Replication and latency issues

  • Hi All,

    I am looking to put some servers into Americas, Europe and Asia to reduce latency issues when accessing a web application. Transactional replication would be used to keep the SQL2008R2/2012 (its not decided yet which to use) in synch between the locations.

    Given that they are so geographically separated I would like to understand if they are any pitfalls in this approach?

    Also, this is for an NGO, so they are invariably in the backwaters where the internet connection is likely to be out ran by a slug! Hence the "local" servers.

    Has anyone suggestions about the placement of the servers to reduce latency issues ?

    This is pretty much a blank canvass, so I don't have many restrictions, but as its for an NGO it cannot be the "bells and whistles" version.

    Any comments or advise would be really appreciated!

    Cheers

    Peter

    I work for 1st Consulting, a small ISV based in the UK. We build web and desktop applications using the MSFT stack. Find us at www.1stconsulting.biz

  • We use network compression hardware to speed up our replication to Australia. Without it would be shockingly slow. We still struggle occasionally with latency but then we do have a massive amount of transactions per second on some of our replicated tables (upwards of 10000 tps at times).

    I believe they are cisco ironport with data compression but I dont know for sure.

    Having the snapshot folder at the destination location reduces network load if you have multiple subscribers but this may not apply to you. This means snapshot changes only travel over the network once. Although it may be worth you considering a publication per subscriber so each location can have its snapshot locally.

    I wold also suggest increasing the packet size of the distribution agent and plpaynig around with the batch size to get the most out of the agent.

  • I am facing this weird problem:

    I have transactional replication with many publications setup and one particular publication is reporting distribution latency by perfmon counters. But when I look at replication monitor the latency is excellent and replication runs just fine without any issues.

    Has anyone seen this problem before with perfmon replication counters? I am wondering if there is any bug that I don't know.This is sql server 2005 instance with sp3 installed.

    Help will be appreciated.

    swimming goggle

    <a href="http://www.swim-goggles.info">swimming goggle</a>

  • We sometimes get that. Restarting the distribution agent resets the counter

  • Hi guys,

    thanks for the replies.

    It looks like we are going to go down the Azure route with mutiple apps deployed to different data centres. Given the cost of servers + licenses etc, going down the Azure route gives us a pretty redundant network.

    The icing on the cake is that MSFT look like they are opening up a DC in SA - so reducing the latency for the African area.

    Cheers

    Peter

    I work for 1st Consulting, a small ISV based in the UK. We build web and desktop applications using the MSFT stack. Find us at www.1stconsulting.biz

Viewing 5 posts - 1 through 4 (of 4 total)

You must be logged in to reply to this topic. Login to reply