memorandum

 

To:             Katalin Voros, Operations Manager

From:        Todd Merport, Computer Systems Manager

Subject:     2007 Year-End Report

Date:         17 January 2008

 

I.              Introduction

Microlab operations have utilized relational databases and applications almost since the technology was introduced at Berkeley. The systems and databases are highly tailored to give management information on lab activities and use the information in the most efficient way. The Microlab computer group is responsible for maintaining, upgrading, and replacing these systems and insuring systems are running with excellent reliability.

II.       Infrastructure  SUMMARY

These are the main systems used for the computing infrastructure:

The five main systems that make up the infrastructure are a diverse set of technologies. System diversification has its good and bad points. The downside is a patchwork system with some duplication. Also computer staff are spread thin as far as technological skills. The upside is not being roped in to one vendor and the ability to incorporate new technology in the mix for a specific need. The lengthy experience the Microlab has with computerized operations ensures infrastructure decision making is not hasty -- not jumping headlong onto the latest bandwagon helps productivity and reliability: a well-maintained, documented, tested open-source software foundation is far advantageous then newly released, buggy, and poorly documented technology.

Figures 1 and 2,  and Table 1 summarize the current infrastructure. Specifics are summarized in the Tasks section.

III.      Organization

I supervise two Programmer/Analysts. We strive to be diverse in our skills and available to help Microlab members. The responsibilities of computer staff are listed below:

Changrui Yin

Olek Proskurowski

Todd Merport

·         Supervisor, UNIX Administration, PC Server Administration, etc.

Eniko Seen (left at the end of June)

With special thanks to Madeleine Leullier who assists members with computer problems and helps manage computer resources. Madeleine also holds the computer group orientation. Her presentations are lively and detailed.  Eniko finished the Java based client for the Mercury system and has left the group.

IV.      TASKS

Unix SYSTEM Administration

Migrations and upgrading continue. All tasks done by the server silicon (circa 1999) has been moved to newer servers with the latest OS releases. Migration started with mercury for the home directories, then silicon2 for the Wand and the Microlab database (off of argon), and finally for mail and the user database, silicon3 has been deployed. This last bit of migration to the new mail and password server entailed creating a shadow system and various migration scripts. A phased approach was necessary to prevent significant disruptions to Microlab operations. 

The artwork conversion server that generates TAP and TIX files for the mask making equipment has been upgraded from a SUN Ultra 5 to a 2-CPU SUN Sun-Blade 1000 with 2GB memory. Significant improvements in conversion speed have resulted.  Members are able to use the conversion software much more efficiently because they no longer have to wait for licenses. Also Solaris 10 is much more robust when dealing with layouts that cause the conversion program to go into unwanted states. The Artwork Conversion Inc. engineer worked with Olek to re-compile (and debug) their software to better utilize the Ultra Sparc III architecture.

Through an act of Congress daylight savings time beginning and end dates were changed this year. As usual, computer vendors did not release patches until the last minute. Olek upgraded the Ingres database with new time zone definitions. He recompiled timezone rules on silicon. Changrui made sure MS based servers had the necessary patches for the new time zone rules.

WINdows Administration

Lab terminals utilize a thin client and a Windows Terminal server “CAPE”. This original system server has been upgraded this year from a 2CPU (1GHZ), 2GB system to a 2CPU (2GHZ) Xeon Processor with 8GB of memory. Performance has improved 10 fold. Changrui took on specifying, ordering, and deploying this system. He also incorporated RAID drives and a hot backup system.

Mercury Software Project

Eniko has completed the Mercury client. This is the program that members will use in the new lab to enable/disable equipment, file problem reports, and view lab activities.  An important aspect of this program is the ability to manage sessions so a member’s location and time can be accounted. This is the real-time aspect of Mercury. Many other forms will be available from the Web.  She also worked on web based tools and the Purchase application.

I revisited the migration and database creation script for Mercury. The procedures were broken up into modules that permitted easier modification and additions. I also developed a script to map Microlab activities to the Mercury database.  Some modifications to the Mercury client have been required since the database schema has changed since Eniko left.

Olek found new development tools for Mercury including Netbeans and BIRT. He implemented a session management, authentication system, and developed several forms and reports. He will elaborate in the Comments section.

Marvell Nanolab planning

Planning is underway for the telcom needs of the new Marvell Nanolab.  The new lab will run the “Mercury” system of application software and database system.  A couple of other areas where technology upgrades will be needed are most notably the interlock system and sensor acquisition system. The current interlock system that was introduced in 1999 is manufactured by Walker systems. It has performed very reliability and integrated well into the Wand system.  The system consists of 25 channel plug-in boards for an ISA bus and individual relay boxes.  The software drivers were initially developed at MIT and maintained by us as Linux kernels changed versions. Now the ISA bus is well obsolete and Walker Systems is no longer is business.  It makes sense to use a switching system from a major vendor such as Agilent or Keithley for the new lab.  These systems use magnetic latching relays, have built in network interfaces, and substantial front panel controls. A mock-up of this system is shown on Figure 3.  A similar upgrade is needed for the sensor acquisition system RUMS. Built with Labview and using a plug-in card, hardware and software upgrades will be needed for the new lab. But it makes sense to use the same hardware types for the interlock system and sensor system. Real-time widgets are seldom needed for lab sensors and a back-end database can handle all processing. For both these systems, using the Mercury server framework for client-server-database communication will be ideal (see Figure 4). Computer staff are quite fluent in Java and SQL query languages. This will give the Nanolab a great deal of flexibility and rapid development capability. Facility staff will have significantly more flexibility with these systems by using the extensive front panel measurement and control capabilities that come with a Keithley or Agilent system. Moreover, managing data with a database will allow other equipment to be integrated into the system in a seamless manner.  Sampling at fixed but settable intervals will simplify processing and analysis. The cost of the proposed equipment interlock system is about 25% of the cost of the Walker system. For the sensor acquisition system, the cost will be about the same as the National Instruments card and Labview.

Voice communications in the Nanolab is under discussion. Some proposals are on the table to utilize VoIP (voice over IP).  I’ve been participating with Bill Flounders and Phillip Guillory to determine standards for the Nanolab. One item that was quickly determined was that VoIP was not nearly reliable enough. A network outage would kill all communications (short of shouting).  It is pretty clear that the Microlab’s telephone system has been super reliable. It will be used as a design basis for the Nanolab. 

the wand and microlab database

Although “Mercury” has not been released, significant technology and knowledge has been transferred to improve the current “Wand” system. The latest transfer is web based reports extracted from the Microlab database. The reports give similar information to labhist but with many advantages. The selection panel allows easier parameter entry using standard widgets. Queries can be run based on equipment or user name, and optionally fixed date intervals. The report output is exportable to Excel and PDF. A report for RUMS has also been developed that extracts data from the RUMS database for a selected sensor.

Miscellany

I designed a plain old UNIX mail spam filter that plugs into a user’s forward file. It determines spam by a set of rules that are based on words and senders. Entries for words and senders are based on regular expressions (regex). This allows a great deal of power when developing rules.  About 80% of spam is filtered out of a user’s mail with the current set of rules. It is a simple PERL program using only 200 lines of code, one plain text file for the rules, and a log file.  It can forward mail to another server or append accepted mail to the mail spool file.

V.       COMMENTS

Olek Proskurowski

This past year I mainly concentrated working on Mercury system. For the first several months I worked on improving of an accounting/billing part of the database. For the rest of the year I worked on the web based interface of the Mercury system. I also fixed several bugs in Wand system and assisted Todd in system administration.

Mercury Database

Mercury Web  (implemented the following modules)

System Administration

Wand

CHANGRUI YIN

In 2007 I have taken care of system administration, computer based equipment support, staff support, web application development, and office moving. I worked 50% for the Microlab, 50% for the BCAM/FLCC research groups. Sometimes I had to take care some of the business on the evenings and weekends (Even though Changrui is pulled from all directions, he maintains his cool and is very responsive to Microlab needs – Todd).

Windows Systems Administration

Cape Deployment

Microlab4 (Staff Terminal Server)   Deployment       

Workstation and Lab Equipment 

Specified and deployed the new plotter

UNIX System Administration

Machine-shop File Management System Project (Startup)

Moving

·         Coordinated office moves for the computer group.

VI.      Future  Plans

Mercury

Rapid development is underway for various sub-systems of Mercury. They should be completed and tested this year. The client will only require minor bug fixes and modifications.  Significant testing and simulation of the accounting system is required.

MARVELL NANOLAB

Planning will continue this year as the building nears completion. We haven’t received word if the campus or EECS will manage the network environment and what re-charge structure will be implemented.  These issues affect port activation, system monitoring, firewalls, hostnames, email services, and account management – just about everything related to computers.  Voice telcom, equipment control, and sensor acquisition systems will need to be further pinned down and possibly purchased.  

MACHINESHOP JOB/FILE MANAGEMENT SYSTEM

Changrui will redesign the current Job Management system (database and website) to incorporate revision based file management. He will create forms to take care of viewing/uploading and downloading.

INFRASTRUCTURE

A new UNIX server to act as a development platform and backup system should be ordered and deployed. All other UNIX servers are now in production mode.

WORKSTATIONS

Another round of new staff computer purchases will need to start this year possibly deploying one or two per month.

VII.    SUMMARY

Staff and members have seen some visible improvements in 2007. And from my point of view we have finally got things on a pretty solid footing.   All servers are running on supported operating systems. Hardware is up-to-date, too. New accounting software has been released that offers reports in HTML, Excel, and pdf format. The Wand replacement software ‘Mercury’ is far along in development and testing. Lab equipment computers have been cloned, repaired, or replaced in a timely manner.

VIII.  TableS & FIGURES

Function

Platform

Operating System

Count

Desktop

X86 Dell (Mostly)

Windows XP

42

Web, FileSystem, Database, Mail, Applications

UltraSparc

Solaris 10

3

Web, LabView

X86 Dell

Windows 2000 Server

1

Terminal Services

Xeon Dell

Windows 2003 Server

1

FileSystem, Web, Database, Anti-Virus

X86 Dell

Windows 2003 Server

5

Cad Workstation, EGLAS

UltraSparc

Solaris 9,10

3

Cad Workstation

X86 Dell

Windows XP

1

Thin Clients

Neoware

Windows CE

16

Interlock Controllers

X86

LINUX

3

TOTAL

 

 

75

Table 1 - Microlab Systems Summary (not including equipment)

I.                 

Figure 1 - Microlab Computer Infrastructure

 

 

Figure 2 - Web Server Infrastructure

 

Figure 3 - Nanofab Equipment Control Proposal

 

Figure 4 - Sensor Acquisition Concepts