March 19, 2014 at 7:51 am
Hi,
I have a data intensive project for which I wrote the code recently, the data and store procs live in a MS SQL 2008 ENT Ed Server. My initial estimate is that the db will grow to 50TB, then it will become fairly static in growth. The final application will perform lots of row level look ups and readings, with a smaller percentile of db write backs.
My piers are recommending that I abandon SQL Server for a NoSQL solution in order to handle the 50TB DB and its transactions. The questions I have are;
-Would SQL Server 2008 Ent. Ed. be able to handle a 50TB db?
-For physical storage my plan is to use either a SAN or DAS solution, are there any recommendations about either solution, that won't degrade performance?
-Should I consider a NoSQL solution or a Hybrid (SQL Server 2012 with the Hadoop plugin)?
Regards,
-r
March 19, 2014 at 10:01 am
raf.figueroa (3/19/2014)
Hi,I have a data intensive project for which I wrote the code recently, the data and store procs live in a MS SQL 2008 ENT Ed Server. My initial estimate is that the db will grow to 50TB, then it will become fairly static in growth. The final application will perform lots of row level look ups and readings, with a smaller percentile of db write backs.
My piers are recommending that I abandon SQL Server for a NoSQL solution in order to handle the 50TB DB and its transactions. The questions I have are;
-Would SQL Server 2008 Ent. Ed. be able to handle a 50TB db?
-For physical storage my plan is to use either a SAN or DAS solution, are there any recommendations about either solution, that won't degrade performance?
-Should I consider a NoSQL solution or a Hybrid (SQL Server 2012 with the Hadoop plugin)?
Regards,
-r
Yes SQL server can handle a database that large.
Does your data require referential integrity? If so, I would go SQL Server before thinking of using a NoSQL solution.
Should you consider a hybrid solution? Sure. It is a pretty common mix when dealing with very large data.
Jason...AKA CirqueDeSQLeil
_______________________________________________
I have given a name to my pain...MCM SQL Server, MVP
SQL RNNR
Posting Performance Based Questions - Gail Shaw[/url]
Learn Extended Events
March 19, 2014 at 11:27 am
Break it down this way, most of the NoSQL solutions are awesome at collecting data, not so good at reporting from it. So, if your main focus is collection, NoSQL is probably a viable alternative, especially during the gather data til we get to 50tb phase of the operation. But after that, as soon you want to run a query that cuts across the data in any fashion other than the id/value pair of HADOOP or one of the other NoSQL solutions, you'll be extremely stuck. Hybrid is probably the way to go, at least based on the little bit of information we have.
"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt
Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning
March 24, 2014 at 8:33 am
SQL 2008 Enterprise should handle that workload, but will you be able to manage that server?
I suggest you look at partitioning of tables, and how these would help query performance and server management. I also suggest you look at Windows 2012 R2 and SQL 2014, as these will run your queries faster than SQL 2008 can do, and have a number of useful additions that will make management of that volume of data easier.
50TB is a large amount of data. Unless you have previous experience of managing a database of this size I suggest you get a consultant who has. Implementing this in an inefficient manner will mean you spend thousands more on kit and get worse performance than if you implement it efficiently.
Original author: https://github.com/SQL-FineBuild/Common/wiki/ 1-click install and best practice configuration of SQL Server 2019, 2017 2016, 2014, 2012, 2008 R2, 2008 and 2005.
When I give food to the poor they call me a saint. When I ask why they are poor they call me a communist - Archbishop Hélder Câmara
Viewing 4 posts - 1 through 4 (of 4 total)
You must be logged in to reply to this topic. Login to reply