Caching/content distribution is completely different (although in some respects, the same) to just having lots of servers... a content distribution network gives it the geographical diversity so that the site is always performing well throughout the world in case the worst should happen.
Sorry, Cache and CDN is different thing.
That's where the load balancing comes in to play.
The main challenge seems to be less about serving up files but more about keeping the database nice and speedy. The reason FB/Google etc do quite well in delivering large amounts of data around the world is a combination of several factors, including sheer number of servers, the content distribution network (geographic diversity) as well as the distribution of physical data: the cluster of servers which houses the databases will have different specifications and optimizations to the servers which house the images, which will have different specifications and optimizations applied to the servers that house the webpages or system which generates the webpages and processes all the data - when you view a Facebook page you may be pulling all sorts of different chunks of data from 3, 4, 5, 10 or more physical servers (I don't actually know the real numbers in Facebook's case, I'm just going by what I do know from years gone by in dealing with much smaller scale setups), and it's those servers working together (in conjunction with the aforementioned factors) which keep the site running smoothly... unlike, for example, a small business which can only afford & deal with a much simpler web server setup which includes all of the above on a single machine.
If optimized properly, a single machine can probably handle a few thousand connections at a time, assuming it's got enough RAM, CPU etc to play with.
That is what I was thinking if hard drive and other systems have limitation then there would be difficult to fetch/obtain data with out getting server overloaded. They have to divide them in to numbers of systems.
I think we can build it at home in NAS. By having two three hard drive and observe the whole thing? ???
If I had to guess, I'd say the likes of Google and Facebook are hiring engineers who are fairly bright
They have to because according to System manager at Fb, Fb server management is a big headache.
Sorry, Cache and CDN is different thing.
That's where the load balancing comes in to play.
The main challenge seems to be less about serving up files but more about keeping the database nice and speedy. The reason FB/Google etc do quite well in delivering large amounts of data around the world is a combination of several factors, including sheer number of servers, the content distribution network (geographic diversity) as well as the distribution of physical data: the cluster of servers which houses the databases will have different specifications and optimizations to the servers which house the images, which will have different specifications and optimizations applied to the servers that house the webpages or system which generates the webpages and processes all the data - when you view a Facebook page you may be pulling all sorts of different chunks of data from 3, 4, 5, 10 or more physical servers (I don't actually know the real numbers in Facebook's case, I'm just going by what I do know from years gone by in dealing with much smaller scale setups), and it's those servers working together (in conjunction with the aforementioned factors) which keep the site running smoothly... unlike, for example, a small business which can only afford & deal with a much simpler web server setup which includes all of the above on a single machine.
If optimized properly, a single machine can probably handle a few thousand connections at a time, assuming it's got enough RAM, CPU etc to play with.
That is what I was thinking if hard drive and other systems have limitation then there would be difficult to fetch/obtain data with out getting server overloaded. They have to divide them in to numbers of systems.
I think we can build it at home in NAS. By having two three hard drive and observe the whole thing? ???
If I had to guess, I'd say the likes of Google and Facebook are hiring engineers who are fairly bright
They have to because according to System manager at Fb, Fb server management is a big headache.
