Author |
Topic |
mailtonoorul
Starting Member
3 Posts |
Posted - 2014-06-04 : 01:07:18
|
How to Position 5 database server for an centralized database for an application?
1) We are providing a e governance solution for an organization,where we are providing a centralized database,Client have provided 5 Database server for the same.how can we position the Database Server? there are 5000 Concurrent users and 25000 users,SAN Storage for approx. 60 TB,Database size of 2 TB and growth of 1 TB every year
2) How many instance can we have for above said Case?
3) How much RAM Required ?
Noorul Ameen |
|
tkizer
Almighty SQL Goddess
38200 Posts |
Posted - 2014-06-04 : 12:12:09
|
If there is only one database, then I would probably setup a 4-node SQL Server 2012 Availability Group. 2 nodes would be at the primary site, both sync replicas. 2 nodes would be at a DR site, both async replicas.
If there's only one database and if you use Availability Groups, then you would have a standalone instance on each replica.
Get as much RAM as can be afforded.
Tara Kizer SQL Server MVP since 2007 http://weblogs.sqlteam.com/tarad/ |
 |
|
jackv
Master Smack Fu Yak Hacker
2179 Posts |
Posted - 2014-06-05 : 01:36:03
|
@tara - if the primary and DR sites are close - would you consider synch replication between Primary and DR sites? This is assuming performance was acceptable
Jack Vamvas -------------------- http://www.sqlserver-dba.com |
 |
|
tkizer
Almighty SQL Goddess
38200 Posts |
Posted - 2014-06-05 : 12:14:36
|
If performance is acceptable between two sites, then yes you can do sync. However, you can only have a total of 3 sync replicas (including the primary). That's in SQL Server 2012 at least, 2014 might give you more. I know you get more async replicas in 2014.
So in 2012 with a 4-node AG, you'd still need one async.
Typically the network latency between a primary and a DR site is too high to use sync. Even at 12-15ms (300 miles was the distance), the performance degradation was huge for us.
Tara Kizer SQL Server MVP since 2007 http://weblogs.sqlteam.com/tarad/ |
 |
|
jackv
Master Smack Fu Yak Hacker
2179 Posts |
Posted - 2014-06-07 : 01:59:59
|
Thanks for the feedback, I've recently been testing sync repliaction with about 25 miles and the performance has been acceptable. Of course, I'd like it to be quicker but the benefit gained , it's worth the slight degradation in performance. Some applications tolerate the performance hit OK , but other , more time sensitive apps are not so tolerant. The unpredictability of the network is the biggest issue
Jack Vamvas -------------------- http://www.sqlserver-dba.com |
 |
|
tkizer
Almighty SQL Goddess
38200 Posts |
Posted - 2014-06-08 : 12:57:20
|
Yeah we have very, very high performance requirements. I used to support a system that could not tolerate anything being over 300 milliseconds. It timed out at 300ms. Adding 12-15ms would have meant many more timeouts.
Tara Kizer SQL Server MVP since 2007 http://weblogs.sqlteam.com/tarad/ |
 |
|
jackv
Master Smack Fu Yak Hacker
2179 Posts |
Posted - 2014-06-10 : 01:42:07
|
In those situations ,we've implemented - similar to above - i.e sysnch replication on one site , asynch to the failover site. With a very level of monitoring. Aiming to keep variability as low as possible
Jack Vamvas -------------------- http://www.sqlserver-dba.com |
 |
|
|