Very noob server question. Yes ive googled it.

[Drjack6]

So im just learning about servers. I now understand the basic purpose of one server. So, having one powerful server for a business makes sense to me. What i dont understand is this...

 

When i google pictures of servers, i dont see one. I see about 2,000 of them in stacks in a giant room. If each one is simply a computer with special software to manage other computers and tasks, how in the world are they working together? Id love to know. 

 

Thanks

[HunterSkater429]

They are just a bunch of hard drives that are storing data 

[Tedster]

3 minutes ago, Drjack6 said:

So im just learning about servers. I now understand the basic purpose of one server. So, having one powerful server for a business makes sense to me. What i dont understand is this...

 

When i google pictures of servers, i dont see one. I see about 2,000 of them in stacks in a giant room. If each one is simply a computer with special software to manage other computers and tasks, how in the world are they working together? Id love to know. 

 

Thanks

Because there are tasks that are way too big for singular servers to do, and people have written software that divides tasks among servers very, very effectively, similarly to how your processor likely has multiple cores, just on another level and massively scaled up.

 

All these different companies have their own different spins on different approaches and all that...

[Zando_]

I think Linus is coming out with a video soon (might be already out), where he looks around a massive server database or something. He's been teasing it on Insta and FB, so it should be out in the next few days. 

[Jarsky]

They aren't necessarily working together. You're talking about racks of them I take it. Theyre typically more for individual tasks (e.g storage, database, application, domain controllers, backups) or they might be hypervisor guests working in a pool (e.g vcenter) for managing a larger number of virtual servers. Especially if the stack is for Datacenter customers so they don't have to co-locate their own gear. Sometimes they may be clusters, the most common being failover. Parallel processing clusters are not really that common as it has to be written at the application layer and can be quite complex 

[Drjack6]

7 hours ago, Tedster said:

Because there are tasks that are way too big for singular servers to do, and people have written software that divides tasks among servers very, very effectively, similarly to how your processor likely has multiple cores, just on another level and massively scaled up.

 

All these different companies have their own different spins on different approaches and all that...

So are you saying that there is basically no giant group of servers that are running Windows server?

Are you also saying that every single company with lots of servers has custom written some sort of server control program?

 

Im so sorry, for some reason in my hours of research, Ive still come up very short of understanding how in the world servers work and what theyre used for. Im trying to piece all information together to understand it.

[leadeater]

32 minutes ago, Drjack6 said:

So are you saying that there is basically no giant group of servers that are running Windows server?

Are you also saying that every single company with lots of servers has custom written some sort of server control program?

 

Im so sorry, for some reason in my hours of research, Ive still come up very short of understanding how in the world servers work and what theyre used for. Im trying to piece all information together to understand it.

Large businesses require many servers, it's not about them all working as one but all collectively providing the resources to all everything that is required. For example we have around 1300 virtual machines (VM) which run on about 50 physical servers. Our VMs are mostly Windows and the physical servers are running VMware ESXi, VMware vSphere 6 with Operations Management to be more exact.

 

Those 1300 VMs can be active on any one server but only one, the VM also only has a share of the total resources on that server which you configure when creating the VM.

 

We also have physical servers directly running Windows that are using to host databases using Microsoft SQL Server and are clustered together using Microsoft Clustering Services, it is very important to note that a single application or instance running on a Microsoft cluster can only be active on one server. You can have multiple applications/instances running on a Microsoft cluster so you can spread around the active copy to different physical servers.

 

It's rare that multiple servers can work in parallel together on one task, as @Jarsky said, but it can be done and it's what most researchers use for massive simulations of weather or DNA etc.

[samiscool51]

34 minutes ago, leadeater said:

(bump)

thats the long exploitation of it

it's also for failover reasons for services that have to say up 24/7 (e.g youtube, google, LTT forum(not really))

say if one server has a problem ans it has to shutdown, another server takes over and continues the broken servers operation

i can't explain how it works as i don't understand it myself very well, but it works

also servers working together to process something are called server farms and are meant for HUGE work loads like leadeater explained, an example of this is for animation, south park studios (where the show is made) has a server farm to render the animations into a video format for ComedyCentral

if you want to learn more about it read this (it's an excellent read):

https://en.wikipedia.org/wiki/Server_farm
and these other resources are also useful:

http://www.webopedia.com/TERM/S/server_operating_system.html

http://whatis.techtarget.com/definition/enterprise-server

https://en.wikipedia.org/wiki/Server_(computing)

https://technet.microsoft.com/en-au/library/dd632514.aspx

[Jarsky]

5 minutes ago, samiscool51 said:

thats the long exploitation of it

it's also for failover reasons for services that have to say up 24/7 (e.g youtube, google, LTT forum(not really))

say if one server has a problem ans it has to shutdown, another server takes over and continues the broken servers operation

i can't explain how it works as i don't understand it myself very well, but it works

also servers working together to process something are called server farms and are meant for HUGE work loads like leadeater explained, an example of this is for animation, south park studios (where the show is made) has a server farm to render the animations into a video format for ComedyCentral

if you want to learn more about it read this (it's an excellent read):

https://en.wikipedia.org/wiki/Server_farm
and these other resources are also useful:

http://www.webopedia.com/TERM/S/server_operating_system.html

http://whatis.techtarget.com/definition/enterprise-server

https://en.wikipedia.org/wiki/Server_(computing)

https://technet.microsoft.com/en-au/library/dd632514.aspx

 

 

In failover clusters, the cluster maintain whats called quorum. The quorum configuration determines how many failures the cluster can sustain before quorum is lost and the cluster is shutdown for integrity. In larger arrays they can use an odd number of servers to sustain a loss of half the nodes (minus 1) to maintain a stable service (e.g in a 9 server cluster, 4 can go offline). With even node clusters the typically use whats called a witness disk - as the odd number tie-breaker in the vote.  (e.g with 8 servers, 4 can go offline if the witness disk is online - if the witness disk is offline then 3 servers can go offline). For what we do at our work, most of our failover clusters are 2 server clusters, so we use a fileshare witness (FSW) to act as the 3rd vote, which is hosted on a network fileshare.

 

afaik the term 'server farm' is a bit of a buzzword - the actual term is Parallel Processing Cluster - such as in a Beowulf, Matlab or Apache cluster. A server farm may not actually be a cluster, but a group of servers that work together in various roles to provide a service - they can be asynchronous.

[samiscool51]

23 minutes ago, Jarsky said:

(bump)

ok

1. i said i had a understanding of how they worked, but not very well, and if i wanted to lean how they worked i would of googled it a long time ago (thanks for the info though, very interesting read, helped me understand it more)

2. i used server farm because thats what most people call them and people will understand what i'm talking about if i say server farm, most won't get what i'm saying if i say "Parallel Processing Cluster" as many server noobs won't understand what i'm saying, yes it's not the correct term but if the individual wants to learn more they would learn the correct terms while reading about it...

 

[leadeater]

1 hour ago, samiscool51 said:

it's also for failover reasons for services that have to say up 24/7 (e.g youtube, google, LTT forum(not really))

Services like youtube don't use failover configurations at all. They use scale out load balancing across servers so if one goes down it doesn't matter as any of the other servers can service requests.

 

There's many different ways to do things and which is possible or the best solution depends on the task that is required. Websites are best to load balance across many servers where as VMs will failover to another server if the host it's on fails. Microsoft SQL only supports one active writing server in a cluster but many in read-only if using SQL Always On, if not then it's a true Active/Passive cluster.

[Drjack6]

Welp. After reading all of that..... I give up.. 

[samiscool51]

41 minutes ago, Drjack6 said:

Welp. After reading all of that..... I give up.. 

how do you think i feel? thats everyday with me and @leadeater, him telling me something, and me going:

me: i know that, i dumb it down for the people who want a simple and straight answer....

leadeater: but you get some info wrong...

me: again, simple and straight, some info may have to be wrong to allow a person to understand it better

leadeater: again, some of the info is wrong

me: f**k this... (walks out of room)

[leadeater]

16 minutes ago, samiscool51 said:

how do you think i feel? thats everyday with me and @leadeater, him telling me something, and me going:

me: i know that, i dumb it down for the people who want a simple and straight answer....

leadeater: but you get some info wrong...

me: again, simple and straight, some info may have to be wrong to allow a person to understand it better

leadeater: again, some of the info is wrong

me: f**k this... (walks out of room)

Pictured above: @samiscool51 

 

On another note, the detail is important which is why I re-clarify it. Using the correct terminology is important since if someone wants to look for further information on it they'll be looking at the right things. Sure I do often over explain things though so yea... .