to manage the "global" virtual >address space of the network, and pieces in use by the individual machines. > > Applications: shared libraries, shared "in core" binaries, shared >"in core" filesystem data, inter-machine process migration, etc etc. This appraoch to global addressing seems like incredibly hard work and fraught with difficulties (maintaining integrity etc..). If you really need this kind of solution in the mentioned application areas, I would strongly suggest you look at the `Linda' work done at Yale. The first paper to read is: Generative Communication in Linda by David Gelernter. ACM Transactions on Programming Languages and Systems, Vol. 7 No. 1, January 1985, PAges 80-112. The Linda solution presents the application with a kind of `global associative memory' in which data resides. The guys at Yale have a number of `off the peg' implementation solutions (see Nich Cariero [not sure spelling is correct] thesis on Linda Implementation submitted this year [I'd love a copy of this please!]). >From a systems point of view Linda provides powerful solutions to many of the other problems confronted in a distributed environment. But you didn't hear this from me (!)! The concept of generative communication in particular simplifies the problem of resource allocation and recovery (won't someone please do some formalisation of this work, then I might be able to justify using it!). If you haven't yet had a brief summary overview of Linda posted I'd be happy to oblige. * Steven Ericsson Zenith Snail mail: INMOS Limited, 1000 Aztec West, | zenith@inmos.uucp Almondsbury, Bristol BS12 4SQ. UK. zenith@inmos.co.uk Voice: (UK)0454 616616x513 ** All can know beauty as beauty only because there is ugliness **