Benchmarking and reliability studies on Advanced Replication

[Henrique Silva_3]

I am trying to get a handle on what kind of performance, reliablity issues are there with advanced replication accros the WAN ( say from US to Europe and Asia ).

I am writing a paper on single server vs multiple servers woth replication topology, and can not find benchmarking at all on such issue.

If you think I can not achieve realiable "almost" real time replication over the WAN ( two-way ), could you let me know ? If you have tried and failed, can you tell me why ? Is one-way replication to a fail over server better ? Or a standby server ? I would like to have at least one way replication so that I could run reports of this second server, isntead of making it standby on a mounted state !!!

Thanks for any help here,

Henrique Silva

[Graham Cameron_1 · Accepted solution]

I did a lot of work with replication on Oracle 7 in the late 1990s, presented a paper to UK OUG in 1996.
We were using the updatable snapshots model. I can provide much more info if this is of interest.
The major drawback we had was that a distributed application whoch worked well in UK fell apart when implemented in USA due to long network latency of packets over the larger distances.
This was SLQ*NET v2 and NET*8 was meant to improve things, but I have moved on since.
Let me know if you want more.

Graham

[Henrique Silva_3]

Definitely Graham.

If you have such paper, could you send it to me or point me on the right direction ?

I am too worried about the network latency, but if we can establish a private HUGE pipe between the regions, and net8 has fixed some problems, I would like to at least try to implement such setup, and see what the short comings are.

Do you have info on how you did the benchmarking also ?

Thanks,


Henrique Silva

PS.: If you want to send info directly to me, please send to my email

henrique_silva@hp.com

[Henrique Silva_3]

Graham.

ANy possibility for you to send me that paper ?

I am looking for a two way replication scenario, and want to physically prove it that it can or can not be done.

I will setup a test environment soon, and it would be good to learn fro others here.

Thanks,

Henrique Silva

[Graham Cameron_1]

Henrique
The article was published in the summer 1996 issue of Relate, the UKOUG magazine at that time. I don't believe it is on-line but I could fax it to you (3 pages).
Our problems were latency - ie the time for a packet to travese the US - and the fact that sqlnet v2 was inefficient in terms of it needing lots of small packets each of which needed an accept reply before the next packet would be sent. A HUGE pipe would not have helped much.
I am not working in this area any more but am trying to make contact with an ex-colleague who is.

[Henrique Silva_3]

Thanks Graham.

If you would not mind faxing me taht paper, my fax number is 781 505 7703.

Maybe you could scan it and send it to me via email !!!!

Also, do you have any clues with Parallel Server accross networks ? I am looking into replication accross continents( america, Europe, Asia/Pacific and South America ), and need to find out HOW to get one instance to work accross this !!! :-(( I am trying to setup a mock install right now to start to measure this.

Cheers,

Henrique Silva