Redundancy For Servers and Services

Video Activity
Join over 3 million cybersecurity professionals advancing their career
Sign up with
Required fields are marked with an *
or

Already have an account? Sign In »

Time
15 hours 43 minutes
Difficulty
Advanced
CEU/CPE
16
Video Transcription
00:00
>> Now, a little bit earlier,
00:00
I talked about redundant servers,
00:00
and we said they are
00:00
multiple servers acting independently.
00:00
I got server A and Server B.
00:00
We're going to talk about that,
00:00
and then we're going to expand that idea
00:00
>> to talk about clustering as well.
00:00
>> Redundant servers, we'll talk about
00:00
load balancing, and clustering.
00:00
This is an illustration that just shows two servers.
00:00
They're each separate entities.
00:00
If server 1 fails, that's okay
00:00
>> because server 2 is there to pick up the work effort.
00:00
>> It's not unusual also though,
00:00
to have load balancing configured here.
00:00
This illustration, looks like everything
00:00
goes to the first server until it fails,
00:00
and then those requests will be sent on to server 2.
00:00
But in many environments we're going to have
00:00
either a hardware device
00:00
>> that balances the load between them.
00:00
>> Or we may have some other configuration
00:00
like round robin,
00:00
and that's a configuration that basically says,
00:00
traffic goes to one, then two,
00:00
>> then one, then two,
00:00
>> then one, then two,
00:00
regardless of hardware performance.
00:00
That's not the best load balancing,
00:00
but it certainly is adequate.
00:00
With the redundant servers,
00:00
you have two independent devices,
00:00
and that's the key element.
00:00
Now, with a cluster,
00:00
you have multiple nodes in the cluster
00:00
>> that share a single identity,
00:00
>> and that's the important piece, they share an identity,
00:00
>> where redundant servers are independent devices.
00:00
>> For instance, when you go to Amazon.com,
00:00
how many servers could be responding
00:00
>> to that Amazon.com request?
00:00
>> Probably a zillion.
00:00
>> There's so many servers
00:00
that respond to the request from Amazon,
00:00
but I don't have to type
00:00
out server1.amazon.com or server2.amazon.com.
00:00
I just go to Amazon,
00:00
and there is a cluster of servers
00:00
>> that all respond to the www piece.
00:00
>> They all function under that request,
00:00
and they're all running the web service.
00:00
We get redundancy not just of servers
00:00
>> but of services as well.
00:00
>> But to a user, it looks like just a single server.
00:00
Most people don't even think the number of servers
00:00
>> that are there capable of responding to www.amazon.com.
00:00
>> I always think of a cluster like a cult.
00:00
All the nodes lose their individual identities.
00:00
You no longer have server 1
00:00
>> that's accessible or server 2,
00:00
>> they are all in a cluster called www.
00:00
Any one of them respond,
00:00
but they lose their individual identity.
00:00
Web farms, server farms,
00:00
they're talking about clusters in those instances.
00:00
Now, the primary purpose of clustering
00:00
>> is usually associated to be redundancy.
00:00
>> Redundant drives, redundant services,
00:00
>> just like we said.
00:00
>> But usually clusters provide load balancing as well,
00:00
distributing the loads across the systems.
00:00
However, clusters don't have to load balance.
00:00
Most of them do, but they don't have to.
00:00
Sometimes you can set up
00:00
>> what's called an active passive cluster.
00:00
>> Let's say I have a small company,
00:00
and we don't have a lot of money.
00:00
I spend a good amount of money
00:00
>> to have a good, quick, fast domain controller.
00:00
>> But if that domain controller fails,
00:00
I don't want the domain to shut down,
00:00
so I have a cheaper, less capable system there
00:00
>> in the event of failure.
00:00
>> I want my functions,
00:00
I want my request to go to server A,
00:00
and server B just sits there waiting for a time
00:00
>> when server A isn't available.
00:00
>> Nodes in a cluster send
00:00
>> what's referred to as a heartbeat out.
00:00
>> That heartbeat says, "Still alive,
00:00
>> still alive, still alive."
00:00
>> But if server B misses that notification,
00:00
it knows to kick in and start handling requests.
00:00
The only reason I take a few minutes
00:00
there to point that out
00:00
>> is you could see a question that essentially says,
00:00
>> you want to ensure no one server is overwhelmed.
00:00
What should you enable?
00:00
One answer is clustering,
00:00
the other is load balancing.
00:00
Load balancing is a different function than clustering.
00:00
Like I said though, many clusters use load balancing
00:00
>> but they don't have to.
00:00
>> In this section, we talked about the difference
00:00
between having multiple servers, redundant servers,
00:00
and having a cluster,
00:00
and then we talked about
00:00
>> distributing the workload across those servers
00:00
>> in a means that we refer to as load balancing.
Up Next