FOCUS on CONNECTIVITY
NETWORK ARCHITECTURES
Layers of complexity in data center networks
Demand for data center resources continues to grow, and while some issues such as power, cooling and security have dominated headlines of late, facility managers also have to adjust, expand, or redesign their legacy networking architectures to cope with the demand.
Today’s data centers can contain as many as tens of thousands of servers, none of which necessarily have the same bandwidth requirements. Therefore, a data center network architecture needs to be flexible enough to cope with these different needs and, as it expands, must be able to cope with demand for more capacity and higher speeds.
This is driving network managers to re-examine their architectures, both inside the data center and its connectivity outside the facility. And with technologies such as cloud computing and virtualization placing increasing connectivity strains on facilities, so the need to establish a flexible and capable network architecture that can handle these strains becomes even more important.
CHANGE DRIVERS The prospect of replacing or upgrading the entire network within your facility can often prove to be too disruptive, so a phased network architecture upgrade is often preferred, which will give more application flexibility and business agility.
One of the attractions with virtualization is that organisations can often reduce their data center costs by moving more and more applications onto fewer servers. This has the added benefit of reducing software licensing costs as well as administration costs. But management is demanding that applications
How many operators are ready to converge SAN and IP traffic on the same network?
and services must be deployed quicker than ever, while not compromising reliability.
This requires that the underlying data center networking architecture must be able to cope by providing suitable performance (ie, handling increased data traffic), availability and security, while also simplifying the day-to-day operations (network management) of the data center.
Going forward, therefore, there is little doubt that data centers will need network architectures that will support more virtualized environments and cloud computing models that can reduce application and service deployment times.
“Traditionally, data center networks were designed in three layers: the edge layer, which connects the servers; the distribution layer (ie, the first layer of aggregation of all the uplinks coming from edge switches); and the core layer (ie, the aggregation links from the distribution layer and where you do most of the server connections,” explains Olivier Vallois, EMEA data center and core category manager at HP Networking.
“If you have the network design of a tree within your data center, with the servers on the left and right, they need to go through the core in the center,”
Vallois says he feels that while most data centers in the past have a storage area network (SAN) in place, this is still separate from the local network.
CONVERGENCE “The model of a data center’s networking architecture used to be three layers,” Vallois says. “Nowadays, however, we collapse the distribution and core layer, and reduce it from three to two. We can do that because we now have very big core equipment and the technology to virtualize these products. So, for example, we have the capability to implement two chassis that operate as one chassis for the rest of the data center.”
“Certainly, the cloud and virtualization are two hot topics in the industry at the moment which are challenging network managers to look at their networking architectures,” says Alastair Waite, head of market management, EMEA, at ADC Krone. “This can be co-hosted
4
www.datacenterdynamics.com
image © Colt
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19