SQL Performance

Por mi parte no puedo opinar sobre otros productos dado que no tengo la experiencia necesaria.
En el caso de Warehouse (con SQL) con bases del orden de 1Tb, el tema pasa bastante por el hardware.
Por ejemplo, te recomendaría estrutura SAN de discos con RAID 5, ya que para esos tamaños RAID 10 es de costos imposibles.

Mucha capacidad de proceso (2 o más procesadores), mucha memoria (para eso SQL 2000 Enterprise, ya que SQL 2000 standard solamente maneja 2Gb).



Luis Martin
Moderator
SQL-Server-Performance.com

All postings are provided “AS IS” with no warranties for accuracy.

Si el RAID 5 es a través de SAN con fiber, la considero correcta. Ahora si es SCSI normal, es probable que tengas cuellos en disco.

La siguiente es una parte del Newsletter que el forum envía cada mes. El artículo ha sido escrito por Brad, que es el Administrador de la Web.

VLDB (Very Large Data Base) Performance Tuning and Optimization

Because of their massive size, working with VLDB databases takes
longer that you would expect. Starting and stopping the SQL
Server services, creating or expanding databases, running DBCC
commands, backing up and restoring, all take much more time than
the average databases most DBA's are used to. These operations
could take anywhere from a couple of hours to a couple of days to
complete. Don't forget to take these time delays into
consideration when scheduling database maintenance. [6.5, 7.0,
2000, 2005]

*****

Because of the sluggishness of most VLDB operations, don't even
bother trying to use Enterprise Manager to maintain VLDB.
Enterprise Manager will appear to lock up under many operations.
Instead, use stored procedures and Transact-SQL to maintain your
VLDB. In a few instances, some of the tasks you need to perform
can only be done through Enterprise Manager. For these cases, you
have no choice but to use it. [6.5, 7.0, 2000, 2005]

*****

Because of the large amounts of hard disk space involved, RAID 5,
not RAID 10, is recommended. RAID 10, while faster than RAID 5,
uses many more physical drives. When dealing with terabytes of
data, the number of extra drives can not only be cost
prohibitive, but it might be difficult, if impossible, to connect
all of the drives to your server due to a lack of available
connections on the server. Fiber channel is preferred over SCSI
because of performance, and the larger the disk cache in the
controllers, the better. [6.5, 7.0, 2000, 2005]

*****

Consider a System Area Network (SAN) for VLDB storage, especially
if you want the fastest possible backups, restores, and data
loads. [6.5, 7.0, 2000, 2005]

*****

Network connections should be of the highest speed, such as
100Mbs Switched Ethernet, or even 1Gbs Switched Ethernet. [6.5,
7.0, 2000, 2005]

*****

Generally, DBCC commands are not run on a routine basis on VLDB
because of the times involved, often greater than a day. Instead,
only run them if you suspect possible hardware-related problems
may have occurred. [6.5, 7.0, 2000, 2005]

*****

Use filegroups to divide up the database into units that are
relatively easy to backup and restore. It is virtually impossible
to manage VLDB backups and restores without judicial use of
filegroups. [7.0, 2000, 2005]

*****

Queries that join two or more very large tables should ideally
have each table in a separate file and on a separate physical
drive, and if possible, a separate I/O controller. [7.0, 2000,
2005]

*****

Be sure you have the "automatically grow file" and "auto shrink"
options turned off on your VLDB. These both take a large
performance hit on VLDB. Instead, perform these tasks manually
only when needed. [7.0, 2000, 2005]


Luis Martin
Moderator
SQL-Server-Performance.com

All postings are provided “AS IS” with no warranties for accuracy.

Proba con estos dos lugares:

http://www.microsoft.com/windows2000/techinfo/howitworks/communications/winsock.asp
http://www.dothill.com/



Luis Martin
Moderator
SQL-Server-Performance.com

All postings are provided “AS IS” with no warranties for accuracy.