Orcmid's Lair

<$BlogItemTitle$> Welcome to Orcmid's Lair, the playground for family connections, pastimes, and scholarly vocation -- the collected professional and recreational work of Dennis E. Hamilton

This page is powered by Blogger. Isn't yours?

2002-12-28

 

Groupware and Collaboration

Collaborative Workspaces

A wiki in your future?

Shared Workspaces.  A commentary by Nancy Cohen, on Open Magazine.  Discusses Wikis and their benefit in creating shared workspaces.  The workspace aspect is considered the critical factor.  Michale Schrage, MIT Media Lab co-director, weighs in for Wikis.

 

KnownSpace knowledge-oriented document management

Helium User Documentation: Features,  A bit more on how Helium operates.  It is coupled to IMAP, so if you don't have mail access via an IMAP server, this won't work for you.  Being a devoted POP3 and Outlook guy, it won't work for me.  I am also trying to figure out what the "LSA clustering" is that Chris Dent mentions being supported here.  That's the price I pay for not having dug deeper into this all along.

 
KnownSpace Hydrogen API.  The basic API model for the entities, simpletons, events and so on that the KnownSpace Hydrogen model requires.  I dunno how this scale and in particular how it distributes.  It would be interesting to abstract some kind of model of the relationships involved here.  Entities are basically the means of expressing information models, the simpletons are generic processes that roam over entities.

 
KnownSpace.  The KnownSpace site itself.  KnownSpace is billed as an open-source document management system.

 
Welcome to KnownSpace Helium.  The Helium e-mail navigation application.  Part of the KnownSpace development of knowledge-oriented desktops.  A Java application.

 
The Helium Manifesto.  "We are working toward a future where anyone with a computer can explore the space of human-computer interactions, all within a universal, open, and free environment." The KnownSpace approach is to come up with a desktop that is readily organized and personalized and accessible and interchangeable and shared.  Java is seen as the universal platform that KnownSpace requires.

And don't forget messaging

IM compatibility closer to reality - Tech News - CNET.com.  Back to the possibilities of simple groupware, including instant messaging between people and/or reactive processes.

There are two IETF efforts. XMPP, based on Jabber, is starting a Working Group. There is already SIMPLE which became an IETF proposed standard in September.

2002-12-27

 

Life in Cyberspace

Spam and Virus Detection

MessageLabs e-mail scanning and protection

MessageLabs is a Managed Service Provider.  MessageLabs home page. The labs provide e-mail scanning and protection as a service.  Viruses and unsolicited mail, including pornographic material, are all viewed as "threats."

 
MessageLabs - Press Release.  Virus scanners have some of the same problems as Spam filters.  We have seen inappropriate responses from server-based filters because of spoofing by virus propogation techniques.  Is this also something for which the solution is doomed to non-algorithmic solution.  I would think so.  And here is something about the valiant deployment of heuristic approaches.

 

Spam conference and resources

UseTheSource | POPFile Beta 1.  This is about a working e-mail classifier now in Beta 1.  The kinds of things to be dealt with just to install a tool like this is particularly interesting.

 
Arc.  While looking over Paul Graham's treatise on Spam, I thought I should blog the Arc page too, as a reminder to keep track of it while I work on Miser.

 
A Plan for Spam.  A manifesto of sorts on Spam and why Paul Graham thinks it should be easy to filter.

 
2003 Spam Conference Speaker Abstracts.  A source of goodies and links to available tools and experiences.

 
Spam Conference.  January 17, 2003 in Cambridge, MA (at MIT's Media Lab), this conference is an informal get-together of people wanting to filter spam.  This will be very interesting, since I have proposed this as an inherently non-algorithmic situation.  We will see what the experts say.

 

Computer Technology

Blade Server Appeal

IBM announces new blade server line - Computerworld.  2002-09-24 article that takes us back to the kind of high-end product IBM anticipates.  These are part of the eServer family.  At some point, we get back to mainframe capacities.  These blades can hold 8GB of memory, two internal hard drives, and operate optionally with fibre switches and Fibre Channel architecture.  Life does get interesting in the data-center broom closet.  Here is where the tradeoff on power consumption and environmental requirements will become interesting and, perhaps, telling.

 
Blade wars: Dell server undercuts IBM, HP on price - Computerworld.&bsp; A little more on what blade servers are all about.  Here we see competitive positioning among the big three and also a hint that management tools for configuration and support become a key differentiator.

 
RLX rolls out new Transmeta-based blade - Computerworld.  An earlier (2002-12-16) article about the appeal of the Crusoe processor in blade-server applications, because of the low power requirements.  The chip runs Windows Server and Linux at the user's option.  We have already seen an article where the Crusoes are being built into supercomputers for the same reason.

 
BP Energy rolls out blade servers for trading floor - Computerworld.  Here is more information on blade servers as a way to consolidate desktop support and keep towers off of people's desks.  This is interesting technology to watch, especially in conjunction with Web Services and ultimately, the 10GB and larger non-volatile tablet PC.  Interesting for technology watching.

 

Groupware and Collaboration

Wiki Wacky

Hey, I should use that, or wacky wiki or ...

Archipelago implementation

Archipelago FAQ.  Well, OK, it is toolcraft-heavy.  Hmm, are my blogs this disconnected and context-dependent?  Probably.  Ponder later.

 
Creative Commons: 2002-12 Archive.  Interesting. Creative commons has a pretty format, but their weblog does not provide the equivalent of "purple numbers" for linking to the archive.  Fortunately, you can find the current month's archive from a link at the bottom of the current weblog page.  This particular current month is where some announcements about machine-readable material appear.

 
Creative Commons Unveils Machine-Readable Copyright Licenses.  Related to discovery, and reached from Archepelago.  This fruit bears all the symptoms of a depth-first traversal, so bear with me.

 

Really simple discovery

RFC: Really Simple Discoverability 1.0.  This seems like a way to make discovery of the requirements for editing a site discoverable so that people can use different editors and other mechanisms.  I find this page rather opaque, probably because it carries a heavy toolcraft assumption that I am clueless about.  Suggested by Gary Johnson on BA-UnRev-Talk, I will blog enough to figure out how to come back to this.

 

Wikis for content management and logging too

Tiki the CMS.  This is Tiki Wiki, a Wiki oriented to Content Management (CMS).  It has templates, themes, and other goodies, including attachments and more.&nsp; The implementation, on SourceForge, is in PHP.

 
SnipSnap :: start.  SnipSnap is a Web Log and Wiki combination built in Java.  The package is licensed under the GNU GPL.

2002-12-26

 

Computer Science

Compression Algorithms

Here's an example of algorithms that aren't algorithms in the sense that there are two functions, and one provides the inverse of the other.  That is, the compression algorithms isn't for anything in the absence of the decompression algorrithm.  This is akin to the protocol situation, except protocols operate interactively and the compression-decompression game is more for a function and its inverse.

Lempel-Ziv coding algorithms.  Another discussion of the LZ and LZQ coding algorithms.

 
Lempel-Ziv Compression.  A tutorial on the L-Z compression variants.  Useful for seeing how no dictionary is transmitted in advance, it is constructed by the receiver in exactly the same way it was constructed by the sender, using the recently-decoded part of the message.

2002-12-24

 

Computer Technology

Computer History

Unisys and B5000 architecture

Unisys eCommunity Login Page.  A place I need to return to in gathering more information about current UniSys products and whatever Burroughs 5000 legacy remains.

 
Unisys History Newsletter.  A newsletter by George Gray, I should get in touch with him and let him know about my experiences as the technical liaison on the Univac ALGOL projects.

 
Historical Computer Literature.  A cool source of manuals from early computers.

 
Burroughs B5000 News Clippings.  We are discussing polymorphic instruction processing in class, so tagged architectures came up. This leads me to wonder if there are any surviving Unisys products that use the B5000 architecture. Here are some of the original 1961 press releases and news articles on the machine.

2002-12-23

 

Memory Systems

More trends

Datasheets of StrongArm MCM.  A collection of datasheets on a project to apply the Intel StrongArm memory.

 
Semiconductor Memory Technology and Application Trends.  A white paper from ETH Zurich, located by classmate Yiannos Kourmouzi.

 
ZDNet UK News - Wanted: Successor to Flash Memory.  A ZDNet think piece on varieties of non-volatile memories.  Cited by classmate John Cotsikoros.

 

Storage Systems

IBM Redbooks | Introduction to Storage Area Network, SAN.  Here's the Redbook abstract on the Introduction to Storage Area Network (SAN) that Jeffrey cited.

 
IBM Redbooks | Just Published.  IBM Redbooks are part of the amazing documentation amassed by what at one time was the world's largest non-governmental publisher.  The difference is that Redbooks are now available on-line and on CD-ROM. Also, there are other manuals that provide deeper materials, functional, interface, and architectural specifications, and training materials.  It is a mammoth operation.  This searchable index was pointed out by classmate Jeffrey Caldwell, who cited the Redbook on "Introduction to Storage Area Network," available as a PDF.

 
SNIA - Storage Network Industry Association: Presentations.  Presentations on storage systems at a 1998 meeting of the Storage Networking Industry Association.  This examination of storage-system approaches, as predicted then, was located by classmate Jeffrey Caldwell.

 

The promise of non-volatile memories

NewsFactor Sci::Tech -- The End of Computer Rebooting.  Another angle on non-volatile memory, with some leads to track down.  It is suggested that these memories will be competitive in 3 years.  Another Martyn Atkins find.

 
Future Trends in Non-Volatile Memory Technology.  Another find from Martyn, this time an Intel Research PDF file.

 
[Cover Story] Flash Memory Threatens to Replace HDDs.  Ah, finally some analysis of where these predictions for multi-GB non-volatile storage can be substantiated. Very useful analysis, turned up by classmate Martyn Atkins.

 

Good Reference

Webopedia: Online Dictionary for Computer and Internet Terms  Another nice resource for information on computer technology.  This one was cited by classmate Yiannos Kourmouzi.

 

Storage Systems

Hard disk evolution

Hard disk drive design and technology.  A nice description of the RAMAC, although the elevator positioning mechanism is missed.  A photograph of a RAMAC would be very nice.  This USByte site provides nice collections of information on computer technologies. (Reference provided by classmate John Cotsikoros.

 

RAID technology

JAD Logic Ltd [What is RAID?] - RAID  A nice capsule description and appraisal of the different RAID levels, with a favorable impression of RAID 5.  Another reference from Ingo.

 

Database optimization and performance

DBAzine.com: ORACLE DBA Resources.  Another citation from Ingo. This looks like a good source of performance information and other DBA-oriented materials on Oracle, DB2, SQL Server and (gosh!) IMS.

 
Oracle Technology: Performance.  Cited by classmate Ingo Wevers, as an example of disk performance techniques applied at the database system level.

 

Hard disk technology

PCGuide - Reference Guide - Hard Disk Drives.  The PC Guide was also recommended in our lecture notes.  This is delightful material.

 

Making disks look like memories

Solid State Disk White Paper.  This paper, on solid-state disks provides a novel system used for Stratus and analyzed by DRA International.  The solid-state disk uses volatile RAM, so there is a hard-disk drive integral to the unit which provides backup.  There is also a built-in battery backup.  An interesting effort, pointed out by Mike.

 

Computer System Architecture

System Architecture.  The PowerPoint slide set provides a nice overview of basic hardware system architecture concepts and trade-offs, including a view of multiprocessor architectures. Recommended by classmate Michael Franklin.

 

Storage Systems

Hard disk technology

PCGuide - Ref - Redundant Arrays of Inexpensive Disks (RAID).  The PC Guide discussion of RAID technology. Also recommended by Mike.

 
PCGuide - Reference Guide - Hard Disk Drives.  A very nice guide to the workings and structure of HDDs.  There are great illustrations and a delightful commentary. Recommended by classmate Michael Franklin and others.

2002-12-22

 

Computer Technology

More Performance Characteristics

Topics in Computer Science.  Some general summaries that are useful to review from time to time.

 
Disk scheduling.  A look at the different ways that disk usage can be optimized.

 
Freecom Technologies is the specialist in multimedia connections and storage solutions for PC, Apple Macintosh and Home entertainment..  A source of flash memories and glass storage solutions.  I was referred to the specifications for the FM-1 USB Memory Stick by a classmate.

 

Man vs. Machine

AI and The Limits of Computation

All that can be computed and more?

I notice that there is some confusion around the Church-Turing thesis and exactly what Turing demonstrated was universal about UTMs.  A marker for more, later.Intelligent Systems: AI's Greatest Controversies.  A little off track for now, but something else to look at later.

 

World Chess Champions vs. Computers

Recent events

ChessBase.com - Chess News.  The Kramnik vs. Deep Fritz drawn match, 4-4, in Bahrain.  Kasparov is scheduled to play Deep Junior in a match beginning in New York City on January 23, 2003.

 

The historical Kasparov vs. Deep Blue matches

IBM Research | Deep Blue | Overview.  Information on the 1997 rematch won by Deep Blue, 3-1/2 to 2-1/2 over Gary Kasparov. There are also links to the previous match, won by Kasparov 4-2 and in which Deep Blue recorded the first won game against a World Champion in match play.

 

Computer Technology

Mainframes and Super Computers

Memory capacities

S/390 Parallel Enterprise Servers - G5 and G6.  Using the S/390 architecture, these configurable multiprocessor systems have processor storage running from a minimum 1GB to a maximum of 32GB.  In some cases of IBM multiprocessor technology, processors can share main memory via coordination of their L3 caches.  I don't know if that works for the S390 architecture.

 
IBM eServer zSeries 900, the platform for enterprise e-business.  I am interested in the memory sizes of these systems.  They start at 5GB and grow to 64GB, depending on model.

 

Microprocessors for mainframes

Mainframe technology uses semiconductor microprocessor chips, just like your cellular phone or desktop PC.  These are very hot chips and they use dramatic architecture improvements to deliver GHz performance.

POWER4 System Microarchitecture.  About the CPU.  These chips are combined into multi-processor configurations.  There are provisions in these designs that are unheard of at the desktop PC level.  The chips use 64bit data paths and processing architectures, but that is just the beginning.  What is startling is all of the provisions for operation in symmetric multiprocessor configurations of 32 or more processors.  There is support for availability and configuration troubleshooting that we don't even think about for PCs.  The ability to partition a high-performance multiprocessor into seemingly independent and isolated sub-computers is just one example.  This is done for load management and for operation of tests and upgrades while main production systems continue.

This processor chip is at the heart of the p690 series of eServers.  Multiple configurations of p690s are used to build supercomputer configurations such as a 7.3TFLOP, 42TB storage system 44-p690 cluster that is the starting point for growing to a 100 TFLP system in stages (article below).

 

Making memory work fast

IBM eServer pSeries 690 with the HPC Feature.  Describes the upgrade that is done to support high-performance computing by improving memory bandwidth. There are 3 levels of cache, L1 is 32kB data, 64kB instructions. L2 is 1.5MB dedicated to a single processor (rather than shared in the dual-processor form of the POWER4 chip) and the L3 is 32MB/chip and these can be coordinated across multiple cpu-chip modules (e.g., 128MB in a 4-way module).

It appears that each L3 cache can "front" a 4-8GB main memory, and that coordinated use of 64-128GB aggregate main memory is by coordination among the L3 caches of the machines.

 
eServer from IBM - pSeries 690.  Now we're getting somewhere.  The @server product family has an ISO 9001 certificate.  They are multiprocessors (up to 32-way), can be partitioned dynamically, are air cooled, and run Linux.  They have 128MB cache memories and grow system memory to 128GB in the 32-way configuration (at a nice $2M).

 
Welcome To SPEC.  The Standard Performance Evaluation Corporation.  Oh, so this is where Spec-marks come from.

 

Super-server and Super computer characteristics

IBM eServer p690 server establishes new Oracle transaction processing record.  The p690 is a multi-processor configuration, with 32-way multiprocessing used in this benchmark configuration.  The system is based on the POWER4 microprocessor, a 64-bit processor that basically provides a server on a chip. It has two 1GHz+ processors, a high-bandwidth system switch, and a large memory cache as well as I/O interface.

 
Bell, Torvalds usher in next wave of supercomputing - Computerworld.  This Los Alamos installation configures hundreds of blade servers based on the Transmeta Crusoe chip. The power advantages allow the computer to be tucked in a closet without aggressive cooling requirements. Other configuration details are not provided.  The machine is called Green Destiny.  An earlier version, MetaBlade, has been running 9 months without a failure, and credit is given to the low transistor count and use of software over software in the Crusoe approach.

 
IBM nets its biggest supercomputer deal yet - Computerworld.  The IBM order for the National Centers for Environmental Protection.  The system will be built up in stages, achieving 100 TFLOPS by 2009. IT starts with 42TB of storage in a cluster of 44 IBM eServer p690 servers. That gets to 7.3 TFLOPS. The system will occupy a room roughly the size of a basketball court and will be used remotely by the customer.

 

Storage Systems

General resources

Computerworld Hardware Knowledge Center.  I am looking for information on main memory trends, and the manufacturers are not much help so far.  Computerworld watches the mainframe industry, so I am looking at their predictions now.

 

Available storage servers

Dell, IBM and Sun to deliver InfiniBand servers - Computerworld.  InfiniBand technology, promises the ability to connect servers and storage systems in a data center at speeds from 2.5GB/sec. to 30GB/sec. vs. the roughly 1GB/sec. for existing Peripheral Component Interconnect. technology.

It appears that the actual transfer rates are lower than that, but using 2GBs and faster connections will support bigger, faster bursts from storage systems.

Hard Hat Area

an nfoCentrale.net site

created 2002-10-28-07:25 -0800 (pst) by orcmid
$$Author: Orcmid $
$$Date: 03-09-06 20:49 $
$$Revision: 4 $

Home