1.2.2 Physical Structure Discussion

Loading...

Del curso dictado por University of Illinois at Urbana-Champaign

Cloud Networking

180 calificaciones

Explorar los videos relacionados

En Coursera, encontrarás las mejores conferencias del mundo. Aquí te mostramos algunas de nuestras recomendaciones personalizadas para ti

University of Illinois at Urbana-Champaign

Cloud Networking

180 calificaciones

Curso 5 de 6 en SpecializationCloud Computing

In the cloud networking course, we will see what the network needs to do to enable cloud computing. We will explore current practice by talking to leading industry experts, as well as looking into interesting new research that might shape the cloud network’s future. This course will allow us to explore in-depth the challenges for cloud networking—how do we build a network infrastructure that provides the agility to deploy virtual networks on a shared infrastructure, that enables both efficient transfer of big data and low latency communication, and that enables applications to be federated across countries and continents? Examining how these objectives are met will set the stage for the rest of the course. This course places an emphasis on both operations and design rationale—i.e., how things work and why they were designed this way. We're excited to start the course with you and take a look inside what has become the critical communications infrastructure for many applications today.

De la lección

Week 1

We will start by looking at the applications running in data centers and the traffic patterns resulting from them. This will help us differentiate the networking needs within data centers from computer networking in general and set the stage for other material in the course. We’ll then dive into the first of these topics, physical network structure, where we will look at how the increasing demand for bandwidth within data centers is influencing the network’s design.

1.1.1 Application and Traffic Patterns17:33

1.1.2 Application and Traffic Patterns Discussion2:11

1.2.1 Physical Structure12:14

1.2.2 Physical Structure Discussion3:13

Conoce a los instructores

P. Brighten Godfrey
Associate Professor
Department of Computer Science
Ankit Singla
Assistant Professor
Department of Computer Science, ETH Zürich

Okay Ankit. So, what did we learn in that lesson?

I guess, um, the first, you know, "big point"

is just that we're at very large scale,

and very high bandwidth requirements between posts.

Tens of thousands of them.

So we can't, uh, any longer use just a few.

Scale up, uh, switches forming the backbone of the network.

That would be very expensive, and even if, um,

we were willing to pay for it,

we just can't get the kind of port density

we need to network that big of a data center.

Absolutely.

As a result, data centers these days are commonly built

with clos or leaf spine topologies.

Where you use a large number of cheaper commodities,

such as, network together to build this large data center.

These networks can be, still arbitrarily large

and use very simple building blocks.

And as much bandwidth as we want, as large as we want.

It becomes feasible,

but it doesn't necessarily become cheap.

You're still paying for a lot of equipment, and

you still have challenges like cabling-huge numbers of cables.

If you really wanted, you know,

a fully nonblocking network

between tens of thousands of hosts, that is,

the host will never wait for the network.

That would be a lot of equipment and a lot of cables.

Absolutely. And cable aggregation, and

physical organization,these things go a certain distance,

uh, but people are evaluating other solutions

to this problem, right?

Yeah, so, you know, something I think is interesting is, ah,

you know, looking more towards some kind of radical, as yet,

unproven solutions in the research community.

Ah, some very cool stuff like using wireless,

as the links among racks in the data center.

Yeah, in fact, one proposal involves using wireless optics

to connect the host or top of rack switches that need

to communicate at the time

based on traffic you're monitoring in the data center.

Connecting them up dynamically, that these lasers

that are configurable at run time.

So, you'll establish a connection in response to

the traffic characteristics,

and you don't have to run cables between every pair of racks, or

even build fabrics as dense as these closed networks.

Cool, so obviously a lot of new challenges to solve there

from the physical layer on up,

and we're not sure whether, You know, something like that

will actually, ah, be cost effective in the future,

but it's a very cool direction.

Absolutely. And you know, another direction that, ah,

these hyper scale data centers might go,

is in trying to tackle that ah, tension between

getting the, ah, economy of scale of a huge data center.

Verses, being physically close to your user

so you get low latency.

Absolutely. In fact, I think it's very attractive to have

some compute-general purpose compute-closer to the edge,

where your customers are, right?

Latency sensitive apps will greatly benefit from that.

I will say, though, that will require

a lot of re-architecting for

both the network as well as the application

You have to change both.

Cool. It's an interesting space to watch.

All right, we'll see you next time.

Explora nuestro catálogo

Inscríbete de manera gratuita y obtén recomendaciones personalizadas, actualizaciones y ofertas.

Coursera brinda acceso universal a la mejor educación del mundo, al asociarse con las mejores universidades y organizaciones, para ofrecer cursos en línea.

© 2019 Coursera Inc. Todos los derechos reservados.

Descargar en la App Store

Consíguelo en Google Play