
Issue #5 1995
Go Back to DBAdvisor Main Page
In
This Issue:
- Yes, DEC DBMS was purchased by Oracle
as part of the sale of the Rdb family of products. In
this issue, we will review what this will mean to you,
the DBMS installed base.
Acquisition Finalized
Redwood Shores, Calif. Dec. 1
1994 - Oracle Corp. announced the completion of the
acquisition of Digital Equipment Corporation's Rdb
[and DBMS] database, repository and database
administration tools and the related support business
for $108 million (U.S.).
- Oracle has renamed DEC DBMS to "Oracle
CODASYL DBMS".
- "We are very pleased to announce
the completion of this acquisition and to officially
bring Rdb [and DEC DBMS] and related businesses into
the Oracle product family," said Jerry Baker,
senior vice president, Oracle Corp. "We feel Rdb
[and DBMS] customers will find the transition from
Digital to Oracle as their product and service
provider to be a smooth and easy one."
Enhancements to Products Now
Underway
Since Oracle's announcement of their
intent to acquire Digital's Database Systems Group (Sept. 1,
1994), over 250 letters of employment offers have been sent
and accepted by the database systems employees, a new Oracle
development center has been established in New England for
database systems employees, as well as a support center in
Colorado. Additionally, work on enhancing the DBMS product is
ongoing, and efforts are well underway to provide greater
interoperability between the DBMS and Oracle family of
products.
Development Efforts Underway
Oracle recently completed and released
DBMS v6.1 that offers significant enhancements over previous
releases, including the ability to run multiple versions of
DBMS concurrently on one system.
In development at Oracle, is the next
version of DBMS, version 7.0. The DBMS engineering group is
working closely with customers to develop a comprehensive
list of features and functionality that will go into this
version.
Customer Driven
A DBMS working group within Oracle's
Customer Advisory Council has been established to provide
input on future product directions. The WG conducted its
first meeting on February 27, 1995.
"We are committed to making
significant investments in the Rdb [and DBMS] technology set,
and we're happy to report the progress we've made thus
far," said Beatriz Infante, vice president, Oracle Corp.
"I'm confident the Rdb [and DBMS] customer will be
pleased and encouraged with the work completed to date and
see this as evidence of our commitment to the maintenance and
growth of Rdb [and DBMS] as a product line within
Oracle."
Product Support
DBMS customers will have the immediate benefits of support
and consulting services designed specifically around their
requirements. The same people who supported DBMS in the past
have now become Oracle employees and will continue to provide
world class support services. Oracle will honor existing DBMS
support contracts. In the future, customers can renew on an
annual basis at one of the Oracle medal levels of service
(bronze, silver, or gold)-(see related Oracle Services
article.)
Oracle Consulting has established a new
group for Rdb [and DBMS]. The group includes individuals who
were previously members of Digital's database systems sales
consulting organization and who have now joined Oracle. They
bring a level of expertise for the entire range of database
products.
Oracle is a registered trademark of Oracle
Corporation. Oracle7, Oracle Media Server, Oracle Cooperative
Applications, Cooperative Development Environment and Oracle
Transparent Gateway are trademarks of Oracle Corporation. All
other product or company names mentioned are used for
identification purposes only, and may be trademarks of their
respective owners.
Contents
DBMS Hints, Kinks & Work-arounds
Q. What safeguards should be made when upgrading DBMS?
A. Software Concepts International strongly recommends
performing a full DBO/BACKUP both before and after the
upgrade. It is also good idea save a copy of the
pre-conversion root file. This will allow you to recover the
old root file if something goes wrong during the DBO/CONVERT
command. Note that once a database has been updated after the
conversion, this old root file should never be used to return
to the prior version.
DBMS 6.1 supports multiversioning. This
should help to resolve some of the upgrade problems
experienced in the past. With multiversioning, it is possible
to run v6.1 on one database while running a previous version
of DBMS on a different database on the same system. Note that
you cannot run different versions of DBMS against one
database-a database is converted to a specific version level.
Q.
We recently upgraded from version 4.3 to 5.1. Many of our
programs would not run under the new version of DBMS. What
happened?
A. In version 5.0 of DBMS, the metadata definitions
contained in the root file were changed. As a result, any
images that contain callable DBQ will need to be relinked.
Q. I am trying to create a new database, and am using
an options file to specify area qualifiers. It appears that
not all the area qualifiers are being correctly applied. Is
there something wrong with my options file?
A.
Your options file is just fine. There is problem in versions
v4.3c, v5.0b, v5.1b and v6.0 that causes the options file to
be processed incorrectly. Oracle has a patch to resolve this
problem.
Q. We
are experiencing periodic image failures when trying to bind
to our DBMS database. The error message indicates the RUJ
file is locked by another user. Since RUJ file names are
unique as of V4.3, how can this be?
A. The RUJ filenames were made "more" unique
in v4.3. Actually, DBMS uses a time-stamp for creating RUJ
filenames. On some of the faster processors, it is possible
for two processes to get the same time-stamp-thus the same
(duplicate) RUJ filespec. One of the processes will fail when
it attempts to create the RUJ file. As a work-around, you may
try defining DBM$RUJ_DIRECTORY as a group logical name to
place the RUJs from different groups in different
directories. This will not guarantee that you won't
experience this problem, but it may help minimize it.
Q. As the leader in
DBMS consulting and software development, is SCI concerned
about the sale of DBMS to Oracle?
A. At first, we were guarded about how the acquisition
would affect DBMS. SCI has had several meetings with the new
Oracle product manager and the DBMS development team since
the acquisition. We have been very pleased with the level of
support and commitment that Oracle has provided since the
sale. While we have considerable regard for Digital, prior to
the sale, they appeared to be preoccupied with "trying
to figure out what their business is" rather than
"doing business". The sale actually seems to have
cleared a "log jam", allowing many positive things
to happen with DBMS in its new home at Oracle.
Contents
Minimizing Down-Time with DBInitialize
Minimize down time! That is the loud and
clear message of DBMS customers. Each hour of down time can
cost hundreds of thousands of dollars in lost revenue. Yet,
databases still require periodic maintenance to achieve and
maintain optimal performance. DBInitialize is a high
performance software tool that helps minimize down time.
DBInitialize is independent from the underlying DBMS software
version. Thus, you do not need to upgrade to the latest DBMS
version to benefit from DBInitialize.
What is DBInitialize?
DBInitialize is a software tool
that allows you to partition your DBMS database into more
manageable pieces. If only a few storage areas of your
database are in need of reorganization, you can quickly
initialize these areas. In essence, DBInitialize is a
"bulk eraser" that automatically
"disconnects" the cross-area pointers of sets with
owners or members in the area being initialized. Besides
storage area initializations, you can also initialize
(delete) all instances of a specified record-type or
set-type. This makes DBInitialize useful for moving record
types from one storage area to another, or for correcting
record placement problems caused by misguided
"RECALC" operations.
Can DBInitialize reorganize any DBMS
storage area?
DBInitialize can initialize
and cleanup cross-area pointers for any DBMS storage
area. However, DBInitialize does not unload or load
data. DBInitialize quickly removes data (and the appropriate
set relationships). DBInitialize would be used in conjunction
with site-supplied utilities that actually unload and load
the data (and set relationships). Customers may write their
own unload/load routines, or contract with SCI for this work.
How much time can it save?
A DBInitialize site recently
reported that the MANMAN procedure to erase all records in
the WOHAREA, RTGAREA, EMPAREA ran for over 27 hours (this
routine also performs updates in the PRTAREA and ACTAREA to
remove cross-area pointers). After the MANMAN procedure
completed, an additional 20 minutes were required for
DBO/INITIALIZE to initialize the storage areas. Using
DBInitialize, the entire process (deleting all records,
disconnecting the sets and the DBO/INITIALIZE) completed in
approximately 20 minutes! Thus, DBInitialize was able to save
over 27 hours of down time!
At another site, it took 8-1/2 minutes to
perform 10000 disconnects using a traditional COBOL/DML
program. DBInitialize completed the same task in only 1/2
minute -- seventeen times faster than the DML approach.
How did DBInitialize do this?
DBInitialize achieves its
outstanding performance characteristics by scanning the
affected storage areas for the appropriate pointer
clusters/index nodes that need to be "cleaned up".
In addition, when performing storage area initializations,
DBInitialize performs the pointer-cluster removal operation
in parallel with the DBO/INITIALIZE operations - In
the above example, DBO/INITIALIZE took longer than the
pointer-cluster removals and thus DBInitialize had to wait
for the DBO/INITIALIZE to complete.
A Consilium WorkStream site reported that
they successfully initialized all LECOPR_DATE_INDEX and
WIPENT_DATE_INDEX sets (requiring a complete sweep of a 2.4
million block WIPLTH area) in only 45 minutes.
Can DBInitialize "disconnect"
FIXED and MANDATORY sets?
DBInitialize disconnects all
sets required to maintain database consistency regardless
of the set retention modes. Thus, even set types with either
FIXED or MANDATORY retention modes are disconnected without
any schema changes!
Database Independent
DBInitialize is absolutely database
independent. It works with WorkStream, MANMAN, MACPAC, LIMS
and all home-grown DBMS databases.
FOR MORE
INFORMATION
Software Concepts International
Nashua, NH 03063
(603) 879-9022
Contents
Oracle's
Award-Winning Support for
DBMS Customers
Large-scale applications, such as
manufacturing and shop floor systems, require a database that
has the characteristics of high availability and throughput,
coupled with system management features that minimize
downtime. Why? Because when these applications are down, the
business is losing money - losses can total millions if
problems are not corrected FAST!
So it is very important that a database
used in these environments, such as Oracle CODASYL DBMS, have
product features that reflect performance and high
availability. And it is just as important that software
services and support be in place for the product when
problems do occur. Does Oracle's Customer Support Program
measure up to the challenge?
The answer is "yes"! Over the past several months,
Oracle has worked hard to revamp the entire program. A
Globalization Task Force of selected Customer Support
managers from around the world was formed to create a new and
improved model. Key goals were put in place to make this
model a reality:
- Time zone coverage for all products
everywhere on a 24x7 basis;
- Common, consistent worldwide systems,
business practices and processes;
- World class product quality with
built-in supportability
- Fulfillment of customers' business
critical needs.
"The Oracle CODASYL DBMS customer can
strongly identify with these objectives," said Tom
Holmes, Vice President of Worldwide Customer Support,
Americas. "For example, time zone coverage on a 24x7
basis is a direct parallel to uptime in their own
operations."
Efforts went well beyond words; core processes were
re-engineered to reduce cycle and hold times and improve the
quality of service delivery. And a number of new services
were introduced including electronic support, a field analyst
program and a SupportNotes CD-ROM. The changes were a direct
result of customer feedback - over 6,000 extensive customer
satisfaction surveys were analyzed to ensure the design of
appropriate solutions.
"Suggestions from our customers were
invaluable," continued Mr. Holmes. "We discovered
what we were doing well - customer education, our phone
system, our handling of 'urgent' situations, access to
problem/fix databases, and the expertise of our support
analysts. And we determined where we need to improve - hold
times, length of time to reach an analyst or deliver an
effective solution, and to offer more proactive assistance.
We certainly knew where customers wanted us to focus our
efforts."
Consequently the Support Organization was
re-engineered. An aggressive hiring program was instituted to
reduce hold times and length of time to get an answer.
Quality was improved by requiring additional training for
analysts and the implementation of improved expert systems.
One of these systems is the new Problem Repository which is
currently being piloted in the Supercenters at Redwood
Shores, CA, and the UK. It will be directly accessible by
customers sometime in the future.
Service offerings also were overhauled to
provide flexibility to meet the needs of individual
organizations. The Bronze level provides telephone and online
support for customers who do not require extensive coverage;
the Silver level provides more extensive around the clock
support (this level is the closest equivalent to Digital's
level of product service); and finally the Gold level
provides a designated Oracle support team dedicated to giving
personalized expert responses whenever needed. All offer
electronic support - a convenient alternative to phone
support that can be designed to be proactive. Examples
include SUPREQ, an automated e-mail server that allows access
to the Real Time Support System (RTSS) over the internet, and
SupportLink which gives complete read/write access to the
support forum on CompuServe.
"Feedback so far has been positive. We are still
gathering input as we need to see how well the new offerings
are working plus looking for additional ways to meet customer
needs," added Mr. Holmes. Methods being used include:
- A worldwide customer satisfaction
survey;
- Monthly relationship surveys;
- Daily transactional TAR (technical
assistance report) surveys;
- Seasonal surveys at major user group
conferences (IOUW, EOUG);
- Bi-annual focus groups at major user
group conferences;
- Face-to-face customer
interviews/visits.
The DBMS customer has an additional avenue
- feedback through the Oracle CODASYL DBMS Working Group, a
subset of the Oracle Rdb Customer Advisory Council. This
group is devoted exclusively to the needs of the DBMS
customer. Service and support input is welcome as are
proposed product enhancements. The working group addresses
the longer term future of DBMS as well as shorter term
concerns.
Overall, DBMS customers win. They retain the support they are
accustomed to as the Rdb Support Center in Colorado and other
locations was part of the acquisition. This means DBMS
questions get answered by the same people; the DBMS customer
has retained the experience and knowledge acquired through
the years by these support centers. In addition, the DBMS
customer can take advantage of Oracle's new, improved
offerings. These services, a direct result of customer input,
have been publicly recognized for the value they provide. In
1994, Oracle won four service and support awards from Sentry
Publishing in their annual survey of IT executives - the
highest number awarded to an independent software vendor!
And what about the future? Oracle is
committed to long term, consistent, superior,
customer-centric support. "Customer satisfaction is
crucial to Oracle's overall success," stated Mr. Holmes.
"The Oracle CODASYL DBMS customer clearly plays a major
role by speaking up about their support requirements. Please
give us an opportunity to keep on improving."
Oracle Worldwide Customer Support offers
U.S.-based customers Oracle Customer Care, a variety of
support programs and options designed with maximum
flexibility to suit the individual needs of each
organization.
| |
Gold
|
Silver
|
Bronze
|
SupportLink
|
| Software
Releases/Updates |
x
|
x
|
x
|
-
|
| Product Updates &
Patches |
|
|
|
|
| Electronic
Services |
|
|
|
|
| Oracle SupportLink
|
Read
Q&A
|
Read
Q&A
|
Read
Q&A
|
Read
Only
|
| O-L Problem-Fix
Database |
x
|
x
|
x
|
-
|
| O-L Real-Time Support
Sys. |
x
|
x
|
x
|
-
|
| Alerts |
x
|
x
|
-
|
-
|
| CD-ROM SupportNotes
|
x
|
x
|
x
|
Option
|
| Add'l Subsc to Support
Notes |
Option
|
Option
|
Option
|
Option
|
| E-Mail Server |
x
|
x
|
x
|
-
|
| Telephone
Assistance |
|
|
|
|
| M-F 5am-6pm PST |
x
|
x
|
x
|
-
|
| 7 x 24 Telephone
Coverage |
x
|
x
|
-
|
-
|
| Toll-Free 800 Number |
x
|
x
|
-
|
-
|
| Proactive
Support |
|
|
|
|
| Proactive Alerts |
x
|
x
|
-
|
-
|
| Technical Newsletter |
x
|
x
|
x
|
Option
|
| Scheduled Onsite
Support |
Option
|
Option
|
-
|
-
|
| Full-Time Onsite
Support |
Option
|
-
|
-
|
-
|
| Personalized
Support |
|
|
|
|
| Priority
Processing/Routing |
x
|
-
|
-
|
-
|
| Designated Support
Team |
x
|
-
|
-
|
-
|
| Account Mgmt. Reports |
Automatic
|
On
Demand
|
-
|
-
|
| Database Center
Expertise |
Option
|
-
|
-
|
-
|
| Add'l
Options/Reqmts. |
|
|
|
|
| Education |
Recommended
|
Recommended
|
Recommended
|
-
|
| Consulting/Installation
|
Option
|
Option
|
Option
|
-
|
| Consulting/Upgrades |
Option
|
Option
|
Option
|
-
|
| Consulting/DB Audit
Service |
Option
|
-
|
-
|
-
|
| Documentation |
Option
|
Option
|
Option
|
Option
|
Contents
Oracle Corp. has formed the Oracle CODASYL
DBMS Customer Advisory Working Group (DBMS WG). This Working
Group is a "subset" of the Oracle Rdb Customer
Advisory Council (CAC) which was formed to:
- establish a forum for user input and
feedback for Rdb [and DBMS] product line directions;
- provide up-to-date information
- understand customer concerns;
- plan for the future.
In essence, the DBMS WG will be the forum
for user input and feedback specific to DBMS. We hope to
obtain input not only for the shorter term next versions, but
most importantly, for the longer term future of the product
and related services and support.
- The criteria for membership for the
DBMS WG are:
- existing DBMS users for a range of
end-user customers, third-party application
providers, and DBMS service providers;
- knowledgeable about DBMS, its
attributes and benefits;
- ability to "forecast" own
database needs into the future, and make
recommendations for DBMS based on that;
- willing to participate in the WG
activities and, consequently, to devote time to
prepare and participate in these activities.
Companies
currently represented in the WG include:
- British Aerospace
- Consilium
- Georgia Pacific
- Intel
- MassMutual Life Insurance Co.
- McDonnell Douglas
- Software Concepts International
- Oracle Corporation
- Short Brothers
- Sikorsky Aircraft
The first WG meeting (teleconference) was
held on February 27, 1995. The objectives of this meeting
were primarily technical in nature-to discuss technical
enhancements to DBMS.
To stimulate discussion, Oracle presented
several product enhancement ideas for V7.0. These ideas [not
commitments] included:
- Automated AIJ Log shipping.
- 64 bit AXP support
- Utilities Improvements
- Combined VERIFY & ANALYZE
- Additional RMU-like features
- DEC C language support.
The customer members of the WG had no
problem in developing a list of product enhancements.
Customer requests included:
- More run-time DEBUG flags, allowing us
to:
- trace DML verbs on a per
stream basis
- trace DBKEYs accessed.
- DBO/SHOW STATS enhancements,
including:
- Monitor a transaction's bind
status (ready mode).
- Summary report on area
fragmentation.
- Rate of SPAM activity per
area.
- Ability to create customized
screens with calculations.
- Ability to capture and dump
I/O stats by area in batch mode (capture
binary area stats)
- Ability to dump CDD path information
only
- Dynamic (run-time) access to DBMS
metadata.
- Official documentation and support of
DBO/ANALYZE/BINARY feature.
- Enhanced documentation on dates when
metadata was changed.
- Ability to know when a quietpoint is
achieved from within an application.
- Modular DBO utilities (allow new
routines to access old databases)
- Fix for DBKscope locking problem
- Enhanced support for 7 x 24 x 365
sites.
- Ability to limit number of concurrent
recovery processes (avoid swamping the system).
- Ability to examine contents of
keeplists (from COBOL).
- No-CDD support from COBOL.
- Enhance DBQ to allows:
- Host variables
- Arithmetic computations
- Move data between fields
- Windows interface
- Improve performance of duplicates in
indexed sorted sets.
- Ability to write AIJ backups in either
RMS or new tape format (like Rdb).
- Maintain valid buffers during changes
in ready modes.
- Address performance issues with
"main memory" databases.
- Ability to set buffer-lengths by area
and process. Alternatively, we would like a way to
decouple the size of an I/O from the buffer size.
Members of the WG were requested to
prioritize their technical "wish list" and return
it to the Oracle CODASYL DBMS product manager.
Future meetings will be oriented towards
developing requirements for a comprehensive product support
program, longer term future enhancements based on customer
directions, and possible service offering enhancements.
Contents
Imagine the performance benefits of caching
your entire database (or significant portions) into
main-memory. Except for writes to the AIJ and periodic
flushes of the buffer pool all I/O is eliminated.
When global buffers were introduced in
version 4.3 of DBMS, it appeared that Main Memory databases
were within reach. After all, DBMS could now share 32768
buffers up to 64 blocks each (about a gigabyte). Could
systems with a gigabyte or more of main memory combined with
a massive global buffering strategy be the answer to all I/O
bottlenecks?
Even with the advances made to global
buffering in V6.0 (use of recoverable latches for managing
the Global Buffer Page Table [GBPT]), and Optimized Page
Transfers (introduced in V6.1), massive global buffering is
still impractical in a production environment.
To understand where massive global
buffering breaks down, we need to understand how DBMS global
buffers are implemented, and what resources they consume.
DBMS uses VMS global pages for caching
global buffers. VMS global pages allow multiple processes on
a VMS system to share the same physical memory. In
moderation, this strategy works well.
DBMS uses one global section that contains
the root information, global buffers and after-image journal
data. The approximate size (GBLPAGES) of the global section
can be determined from the "Derived Data" section
of DBO/DUMP/HEADER.
System Parameters (SYSGEN)
GBLPAGFIL: Global sections must be
able to be entirely mapped to the global page file (which is
part of the system page file). This value is directly related
to the size of your global buffer pool. If this is too small,
then processes will fail with an "exceeded global page
file limit" exception.
GBLSECTIONS: DBMS uses only one global section to
map the root file and global buffers. Therefore, use of
global buffers does not affect the number of global sections.
GBLPAGES: This must be large enough for all global
section pages on the system. The number of GBLPAGES required
for each database with global buffers can be estimated
from the "Derived Data" section of DBO/DUMP/HEADER.
When implementing a "massive global buffering"
strategy, GBLPAGES must also be enormous.
VIRTUALPAGECNT: When a process binds to a database,
it maps the entire database global section into virtual
memory. Prior to v6.1, each stream maps to the global
section, requiring a cumulative number of virtual pages to be
mapped.
LOCKIDTBL: Each page in the global section (whether
referenced or not) requires a page lock. Thus, this parameter
is directly impacted by the number of global buffers.
RESHASHTBL: This parameter is directly tied to the
number of locks used on the system. Under allocating this
resource will increase CPU utilization.
Working Set Adjustments...
AWSTIME: Determines how frequently working set
characteristics are sampled by VMS.
PFRATH/PFRATL: Determines the upper and lower page
fault rate thresholds to begin adjustments.
WSDEC/WSINC: Determines the size of the adjustments
to the working set. Consider setting WSDEC to 0 (zero).
WSMAX: Determines the maximum working set possible
on the system. A processes working set cannot grow larger
than this value regardless of their WSEXTENT.
Lookaside lists (pre OpenVMS v6.0)
IRPCNT:
SRPCNT:
NPAGEDYN: Contains the lookaside lists for OpenVMS
v6.0+. Use AUTOGEN with feedback to assist in correct sizing.
User Parameters (AUTHORIZE)
PGFLQUOTA: This parameter determines the maximum
number of pages that a process may use in the system page
file. Mapped global section pages (global buffers) do not
count against PGFLQUOTA. However, during an unbind, when the
global section is unmapped, memory becomes local. Thus, the
page file may take a hit at unbind time, as well as physical
memory requirements.
WSEXTENT: This parameter determines the maximum
size of a processes working set. Memory use beyond this value
will require page faulting. A value too small will cause
excessive global valid page faulting to occur.
The DBMS Monitor
It is the DBMS monitor process that
initially creates and maps each database global
section. Thus, if multiple databases are accessed
simultaneously, the DBMS MONITOR process must have sufficient
resources to map all database global sections.
What it all means...
Ok, so now we know all of the SYSGEN and
AUTHORIZE parameters that need to be adjusted when working
with global buffers. When used in moderation, these
parameters can be set to reasonable values, and global
buffers work well.
When we begin working with massive global
buffers (on the order of ¼ gigabyte or more), the process
begins to break down.
Why? If we review the SYSGEN and AUTHORIZE
parameters required to bind and unbind a database with
massive global buffering enabled, we can see that these
values must also be massive. This not to say that the
"massive" values by their self are bad.
However, mapping and unmapping global
sections is not free. The larger the global section, the more
expensive. When you consider the amount of time
required for each process (stream) to map and unmap a ½
gigabyte of memory, the bind and unbind time becomes massive.
Even in an OLTP-style environment with
database servers that pre-bind to the database, massive
global buffers are impractical. In the event of a database
shutdown or system failure, all processes will attempt to
unbind simultaniously, causing an overload of system
resources.
During a recovery scenario, multiple with
recovery processes attempting to bind simultaniously, system
resources will again be overloaded.
Even the recovery of a single process is
unacceptable-DBMS takes out a FREEZE lock which prevents any
other process from performing database work until the
recovery completes. Assuming the recovery process takes only
3 minutes to bind to the database, all other database
activity is suspended until this single recovery process
completes. Thus, if you have 500 processes attempting to use
the database, all 500 processes are suspended for the
duration of the recovery!
That's not to say...
While massive global bufffers (Main Memory
Databases), are not possible at this time, this does not mean
that global buffering itself is bad. In fact, many
environments can realize significant reductions in I/O by
utilizing moderate levels of global buffering to cache
database "hot spots".
Buffering options...
While main memory databases still elude us,
there are still many options that will help to minimize I/O
bottlenecks. First, determine the following:
1. Is access to the data evenly distributed
across available disk drives and controllers?
2. Is there a sufficient number of disk
drives available to sustain the required I/O rates?
3. Have conventional I/O spreading
techniques such as I/O balancing and disk stripping been
explored?
If these issues have been addressed, then
further steps must be taken.
Disk caching
Disk caching can also improve throughput. Software
Concepts International recommends using a write-through (to
disk immediately) cache rather than write-back (deferred
writes) unless you are operating in a single-node system.
When a disk cache is used in conjunction with DBMS
buffering, a double buffering effect occurs - DBMS first
checks its buffers for the data, then checks the disk cache
for the data. Thus, disk caches are generally serve best when
used for massive caching, and DBMS buffering is minimized.
When attempting to cache moderate amounts of data, DBMS
global buffers typically performs better than in-memory disk
caches of the same size. However, disk caching may allow
significantly more data to be cached than is practical using
global buffering.
As always, a solid recovery strategy that has been tested
is recommended.
Solid-State Disks
Solid-state (Electronic) Disks (SSDs) also offer a good
way to minimize I/O bottlenecks. Rather than eliminate
physical I/Os, as buffering or other caching strategies do,
SSDs minimize I/O bottlenecks by eliminating the seek and
rotational latencies associated with conventional disk
drives. Thus, SSDs are able to achieve I/O rates in excess of
2000 I/Os per second. SSDs also utilize non-volatle RMA, with
a battery backup that is able to survive power failures.
SSDs appear as a conventional disk drive on your system -
they connect to your system through a standard disk bus.
Thus, if your I/O bottleneck is at the bus level, then a SSD
will not help to eliminate the bottleneck.
RAM disks (a non-option)
"RAM disks" differ from SSDs in many ways. A RAM
disk is a virtual disk that resides in main system (volatle)
memory-making this too risky for a production environment. A
RAM disk is not connected a disk bus, and thus is not subject
to the bus bottlenecks of SSDs. However, this also means that
RAM disks cannot be shared across multiple systems.
In the event of a system failure, the contents of your
"RAM disk" will be lost. "RAM disks" are
far too risky to be used in a production critical
environment.
Contents
REDWOOD SHORES, Calif., March 1, 1995 -
Oracle Corp. announced the immediate availability of its
CODASYL-compliant network database product, Oracle CODASYL
DBMS V6.1. This product was formerly known as DEC DBMS and
was acquired by Oracle as part of the Rdb product family.
This version brings added power and performance to large
scale applications such as manufacturing and shop floor
control.
"Consilium has had a long history of
DBMS support with our MES application, WorkStream," said
Edward O'Rourke, WorkStream Lead Product Manger, Consilium
Inc., Mountain View, CA. "As our WorkStream customer
base is primarily comprised of DBMS users, we are indeed
pleased that Oracle is committed to the active support of the
product."
This release builds on the reputation of
previous versions by offering increased high availability and
throughput required by applications, such as WorkStream. In
addition, powerful system management features preserve the
stability needed in production environments.
"Our customers must be constantly up
and running; otherwise they are loosing money," said
Chuck Rozwat, Vice-President Oracle Rdb [and DBMS] Products
Division. "The new features offer significant reduction
of downtime with dramatic improvements in performance and
reliability."
A major enhancement supporting this demand
is multiversioning support which facilitates the upgrade
process in production environments. The new version of Oracle
CODASYL DBMS databsae can be installed and tested, while the
older version continues in production, thus reducing downtime
usually associated with such an upgrade.
"The fact that our clients can use the
new multiversioning feature to perform 'rolling upgrades' of
their DBMS environment will mean they can transition to this
new version sooner and with more confidence," said Bryan
Holland, owner of Software Concepts International, East
Nashua, NH, the leading supplier of DBMS expertise.
"It is clear that ;Oracle is listening
to and responding to the needs of DBMS customers,"
continued Mr. Holland. Significant improvements have been
made to the database verification routine. Based on
methodologies proposed by Software Concepts International, an
option to that routine allows it to perform a verification in
significantly less time than in earlier versions. "Our
experience shows that this optimization of DBO/VERIFY runs 5
to 10 times faster than previous versions," added Mr.
Holland. "And at the same time, customers get a more
thorough verification with improved diagnostics which helps
them in maintenacne tasks."
V6.1 Offers Multiversioning Support to
Further Reduce Downtime in Production Environments.
The database statistics package has also
been significantly enhanced to provide detailed information
on the real-time usage of database resources. The RELOAD AREA
utility can be used to split a database are into more
manageable pieces in a shorter period of time than previous
versions. Together, these features will allow knowledgeable
DBAs to fine-tune the database environment.
Other recent enhancements include the SQL
interface to Oracle CODSAYL DBMS through the DB Integrator
Gateway from Digital Equipment Corporation. This gateway
allows users to directly access their DBMS databases using
many of the popular front-end tools designed to use SQL.
Oracle CODASYL DBMS is available now on
OpenVMS for Digital VAX and AXP platforms. This product can
be purchased through Oracle for both named user and
concurrent usage. The product can also be ordered through
Digital Equipment Corp.
Oracle Corp., a $2 billion company with
headquarters in Redwood Shores, Calif., is the world's
leading supplier of information management softwware. Oracle
develops and markets Oracle Media Server and the Oracle7
family of software products for database management;
Cooperative Development Environment (CDE), a complete set of
tools for enterprise-wide, client/server applicatin
development; and Oracle Cooperative Applications, packaged
client/server solutions for accounting, manufacturing,
distribution, human resources and project control. Oracle
software runs of personal digital assistants, PCs,
workstations, minicomputers, mainframes, and massively
parallel computers. The company offers it sproducts, along
with related consulting, education and support services in
more than 90 countries around the world.
Consilium is the leading supplier of
manufacturing execution systems (MES) software and services.
For more than 15 years, the company has been helping World
Class companies achieve manufacturing Best Practices and
stronger business performance through better visibility and
control of their manufacturing operatoins. With its
WorkStream and FlowStream product lines, Consilium has helped
50,000 users in 20 countries solve critical business problems
by providing solutions that optimize manufacturing
performance. Found in 1987, and headquartered in Mountain
View, Califl, Consilium's customers include such Global
Fortune 200 companies as Baxter Health Care Corporations,
E.I. DuPont de Nemours, Intel, and Philips N.V.
Software Concepts International is a
computer consulting and software development firm
specializing in Digital Equipment Corporation computer
applications. SCI is the leading authority on Oracle CODASYL
DBMS-based applications. Clients include Fortune 100
corporations spanning several industries including aerospace,
insurance, electronics, automotive, precision optics,
super-comouting, software development and environmental
engineering.
Contents
Software Concepts International will be
building and maintaining a list of talented individuals with
experience in DBMS technology and related applications.
Looking for help in a DBMS related
environment?
CA/ASK MANMAN
CONSILIUM WORKSTREAM
DATABASE ADMINISTRATION
Do you have experience with a DBMS related
application?
CA/ASK MANMAN
CONSILIUM WORKSTREAM
DATABASE ADMINISTRATION
Whether looking for help or work, Software
Concepts International can help put the right people in
contact with the right opportunity.
Contents
MACPAC/D
User Conference
May 4-5, 1995,
Montreal, Canada.
Contact Dave Quilla
(310) 915-8154
Database Client/Server World
June 13-15, 1995,
Boston MA.
Contact Mary Lou Dopart
(603)881-2024
International Oracle Users Group
September 17-22, 1995,
Philadelphia, PA
Contact Mary Lou Dopart
(603)881-2024
CA/ASKUS World'95
(MANMAN)
July 16-21, 1995,
New Orleans, LA.
Call (800) CAINFO95
Consilium International User's Group
(WorkStream)
September 12-14, 1995,
Contents
Give Your DBMS Data SQL Connectivity
from Your Favorite PC Tool!
Oracle CODASYL DBMS (formerly DEC DBMS) is
still one of your rock-solid databases, supporting
applications that are critical to your business.
However, as business needs change in this
crazy world of cost-cutting, mergers, and acquisitions, maybe
you'd like a way to bring your DBMS data into the new style
of computing for the 90's - client/server.
Digital's DB Integrator (DBI) Gateway for
DBMS gives SQL connectivity to DBMS. It makes it easier for
PC users to access your DBMS data, and supports hundreds of
PC applications and tools, such as Visual Basic, SQL Windows
(Gupta) and PowerBuilder. With the purchase of just one
product - the DBI Gateway for DBMS - you can provide a
seamless link between your desktop environment and your DBMS
database, with no SQL programming required!
The DBI Gateway for DBMS
provides a seamless link between your desktop environment and
your DBMS database, with no SQL programming required!
DEC DB Integrator Gateway for DBMS provides
applications and tools with direct, read and update (modify)
access to DEC DBMS databases. This access can be via
interactive or dynamic SQL interfaces. Support for
client/server access is provided by the SQL/Services facility
and the DEC ODBC Driver for MS Windows clients, both of which
are included with the Gateway software at no extra charge.
The DBI Gateways are optimized to reduce
response time and improve network efficiency. The DBI Gateway
for DBMS selects access strategies that minimize data
transfer activity. Wherever possible, selection, filtering,
and statistical operations are done by DBMS.
Benefits of Three-tier Client/Server
The DBI Gateway for DBMS is part of
Digital's ACCESSWORKS family of software for data access and
integration. Following a three-tier client/server model,
whereby the client interface on the front-end is separated
from the data sources on the back-end by a middleware tier,
this family of software tools provides flexible, powerful,
secure access and integration to over 60 data sources from a
wide variety of clients. The ACCESSWORKS middleware can be
installed either on your existing VAX or Alpha DBMS system,
or on a third middleware server - your choice!
At the heart of the ACCESSWORKS family is
DB Integrator (DBI), a multi-database manager that allows
cross-database operations between two or more data sources.
The DB Integrator Gateway for DBMS provides native
integration with DB Integrator, further extending the
functionality of the DBMS Gateway. With ACCESSWORKS, you buy
what you need. ACCESSWORKS software can be configured to
provide connectivity to one or more DBMS data sources as well
as to integrate heterogeneous data distributed throughout
your enterprise.
Try Before You Buy
Prove to yourself how easy it is to link
desktops to DBMS. For a limited time, you can evaluate a
full-function Gateway at no cost or obligation. This way you
can experience the benefits of three-tier client/server data
access. Please find the correct software loan part number
below.
For a limited time, you
can evaluate a full-function Gateway at no cost or
obligation.
In the U.S., to receive information on DB
Integrator Gateway for DBMS call your local reseller, or call
800-DIGITAL, press 2, reference code AAF. To order a loan of
software in the U.S., call 800-DIGITAL, press 2, then 9 at
the prompt, reference code AAF. For Canada, Europe, or
Asia/Pacific Region, please call your local Digital sales
office.
To access DBMS V5.1-6.0 data (local or
remote via DECnet or TCP/IP network transports), order one of
the following:
FREE Software Loan
Try the DBI GATEWAT for DBMS free for 90
days! When ordering the FREE 90 day loan, please refer to the
following:
| Part Number
|
System |
| QL-2DAA9-LA |
OPENVMS/VAX |
| QL-2DBA9-LA |
OPENVMS/ALPHA |
Note: Client software is included with the Gateway.
Appropriate network transport software must be installed on
each client and server tier. Please refer to your Digital
ordering channel for details.
FOR MORE
INFORMATION
DIGITAL EQUIPMENT CORP.
1-800-DIGITAL
Contents
PALO ALTO, Calif., May 1, 1995 - Taurus
Software, Inc., specializing in archival software on the HP
platform for the last six years, announces its DEC DBMS
product and launches an ASK software interface for DEC VAX
users.
Managing Historical Data
Finally a way to manage historical data in
the DEC environment-Warehouse from Taurus Software.
So, how do you satisfy users wanting to
keep information forever and still maintain system speed, and
the money managers who resist buying more disc space?
ASK's UT functions to archive data provide
a partial solution. However, unless you have taken a
backup your history is gone. Getting the information back
means overloading the system all over again. If you have
upgraded your ASK software your historical data may not be
accessible with the screens and reports your users want to
use.
In response to this need voiced by ASK
users Taurus has ported their Warehouse product to the VMS
environment and has developed an interface to DBMS. In
addition, Taurus has developed specific Warehouse routines to
address the needs of the MANMAN users.
Warehouse lets you pull burdensome
historical data out of production files. Archive it to a set
of historical files which are accessible through the same ASK
screens and reports you use every day.
Archive your history later directly to
tape. And then if you need to bring anything back, simply
mount the tape and select just want you want. There is no
need to bring all of the data back onto the system. It works
even if it was archived several ASK releases ago. All needed
data conversion issues are handled automatically.
Warehouse lets you pull burdensome
historical data out of production files, archive it to a set
of historical files which are accessible through the same ASK
screens and reports you use every day.
Customer are saying...
"Warehouse paid for itself twice over
in reduced development time alone." Bill Perlack, MIS
Project Manager, Stonco Lighting.
The growth of closed sales order files at
Stonco Lighting of Union New Jersey, was creating poor system
response times, not to mention a shortage of disc space.
Stonco needed a flexible archive system that would
automatically provide a "rolling" four months of
orders, shipments and invoices, plus three years of history
in archive databases. What's more, customer service needed
reports of and fast access to data in all of the archives.
Warehouse provided the answers. Set up was easy. Complex
archiving procedures were defined, processed and tested in
less than an hour. The whole system was up and running weeks
ahead of schedule.
In discussions with ASK clients that used
both HP and DEC, Taurus heard the same issues: "We are
buying more and more disc, our application performance is
suffering because of huge amounts of data to search through,
our test databases are the same size as our production
databases so our tests take forever."
Archiving MANMAN
Taurus has worked with ASK experts to
develop archiving routines for the most common problems:
Obsolete parts, closed work orders, closed purchase orders,
closed sales orders, sales order change detail, shipping
history, GL posting detail, and closed AR items.
Test databases can also be generated simply
by not deleting the data you are pulling from your production
environment.
What's more, our technical staff is ready
to help you develop your own applications for those portions
of ASK which you may have supplemented or modified.
Taurus software offers demonstrations of
Warehouse, just call Taurus at (415) 961-1323 x110.
FOR MORE
INFORMATION
TAURUS SOFTWARE, INC.
1032 ELWELL CT, SUITE 245
PALO ALTO, CA 94303
(415) 961-1323
Contents
By Dan Esbensen
Perhaps one of the most frustrating aspects
of working with a network style database, is the apparent
lack of productivity tools. INTOUCH, a high-performance 4GL
development environment provides an effective solution to
this problem.
Touch Technologies, Inc., of San Diego, CA,
recently announced their Oracle CODASYL DBMS (DBMS) interface
to their high performance 4GL language, INTOUCH.
The INTOUCH interface to Oracle CODASYL
DBMS (DBMS) provides 4GL access to the DBMS database engine.
The design goals for this project were to provide a high
performance interface to DBMS that provides the ease-of-use
of a 4GL while maximizing database concurrency and minimizing
locking.
Set types and currency indicators are
concepts that are unique to DBMS. Although they provide DBMS
with efficient access paths among records, they also serve as
a source of confusion for 3GL programmers. A set type
is a defined relationship among different record types in a
database. A set occurrence is a single instance of an
owner record that has zero or more member record occurrences.
Currency indicators serve as place holders in the
database. Currency indicators mark your position within the
database, each realm (one or more files), each record type,
and each set-type.
From a performance and locking standpoint,
each currency indicator places a lock on a record in your
database, thus restricting access to that record. Excessive
active currency indicators mean excessive record locking.
The 3GL problem...
The biggest problems faced by most DBMS
development environments are low productivity and excessive
locking. Productivity suffers as a result of dealing with the
complex issues of set types, set occurrences, and currency
indicators, all of which are necessary to navigate a network
style database using a 3GL programming language.
Traditionally, most DBMS development
environments have been weak in their management of currency
indicators and thus become burdened by excessive lock
conflicts and unnecessary overhead. Mismanagement of currency
indicators is a natural result of dealing with the
complexities of currency indicators when programming in
traditional 3GL/Data Manipulation Language (DML) languages.
This results in poor database performance, low transaction
throughput, and in extreme cases complete database deadlocks.
Currency indicators, a necessary evil...
Programming for a DBMS database would be
impossible without the use of currency indicators. However,
each currency indicator places a read-only lock on your
records. By default, DBMS updates all possible
currency indicators for each DML access -- thus locking
records with unused currency indicators. Programming around
this in traditional 3GL languages requires significant
expertise in the use of currency indicators, and additional
code.
Most other 4GLs either do not support DBMS,
fail to make use of its sets, or are so cavalier with the use
of currency indicators as to render them unacceptable for
multiple user environments.
Most other 4GLs either do not
support DBMS, fail to make use of its sets, or are so
cavalier with the use of currency indicators as to render
them unacceptable for multiple user environments.
The INTOUCH solution...
INTOUCH solves these problems by providing
an easy-to-use 4GL interface to DBMS that minimizes the
programmers' need to manage complex currency issues. INTOUCH
makes full use of DBMS sets, including indexed access by key
values, range retrievals, and access to the owner record of a
set occurrence. INTOUCH understands database set
types, and automatically handles' currency indicators.
In fact, INTOUCH automatically establishes currency
indicators only when necessary, and releases
them as soon as they are no longer needed.
Mixed database environments
Most application environments require
access to more than just one data source. Using INTOUCH, it
is easy to access data from different databases from within a
single application, all sharing the common INTOUCH syntax!
Thus, programmers don't need to know to access RMS, DBMS,
Rdb/VMS, or ADABASE -- INTOUCH provides a common interface to
your data. Within a single program, it is easy to mix and
match data from all of these sources!
Programmer Productivity.
Because INTOUCH understands set types and
currency indicators, INTOUCH relieves the programmer from
dealing with these complex issues. Programmers spend more
time solving business problems and less time thinking about
how to navigate the database.
Most application
environments require access to more than just one data
source. Using INTOUCH, it is easy to access data from
different databases from within a single application, all
sharing the common INTOUCH syntax!
Using INTOUCH it is very easy to make
sophisticated user interfaces and screens, including pop-up
windows, for interactive applications. While many languages
require the use of external libraries or outside routines,
the INTOUCH language has all screen handing and keyboard
mapping tools built right in. The powerful formatting
capabilities of the INTOUCH language make report generation
equally easy. Creating complex reports is easy, since
programmers concentrate on the report format, rather than how
to access the data needed for the report.
Locks Less, works great!
Because INTOUCH only updates those currency
indicators that are required to navigate your database,
INTOUCH locks less. This vastly improves database
concurrency, allowing more users to simultaneously access the
database with greater throughput.
Can INTOUCH really help?
To find out how much INTOUCH can improve
your development productivity, call Touch Technologies for a FREE
30 day trial.
To understand the significance of INTOUCH's
lock management, try the following example on your DBMS
database.
The following two DBQ* examples illustrate
the reduced locking achieved when programming in INTOUCH
versus typical 3GL programming. These examples illustrate the
lock reduction possible when using INTOUCH. They are not
examples of the INTOUCH language syntax.
Typical locking in a 3GL
dbq> FREE ALL CURRENT
dbq> LOOP
dbq> FIND NEXT
dbq> END
dbq> SHOW LOCKS
Typical locking using INTOUCH:
(This is NOT an example of INTOUCH code)
dbq> FREE ALL CURRENT
dbq> LOOP
dbq> FIND NEXT RETAINING
dbq> END
dbq> SHOW LOCKS
Compare the number of locks held after each
test. INTOUCH locks less!
FOR MORE
INFORMATION
Touch Technologies, Inc.
9988 Hibert St., Suite 310
San Diego, CA 92131
(800) 525-2527
Get INTOUCH !
Discover the difference with
INTOUCH, the high-performance 4GL for VAX/VMS systems!
Call Touch Technologies,
Inc.
for additional information or
a FREE evaluation copy of INTOUCH!
Call Touch Technologies,
Inc. at
(800) 525-2527 or (619) 455-7404
* DBQ is a database query utility included
with DBMS. It is a utility that lets you interactively
retrieve, update, and store database records using COBOL-like
DML syntax.
Dan Esbensen is an associate VAX/VMS
consultant with System Management Software, Inc. and
President of Touch Technologies, Inc. Dan has over 11 years
experience with VAX/VMS.
Contents
For Sale:
Good Used Equipment
Decorative Specialties Int'l is looking to
sell the following used hardware. If you are interested, call
Mark at (413) 539-5259.
- Cabletron Systems (2)
- FR-300 W/Landview
- Fiber Optic Repeaters
- Printronix MVP150B Printers (6)
- TK70 Tape Drive (new)
- T1035-A1P4
- VXT2000 X-window Terminals (4)
- 14 meg - VMR17-HA
- LA210 Printers (10)
- LN03 Printers (8)
Contents
COPYRIGHT © 1995 by Software Concepts
International. All rights reserved. No part of this
publication may be reproduced in any form without written
permission from the publisher, and the inclusion of this
copyright notice.
| Phone |
USA Toll-free (Sales) |
Fax |
e-mail |
| (603) 879-9022 |
1-888-CODASYL |
(603) 879-9023 |
Copyright © 1996 Software Concepts
International Inc. All rights reserved.