ORIGINAL: RCadmin
ORIGINAL: doxilia
Marc,
I see. I would have imagined that a site such as RCU would have had multiple servers in the US located at different ''network geographical points'' in order to balance traffic from different points on the globe. But maybe it's easier to have a single point of presence network wise.
I appreciate your looking into this as it seems that users from around the world are having issues. Don is a few hundred miles south from the servers. I'm several thousand. People in Europe may be routing along the Eastern seabord much like I do and head west along the similar paths to mine.
At the moment I can't upload anything to RCU. I also can't navigate the site with Firefox. Chrome is working on the other hand.
David.
David,
With a bad traceroute like that I think this is network route related. It was passed off to the engineers to look at. Those ping times will cause browsing and upload issues for sure.
As far as multi-geographical servers this was never done because of database consistency. If we had an east and west coast server farm we would have to have a master db somewhere. That would mean the other location would have latency from using a far away db server as its source or alternately an outdated slave server. With forums where you refresh and must see your post instantly this is not a good option.
What is distributed are the images and other static site assets. These make their way to you through closer servers through the CDN (content distribution network). These will usually only be a few hops away with low latency.
What a mess....
I do this stuff for a living for a HUGE bank... we don't allow any impact to our customers for images and data, that my main job as a Systems Architect. Admitedly there are the occastional project leaders who f- things up, but the outcome is usually seen before it hits the customer.
First off not knowing what your using for a back-end, but I believe Oracle, SQL Server, and the maybe still free MySQL (which one of the three and Oracle I would go with) all provide for instant log shipping. Which I would not reccomend since the SQL round trips for selects when the DB is optimized even somewhat are nil. Having multiple web servers running asp code with distributed images is a disaster wating to happen when you have only 1 site on the west coast. My opinion a waste of effort. You'd either stack 2 sites redundant on each coast with BigIP in case 1 site goes down and fail over or just suck it up with 1 provider in CA. Using L3 with your hop count is a big indication that they don't know what they are doing and bouncing you around like a ping pong ball. Sure you'd be going a few hops router to router but the latency posted is what is causing your main issues.
From just reading this much... I say things look like it's better to hit this site wearing long boots to wade through the mess. A project like moving a site takes a long time, and effort, that I understand, but the outcome usually is a weekend down, you bite the bullet, and then come up fresh without issues. Plus it ALWAYS proven out before hand before the swap is made. While it's appreciated that the site exists... mucking up a server swap, distributed images, multiple web servers, one backend (gawd only knows and prays it itsn't some free crud). I would have been easier to have 2 back ends sync'd (Oracle), dual image filers, and a big single web box on each coast load balanced by your ISP. cheaper, faster and much better than serving up the world from the west coast. And get off this cruddy L3 backbone. Worst providers in the business.
Thank (fill in your diety here), we think things though before it hits the users. Thats what beta sites are for kids.