I am trying to build a x64 c# application using the .Net 2.0 framework. However, I do not have the x64 version of the system.data.oracleclient. Can I get that dll only? Or must i build and deploy with the .Net 3.5 framework? Is there a way to just download the x64 version of this dll?

Thanks very much.

There is, visit oracle.com. There's an interesting link in the right rail of this web page.

---

Hans Passant.

Thanks, but I am trying to find the Microsoft driver for Oracle, i want to make the smallest amount of changes possible (at least at this pass).

There isn't one, you'll need Oracle.

---

Hans Passant.

Thank you again, but that sounds like a disaster. I read a link that .NetFramework 3.5 SP1 has the x64bit version. Is that not true? I have reached what I believe to be a limit on performance under x32. I have a mutli threaded app on an x64 windows server with plenty of cpu's and ram, but x32 runs only so fast. When i made 2 copes of the same program and ran them i had much greater performance. Since i am going to run my code as a service, i would the full capacity of in 1 instance. Is there any way around this under x32 or must i switch to x64and as a result move to the Oracle ODP.NET?

Andrew

You ask me to comment on a link but didn't provide it so I could double-check the claim. As long as I can't, I'm pretty sure it is not true. If two copies of your program run faster than your multi-threaded single program, there's something wrong with the threading code. 64-bit code is no faster than 32-bit code, the only advantage is that you have more virtual memory space so you're much less likely to run out of memory. And get a benefit from having more than 4 gigabytes of RAM when you program actually needs that much.

Oracle is supposed to be a reputable company, using their product ought not to be a disaster to any of their customers. If it is, I'd have to recommend to quickly drop them. If it isn't, installing their provider should help you verify your assumptions in but a few hours.

---

Hans Passant.

Hans,

Thank you again.

Here is the link:
http://social.msdn.microsoft.com/Forums/en-US/netfx64bit/thread/3cbc1c81-e0a8-4d00-b0a6-55be2c099bf5

While I have downloaded SP1, I have not found the file, but perhaps that is because my desktop is a 32bit desktop and the installer did not include those files (i am not sure).

Nothing to do with the reputation of Oracle, just am trying to minimize the code that I need to change. I already tested the oracle driver conversion and found that there are some functional differences (e.g. No Discover Parameters). So i would need to be ready to implement new functions to keep our current API the same. My comment regarding disaster was directed towards Microsoft as I would have expected with all the x64 servers, they would provide a version of their own libraries so that developers would not have to look elsewhere.

I agree with you that x64 is not faster, but it is exactly the memory squeeze that i believe was slowing my app down. I developed a library that will run as a windows service, but i test it as a windows application as well.

The function that i was testing is the reading data from a file and writing to a database. It is a 'highly' multi threaded application using many threads to basicaly write entity data in parallel to the db. In addition, it was designed to process more than 1 file in parallel.

Here was my observation which lead me to believe that x64 would provide greater capacity and increased throughput (and hence, seem as though it runs faster).

When the 2 files are being procesed in parallel under a single x32 application, i observed my trhoughput of mastering enties to the database of a rate at 2500/minute.

Changing nothing else on the databse or application, other than processing a single file under a single instance of the application and running the other file on another instance, i was able to processes 18,000 entities per minute!

The processing of the files is still mutli threaded and they all use the same class, that is why I do not believe it was a threading problem, but i wanted to verify that if i ran as a single x64 instance, would i get the same throughput.

If i did, then i solved my problem; If i didnt, then i would have confirmation that i need to look at my threading again.

Sorry about the detail, but i wanted to give you all the background.

Appreciate any assistance.

A

I cannot explain your observations, it makes no sense. As such, predicting behavior of your code in 64-bit mode would be nothing but a wild guess. Why don't you try it so you'll know for a fact.

---

Hans Passant.

Check out this performance comparison by Microsoft:

http://www.microsoft.com/downloads/details.aspx?FamilyID=35E2738A-79A3-4FA0-A9BF-06C52077592D&displaylang=en