November 8, 2019 at 12:00 am
Comments posted to this topic are about the item The Network Bottleneck
November 8, 2019 at 11:32 am
When I eagerly walked into my current role nearly 2 years ago, I was staggered by the number of linked servers (often 5 or 6 instances of 3 different linked servers in one query).
The company had never had a DBA before and when they hit performance issues they just purchased another server and moved databases to another set of hardware... little realising that network I/O is your slowest resource.
I'm slowly dragging systems back from the dark ages and the number of network blips we have is much better. but I've had to patch a few things while we work to get it better - replication has been my friend, - slow gentle dribbles of data that I can use locally rather than query 100,000 records via a linked server... it's not a fix, but it keeps my query times down
Our network team very rarely let us down, it's our code architecture that puts pressure on our network (which we have to share with email, fielshare etc etc) - yesterday we had a 2 minute network outage... I lost all of my current connections to SQL and all of our linked servers failed - but some of my colleagues were more concerned that facebook was not available 🙁
MVDBA
November 8, 2019 at 12:26 pm
I agree with you. I'm working on a team rewriting legacy apps into newer technologies. However one thing we're instructed to do is bring everything back. Under curtain circumstances that can be a lot of data. What our boss is having is do is return only the 500 most recent records. Under certain circumstances that would be OK. Under other circumstances, such as when the user might want to do some aggregating of data, returning only the last 500 records is only going to yield the wrong results .
Rod
November 8, 2019 at 2:09 pm
We have a datacentre in London with many users in Glasgow. The Glasgow users frequently experience dire application performance while London users of the same applications have no problems. Our infrastructure team assure us that the network has ample capacity and no bottlenecks are detected. I can only assume that data slows down as it travels north (or uphill).
November 8, 2019 at 2:47 pm
We have a datacentre in London with many users in Glasgow. The Glasgow users frequently experience dire application performance while London users of the same applications have no problems. Our infrastructure team assure us that the network has ample capacity and no bottlenecks are detected. I can only assume that data slows down as it travels north (or uphill).
there was an article about this a fair few years ago - south Africa's network infrastructure was notoriously slow. they send a carrier pigeon with a 4Gb data stick attached - it got there faster than via their network.
i'm sure they did the same thing in australia as well... get some pigeons and your network will speed up in glasgow
MVDBA
Viewing 6 posts - 1 through 5 (of 5 total)
You must be logged in to reply to this topic. Login to reply