Emmanuel,
Thank you for your response. In the coming month we'll be going ahead
and testing Sequoia in a lab environment to verify compatibility with
our various tomcat and jboss instances. We're really not looking for a
big performance boost so much as high availability in the event of a
server failure. Management is not a big deal since we are a fairly small
shop and the reason we're even considering Sequoia is due to cost
constraints. We will gladly write our own management scripts to save a
few dollars.
You mentioned the Myosotis project which provides functionality for PHP
and stated that transparent failover is not possible. How is this so?
Can you explain in further detail? I would expect that since the client
apps are making calls to the Sequoia virtual database there wouldn't be
any issues if a backend server were to go down.
All of our database servers reside in a single physical location so
network partitioning is not a concern of ours at this time.
One of the notable unsupported items is distributed joins. Are there any
other limitations that you can explain easily via email?
Thanks again and I'm sure I'll have additional questions to post to the
list once we get started with our testing. Is the community fairly
active? Why do you state that not many users have been willing to share
their experiences with Sequoia? One of the hesitations my developers
have with deploying this solution is the lack of verifiable real-world
feedback from users.
Thanks,
Ryan Manikowski
Emmanuel Cecchet wrote:
Hi Ryan,
Hello. My company is currently looking to evaluate Sequoia as a mysql
clustering solution and I am wondering if anyone can share their
experiences with the software.
Very few people are willing to share their experience in using open
source Sequoia even though there are quite a bit of production users
out there. However I can tell you what I have seen so far (my opinion
is probably biased since I was the designer of Sequoia ;-)).
We may potentially be rolling out the service in our production
environment and would like to know of any glaring limitations of the
software if there are any. Also, can anyone point me to configuration
documents for getting the software up and running?
You can find the documentation at http://sequoia.continuent.org/Manuals
There is no fundamental difference between 2.x and 3.x but 3.x is now
deprecated and not supported (4.0 is in the works). Config file format
is slightly different.
Besides what the bug tracker can tell you, I would say Sequoia is a
solution that is more geared towards high availability than
performance. Depending on the parallelism you have in your workload
you may see no or little improvement in performance. Note that it is
still the same database that will in the end execute the query, so
individual queries won't execute faster, we just spread the load on
multiple nodes. Writes are replicated on all nodes, so you will never
see improvement on write speed performance.
Sequoia has all the required management mechanics but it does not
provide automated management, you will have to write your own scripts
for that. Continuent provides a commercial product based on Sequoia
that will handle that nicely. If you prefer staying the open source
route, you can seek for help on the mailing list or get help from
consultants (I do consulting ;-)).
How well does the software work with PHP? Most of our backend systems
are tomcat 5.5/6.x but we still maintain legacy PHP software where
needed.
The software was originally designed to work with Java but the
Myosotis project adds direct MySQL protocol support so that you can
run with legacy PHP applications. Just be aware that you won't have
transparent failover with Myosotis. Another option is to use
ODBSequoia, an ODBC implementation if you PHP application uses a db
library that can take an ODBC driver.
Sequoia is highly configurable which sometimes confuses users and can
lead to poor design or integration. We are trying to simplify
configuration for Sequoia 4 but there will still be application
specific optimizations or architectures that cannot be automatically
guessed. One thing to keep in mind is that Sequoia provides
synchronous statement-based replication and the strong consistency
provided does not support network partitions. Therefore, you have to
think twice if you want to use it in a WAN environment.
If some of the things I mentioned are not clear to you or don't ring a
bell, you can read this article that should give you an overview of
the problems you are going to face depending on what solution you go
for (Sequoia or not): http://infoscience.epfl.ch/record/118488
Thanks for your interest in Sequoia,
Emmanuel
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
