Hi,

How can I improve the communication and the data transfert betwee n 2
servers across a 1ghz LAN?

I have a data warehouse on a server and my OLAP cubes on a second server.
the data transfert never takes more then 5% of the 1gbps bandwidth
My SQL query is optimized but I have 3 tables linked

my fact table contains 40 millions of rows and its a partitioned table.

my database server has 4 proc / 4Gb ram
during the execution of the select statement my server suffer CXPACKET wait
types. I know I have an hard drive issue on the server (its a SAN).
only my tempdb database has been moved to a local non raid drive to share
the performance.

my source database is in readonly mode.
I know that I can reach more then 30% of usage of the bandwidth between the
2 servers.

but there is anything to do?
any option to change? (in sql server and/or the olap server) to improve
this.

thanks for your comments.

jerome.

"Jéjé" <willgart (AT) AAAhotmailBBB (DOT) com> wrote

Which server is performing the join? If it's the OLAP server, consider
pushing the join to the source server with OPENQUERY or a remote view.

David

the join is a simple select * from table1, table2 where table1.key =
table2.key
so the query is executed in the SQL Server database on the database server.

OLAP Server = Analysis Services

"David Browne" <davidbaxterbrowne no potted meat (AT) hotmail (DOT) com> wrote in
message news:ub1yxHCoFHA.2152 (AT) TK2MSFTNGP14 (DOT) phx.gbl...

You might also take a look at the IO throughput which is likely the
performance bottleneck...

--
Wayne Snyder, MCDBA, SQL Server MVP
Mariner, Charlotte, NC
www.mariner-usa.com
(Please respond only to the newsgroups.)

I support the Professional Association of SQL Server (PASS) and it's
community of SQL Server professionals.
www.sqlpass.org

"Jéjé" <willgart (AT) AAAhotmailBBB (DOT) com> wrote

I know I have issue here, and I can't do anything.
I'm on a SAN.

the IT Team in charge of the SAN has not configured the system like we have
recommanded, and also there is performance issue not solved.
And we don't know if the IT team can change anything in the configuration.

I think our server is connected to the SAN through a shared system
instead-of having a dedicated communication, set of disks and controllers.

The problem appear when I read & write simultaneously. an older server from
another client perform better than this new server, but simple requests are
better on the new server (like create index or select statement)

Due to the lack of support and knowledge from this team, I have to find
other solutions.

"Wayne Snyder" <wayne.nospam.snyder (AT) mariner-usa (DOT) com> wrote