departmental vs corporate database
Doug Stone
For years, ASA has been cast as a Departmental Database.
ASE has been mentioned as the Corporate or Enterprise Database from Sybase.
From my perspective, when I hear those labels, I think in terms of
concurrent users, resource consumption, performance, and (with our product)
to a lesser extent, failover capability.
How do ASA and ASE compare in those categories?
Also, how do ASA and Oracle compare?
These questions do not spring from just idle curiosity.
Our industry is in a consolidation mode. The question "which database?" is
often asked.
Thanks,
Doug
Andre Schild
> For years, ASA has been cast as a Departmental Database.
> ASE has been mentioned as the Corporate or Enterprise Database from Sybase.
>
> From my perspective, when I hear those labels, I think in terms of
> concurrent users, resource consumption, performance, and (with our product)
> to a lesser extent, failover capability.
>
> How do ASA and ASE compare in those categories?
I think ASA provides everything ASE provides, with a lot af addons.
In general the "only" benefits of ASE are in number of concurrent users
(above several thousands)
>
> Also, how do ASA and Oracle compare?
>
> These questions do not spring from just idle curiosity.
> Our industry is in a consolidation mode. The question "which database?" is
> often asked.
>
There is only one answer: ASA <vbg>
Seriously: It depends on your needs, but with ASA 10 you will also get
hot-failovers, and you already have support for the range from handhelds
up to many GB-Size databases.
André
Jack T.
I agree with Andre. We work with MSSQL, which came from ASE, Oracle, and
ASA. When we have a choice, like we normally do, we get the customer into
ASA, even if they own one of the others. How I escaped DB2, I'll never
know. It's the number one database, but I don't know anything about it
anymore.
>From my perspective, when I hear those labels, I think in terms of
>concurrent users, resource consumption, performance, and (with our product)
>to a lesser extent, failover capability.<
It's difficult to give good advice without having real numbers. When I hear
"to a lesser extent, failover capability", that's a smaller environment than
I'm accustomed to working in. Concurrent users and gigabytes doesn't mean
anything to me. How much and what kind of activity in large tables and how
many tables at a time is what matters.
Some years ago I was featured in ENT and Application Development Trends for
developing an ERP system. I considered Oracle and MSSQL, and money was not
an issue. It later ran a $300 million dollar company. That ERP system used
SQL Anywhere. What I want to know is what you mean by Enterprise. How much
data are we really talking here. ASA can handle an aweful lot of companies
where DB2, Oracle, and MSSQL are being mis-applied today. It's a lot more
than a numbers game when it comes to speed. Some get their speed by being
stupid. MySQL and Btrieve, which isn't a DBMS at all, are a couple of
examples. Sybase calls ASA a workgroup product. It can can handle a 300
million dollar workgroup with inputs from factory floor, office, SPC,
barcode, EDI, MRP runs.
- Oracle will hold a lot of data. Has good integrity. No way on earth to
beat this thing at two things. One is row inserts and the other is table
scans. You won't believe they actually happened. It's a good thing it's
good at table scans because it has a pitiful query optimizer. If you don't
hand optimize queries, it will annoy a saint. If you don't make a special
version of your app for Oracle, queries will run faster if you drop the
indexes. It's nickname in the ERP industry is Dr. Slow. People often try
to fix it by throwing iron at it. If you have a special version of your
program from just Oracle, like SAP, then it works. Oracle is different than
any other database that you will work with. They have non-standard names
for industry standard functionality. If they made cars, they would call
them pneumatically suspended personal transporters. It helps to know DBMS
concepts before Oracle so that you can understand the documentation. It's a
solid database. Can be very fast if you are willing to work. It's got more
buttons than your mother's sewing box. It's great if you need it. The
tools that used to ship with it look like they were put together by a 9 year
old. At least with later versions they don't GPF and corrupt. There are a
lot of companies that offer tools to compensate for Oracle's deficiencies,
including query optimizing tools. It's position in the market is kept high
by plenty of hot air from Larry E. and advertising. The latest that I've
used is 8i. They apologize after every release for their query optimizer
and how the new one is much better.
- MSSQL will hold a lot of data. Good query optimizer. You will run into a
lot of instances in development where DDL RI actions can't be defined.
(Cascade/Restrict, etc.) This means you have to write triggers or work in
an app development environment that maintains and generates the language
statements for you so you can maintain the relationships. It's "lets
pretend we have data integrity". Not great concurrency because lots of
instances where locks get escalated to page locks. That is not always a
problem though if you don't have big batches to post that require atomicity.
Poor trigger control. (No, insead-of-triggers are not a substitute.)
Decent tools that are better in 2000, than 2005. OTOH, 2005 has separate
DATE and TIME types. It has the same RI action limitations/bugs/"features".
It's pick your poison. It's easier to administer than Oracle but less you
can do.
- I know zero about recent versions of ASE. Apparently not many people do
since it has less than 3% of the enterprise database market, and that mostly
from legacy. MSSQL was built from this code base. The last time I looked
at it was when Sybase bought Powersoft along with SQL Anywhere. Sybase said
this is where you go when you grow up. I decided then that I didn't ever
want to grow up. I haven't used recent versions.
- ASA is heaven if you can live there. It is significantly slower than
MSSQL at doing record inserts on large datasets, and way slower than Oracle.
It's query optimization is very good if you set up the database right. The
tools in ASA are first class and better than you will see in the other two.
ASA's integrity is rock solid. You have all the trigger control you could
ask for. It has text book DDL RI with actions. Backups and recoveries from
real-world data loss is better an easier than the other two. Real world
data loss isn't a drive going bad or a server down. Real world data loss is
an OOPS! Then you edit the oops out of the log file, restore last nights
backup, and replay todays log minus the oops. Going to last nights backup,
and entering 3 shifts of data from memory to get around an oops is a joke.
Even playing forward to a time stamp is a joke and then trying to remember
what happened before and what happened after. It takes them half an hour
before they get the courage to tell you what they did, and the rest of the
place has been working all that time. Their best estimate is sometime after
break. Then what? Tell people to only re-enter what they entered after
some arbitrary time you pick? ASA is the database I have the least trouble
with in the field, and the one that you can get back on its feet the
quickest.
If you can use ASA, by all means use it. You know the issues with MSSQL.
If you get pushed into that area, I'd recommend checking other options
first. Personally, I lineup with someone who knows DB2 and see once if the
PC version of that DB is real. If it looks good, let me know. I looked at
it once when the PC version came out and it was useless. IBM was using it
for Midrange bait. It may be real on the PC today, but I'm not the guy to
ask. Since then they bought out Informix and moved to the number one spot.
It may not even be the same product.
Low cost/Free: Firebird, poor query optimizer, no mixed order indexes, no
log file, slow, unreliable, but good story line. It can do backups, sorta.
Postgres I've seen perform on FreeBSD very well. New to Windows. The
backups are points in time but non-editble log files, but you can restore to
a certain time. I don't remember everything that cooled me down on that
one. It seems like one thing is it had no mixed order indexes as with
Firebird. I'm not an expert, but I learned what I need to know. Ingres I'm
just starting to play with. Mediocre query optimizer. No distinct date and
time types. Good replication. Good docs. MaxDB currently won't even
install in Windows. I'm on their list and they said they would fix it soon.
MySQL, very limited. Can be fast if the traffic is local, meaning it's not
too smart. Like everyone else, I use it on web sites. I spent a lot of
time prospecting in open source. There are a lot of people out there with
religeon making fantastic claims about their favorite open source database.
I went throug a lot of excitement and a lot of disappointment. I wouldn't
be on a Sybase forum if I had hit pay dirt. I hate to be a realist, but I
have to say that I can pay for ASA out of my project budget cheaper than
using one of the free ones.
I have found that a lot depends on what people have used before. If they
used the Access database, then they think MSSQL is a miracle. Oracle guys
looking at MSSQL are reminded of their favorite Disney Character. They
think ASA is a toy. ASA people that have used all three are thinking
favorite Disney Character and work. They don't wanna go either place.
Good luck,
Jack T.
"Doug Stone" <nospamdstoneatres-qdotcom> wrote in message
news:43185aae$1@forums-2-dub...