Menu
Community /Pin to ProfileBookmark
@NogDogMay 10.2017 — #Typically there is a "load-balancer" (L? server between the web browser and the actual web servers (WS). The domain name service will direct the browser request to the LB, which will use its algorithms to decide which WS will receive the request, and then sends it to that WS. The WS does its stuff, then outputs its response to the WS, which then does more magic ( ? ) to send it back to the requesting browser. Things can get further complicated if/when you also need to split up the database server into more than one, in which case there may be another LB just for the DB requests form the web servers.
Basic diagram:http://www.serverlab.ca/wp-content/uploads/2014/04/51-figure-02c.png
With DB balancing, too:https://itprimer.files.wordpress.com/2011/08/haarchi.jpg @NogDogMay 11.2017 — #Each web server can talk to the same DB server over the LAN if you want. Conversely, if the database is the choke point, one web sever might talk to multiple (possibly sharded) DB severs. The potential different configurations are many, and no single one is right for all situations -- plus you get into redundancies and fail-over servers in case any one (or more?) servers fail, and so forth. That second image link diagrams two load-balanced web servers sharing their requests across 4 separately load-balanced DB servers, for instance. @NogDogMay 11.2017 — #Facebook probably has multiple data centers across the country/world, each with dozens or even hundreds of physical servers, each physical server potentially running many virtual servers. By the time you are getting enough traffic to even have to think about more than a couple web servers and database servers, you'll be able to afford to hire an expert to expand your production deployment as needed. ?
(The specific system I work on right now gets by on two web servers and one DB server, processing over a 1000 requests per minute during peak hours.)@NogDogMay 11.2017 — #
No idea -- but I'm sure it changes every few months. ?
We're mostly migrating to Amazon AWS services now. Here's some descriptions of their options, though they don't necessarily list physical box manufacturers/models:https://aws.amazon.com/ec2/instance-types/ @NogDogMay 12.2017 — #It's a little like GoDaddy and other such services, but gives the user more flexibility on how each virtual server is configured, along with the ability to spin up new (virtual) servers as you need them. Instead of having to fill out a form and pay to get the additional servers you need, then wait for them to be set up; you instead have a billable account that lets you essentially bring them on-line as and when you need them, configure the way you want -- and in a matter of minutes it's ready for use. Of course, it all then gets automatically added to your billing account.
Microsoft provides a similar service, and I think maybe Google does, too?
multiple server question
so I’m kinda new to web development and something has been on my mind a while now. when building a website that requires multiple servers (just because of site demand and size) how is it managed I guess? I mean, if for talks sake one server can handle 10,000 users at any given time (that number is a complete guess) and the site is averaging 20,000 users at any given time, the obvious solution is to add another server. so how does the clients web browser know which server to go to? and what if the stuff their looking for is on one server but they get redirected to the other server because the first one is busy, what happens then?
Sign in
to post a comment17 Comments(s) ↴
0
@NogDogMay 10.2017 — #Typically there is a "load-balancer" (L? server between the web browser and the actual web servers (WS). The domain name service will direct the browser request to the LB, which will use its algorithms to decide which WS will receive the request, and then sends it to that WS. The WS does its stuff, then outputs its response to the WS, which then does more magic ( ? ) to send it back to the requesting browser. Things can get further complicated if/when you also need to split up the database server into more than one, in which case there may be another LB just for the DB requests form the web servers.Basic diagram:
With DB balancing, too:
0
@NogDogMay 11.2017 — #0
@NogDogMay 11.2017 — #Facebook probably has multiple data centers across the country/world, each with dozens or even hundreds of physical servers, each physical server potentially running many virtual servers. By the time you are getting enough traffic to even have to think about more than a couple web servers and database servers, you'll be able to afford to hire an expert to expand your production deployment as needed. ?(The specific system I work on right now gets by on two web servers and one DB server, processing over a 1000 requests per minute during peak hours.)
0
@NogDogMay 11.2017 — #one other thing I'm curious about is what brand machines are out there, who are "apparently" the best manufacturers and how many requests can the average server actually handle?[/QUOTE]
No idea -- but I'm sure it changes every few months. ?
We're mostly migrating to Amazon AWS services now. Here's some descriptions of their options, though they don't necessarily list physical box manufacturers/models:
0
@NogDogMay 12.2017 — #Microsoft provides a similar service, and I think maybe Google does, too?