June 11, 2009 at 9:31 am
Hi,
We are planing to scale out our Database and if the number of users will increase to 5000 then what all the parameter I need to consider?
Right now 500 GB HDD & RAM 2 GB
Cheers,
Got an idea..share it !!
DBA_Vishal
June 11, 2009 at 9:10 pm
You need to consider scaling up your processor, RAM, Netwrok Speed, Fast disk subsystem apart from DB load balancing. If number of users increase then start thinking for database load balancing, by configuring Database partitioned views , Shared server or Replication.
"More Green More Oxygen !! Plant a tree today"
April 30, 2012 at 4:04 pm
Minaz Amin (6/11/2009)
You need to consider scaling up your processor, RAM, Netwrok Speed, Fast disk subsystem apart from DB load balancing. If number of users increase then start thinking for database load balancing, by configuring Database partitioned views , Shared server or Replication.
There isn't db load balancing in sql world 🙁
April 30, 2012 at 4:08 pm
sqldba_newbie (4/30/2012)
Minaz Amin (6/11/2009)
You need to consider scaling up your processor, RAM, Netwrok Speed, Fast disk subsystem apart from DB load balancing. If number of users increase then start thinking for database load balancing, by configuring Database partitioned views , Shared server or Replication.There isn't db load balancing in sql world 🙁
No, but you can fake it. IE: All read queries pull from replicated copies of the DB while you maintain the primary for DML only.
That said, however, 5000 users shouldn't really be a lot, but it depends on the use case of the database. I'm pretty sure the HP website (as a random example) handles a lot more than 5000 simultaneous users browsing their products.
If you're scaling the usage, you need to look at the current usage and determine if you have any existing bottlenecks or are close to hitting one. Then review the use case of the additional users and attempt to determine if they're going to affect those bottlenecks heavily.
This is one of those cases where I recommend you bring in an expert for a week or two to review your system and intended changes and get you a recommendation for a design path for your regular staff to review and implement. It's not a linear kind of thing, it all revolves around the nuances of the load increase.
EDIT: I just realized this was a 2 year old necro. Well, the OP won't care but anyone else who trips on it might.
Never stop learning, even if it hurts. Ego bruises are practically mandatory as you learn unless you've never risked enough to make a mistake.
For better assistance in answering your questions[/url] | Forum Netiquette
For index/tuning help, follow these directions.[/url] |Tally Tables[/url]
Twitter: @AnyWayDBA
May 1, 2012 at 8:39 am
Evil Kraig F (4/30/2012)
sqldba_newbie (4/30/2012)
Minaz Amin (6/11/2009)
You need to consider scaling up your processor, RAM, Netwrok Speed, Fast disk subsystem apart from DB load balancing. If number of users increase then start thinking for database load balancing, by configuring Database partitioned views , Shared server or Replication.There isn't db load balancing in sql world 🙁
No, but you can fake it. IE: All read queries pull from replicated copies of the DB while you maintain the primary for DML only.
That said, however, 5000 users shouldn't really be a lot, but it depends on the use case of the database. I'm pretty sure the HP website (as a random example) handles a lot more than 5000 simultaneous users browsing their products.
If you're scaling the usage, you need to look at the current usage and determine if you have any existing bottlenecks or are close to hitting one. Then review the use case of the additional users and attempt to determine if they're going to affect those bottlenecks heavily.
This is one of those cases where I recommend you bring in an expert for a week or two to review your system and intended changes and get you a recommendation for a design path for your regular staff to review and implement. It's not a linear kind of thing, it all revolves around the nuances of the load increase.
EDIT: I just realized this was a 2 year old necro. Well, the OP won't care but anyone else who trips on it might.
Oh yeah makes sense now. We actually have a sort of similar scenario what you have, like all reads going to local-subscribers and writes to one-main- publisher...Works pretty good but we will eventually want to get rid of linked servers and may be use merge replication....ofcourse we still need to do lot of testing before we deploy in production.
Viewing 5 posts - 1 through 4 (of 4 total)
You must be logged in to reply to this topic. Login to reply