Choice between cluster, log shipping and repl.
-
We have to install an SQL2K Ent Ed. environment (2 servers + External SCSI RAID box)
Which would be your recommandation of choice between cluster, log shipping and replication. (i.e. what is the main diff between those technics)
The issue is performance and availability
Thanks
Bye
Gabor -
We've got a few posts here on the site about this issue. All have their merits, comes down to your specific needs. I use replication because I also use it for reporting. Probably later this year we'll cluster the publisher for fault tolerance, leave the replication in place for reporting. I'd rate log shipping as easiest to use/maintain, followed by replication, then clustering. For max availability you have to go with clustering, then use either log shipping or replication for DR scenarios - server room goes up in smoke.
Andy
-
For availablility clustering is really your only solution unless you app is coded to check multiple servers for the same data. I have always found clustering to be easier to setup than replication and generally more stable. Log shipping is easy to setup and maintain as well. I have to say without knowing what you are shooting for besides performance and availablility it makes it hard to recommend a solution and point you in the right direction for more information.
Wes
-
In fact the performance is the real issue.
We will have 3-4 applications connecting to several DB's. Approx. 1500 user, most of them are concurrent
I was thinking to make 2 servers (if cluster then active/active or log ship or replicate - and here is the question
) to balance the load between the applications.From the availability point of view We can support 10-15 min of downtime but as few as possible of data loss.
Of cours in case of a DR situation the users can accept a slower response time until the productive box is repared.
Bye
Gabor -
If the money is available you could do worse then take a look at Marathon Endurance Server. Once setup its just like running SQL on one machine but with assured availability and increased performance.
Nigel Moore
====================== -
Based on your equipment, performance needs, and allowable downtime, your best best bet would be clustering. That doesn't come easy from me, I like log shipping. But If I was in your shoes, clustering is the way to go.
One quick thought, watch your system utilization on a dai;y/weekly basis. The minute you start sustaining a combine server cpu utilization of over 100% will be the day you need to reevaluate your situation, ie new hardware, or apps and databases that can be kept offline until the down server is rebuilt.
Here's a discussion we had on the topic
John Zacharkan
John Zacharkan -
For availability, has anyone looked at the hardware based redundant windows servers offered by Stratus? (http://www.stratus.com/)
No one recommends using software based RAID 5, and for this reason, it makes sense that for availability, hardware based redundancy might be a better choice.
For warm backups, since we did not install the enterprise version, I created my own version of logshipping. I call it "Logshipper lite"
I have found that replication gets a bit picky with Identity columns, so I have avoided replication for availability.
Andrew
What's the business problem you're trying to solve?
) to balance the load between the applications.Viewing 7 posts - 1 through 6 (of 6 total)
You must be logged in to reply to this topic. Login to reply