Data Traffic is about to Explode: Don’t Get Caught in the Blast
Using what you have in a smarter way
If you like theory, then you’ll be interested to know that many are predicting that data center traffic is set to sharply increase. As cloud computing centralizes, more computing resources and more devices, such as mobile phones, tablets and TVs, are being used to exchange data.
However, if it is statistics that you prefer, then ABI Research predicting that global annual data traffic will exceed 60,000 petabytes in 2016 — over seven times more than the 8,000 petabytes expected in 2011 — will be of more interest. Just for good measure, it is also predicting that the fastest year-on-year growth will occur in 2012, at 58 percent.
Whichever way you look at it, more traffic means you’ll need more server power in your data centers. And you’ll need to find a way to meet this increased demand, as well as space to accommodate it all. While a few years ago money would have provided the answer, the fact that belts have been tightened so you can barely breathe means you need to find a different way. The time has come to use what you’ve got in a smarter way.
The starting point
As more computing resources are centralized, monitoring, analyzing and securing these resources become more important than ever. Yet, network appliances are typically single server implementations, with few providing more than one application. It is not uncommon for several network appliances to be accessing a single monitoring location. For example, a typical scenario could be three appliances monitoring the same connection with one monitoring specific flows, another providing performance analysis and a third providing intrusion detection functionality.
Since cost, space and power are major issues for data centers, reducing the footprint of network appliances also becomes a major consideration. Many network appliances require all the processing power they can get and, thus, cannot share processing resources with other applications. Examples are 10 GbE Intrusion Prevention Systems or Application Performance Monitoring systems.
But, there are also a large number of monitoring, analysis and security appliances that run at lower speeds or do not require as much processing resources. Here, there are opportunities to consolidate these appliances into a single server solution.
If all appliances are based on the same operating system, it is possible to consolidate them using intelligent network adapters that can distribute data and share data between multiple applications. Such solutions exist today.
Virtualization provides some cover
If the appliances are based on different operating systems or environments, or expect to have full control over available hardware resources, then an alternative solution is required. Virtualization can be used in such instances to consolidate these very different applications.
It will depend on your data sharing and distribution needs as to which solution provides the best possible outcome. The following describes various options based on VMware that can be used to consolidate multiple network appliances onto a single physical platform.
Using VMware Direct Path
 |
| Figure 1 |
VMware Direct Path allows a virtual machine to control a physical network adapter. This allows existing network appliance applications to be transferred to a virtual environment, as shown in Figure 1.
This is the first step in consolidation. To the network appliance application software, it still appears as if it is running on its own server with full control of the intelligent network adapter. The driver software has been updated to support VMware Direct Path but, otherwise, no changes need to be made.
| Figure 2 |
With this solution, a consolidation can be performed for multiple network appliances (Figure 2).
As this model shows, each network appliance can be based on a different operating system and execution environment, yet still be supported on the same physical server. The only restriction is that each virtual machine needs its own network adapter. This is because only one virtual machine can control a given network adapter at one time.
Sharing network adapters
While the above implementation works, it still requires a dedicated network adapter for each virtual client. This limits the number of applications to the number of slots in the server. If all the virtual clients need to access the same point in the network, a separate load balancer would be required to distribute the data between the network adapters.
 |
| Figure 3 |
By distributing data within VMware, we can eliminate the load balancer and reduce the number of network adapters required (Figure 3).
By using a data distribution virtual machine as a server virtual machine based on VMware’s Virtual Machine Communication Interface (VMCI), it is possible to distribute and replicate data to multiple virtual machine clients. The data distribution virtual machine can, thus, distribute or replicate data captured by a single intelligent network adapter to multiple client virtual machines each supporting a separate network appliance.
Distributing data on a per physical or virtual port basis
 |
| Figure 4 |
One method of distributing data to multiple client virtual machines is by physical port, as shown in Figure 4.
In this example, data on each port of the network adapter is mapped to a separate client virtual machine. However, this limits the solution by the number of physical ports on the network adapter.
A more interesting solution is to use logical ports (Figure 5).
| Figure 5 |
Some intelligent network adapters are capable of identifying flows and, thus, defining logical ports providing specific flow data. These logical ports can be mapped to VMCI ports allowing specific data to be distributed to dedicated network appliances running on client virtual machines. The number of virtual ports that can be supported is limited by the implementation on the network adapter, but can be up to 32.
Sharing data between multiple virtual machine clients
 |
| Figure 6 |
As mentioned earlier, it is not uncommon for multiple network appliances to need to access the same data at the same point in the network at the same time. The captured data needs to be shared and replicated to multiple network appliances (Figure 6).
The data distribution virtual machine can be used to replicate the data captured by the intelligent network adapter to each virtual machine that requires that data. The only limitation is the bandwidth of the VMCI interface itself, which is dependent on the processing power of the supported CPU chipset.
Performance expectations
Implementations of the solutions described above have been made providing a benchmark for expected performance. Napatech has successfully demonstrated that the VMCI interface can support up to 30 Gbps of data replication and distribution to multiple virtual machines. This allows any combination of port speeds and number of virtual clients to be implemented, as long as the total consumed VMCI bandwidth does not exceed 30 Gbps.
Benefits of network appliance Virtualization
As stated previously, not all network appliances can be virtualized, especially high-speed, high-performance appliances that require all the processing resources available. However, for less processing-intensive appliance applications, there is an opportunity for consolidation that is compelling.
One of the advantages of using virtualization for consolidation is that each network appliance can be re-used to a large extent with the same operating system and environment. This also means that it is possible to upgrade the physical hardware without needing to upgrade the supported network appliance virtual machines. As physical servers continue to increase in power and performance, even more appliances can be consolidated onto a single physical server.
As network interface speeds change, it is possible to upgrade the intelligent network adapter to support a higher speed interface without having to change the support network appliance virtual machines. This possibility also can be used to upgrade existing network appliances to support higher speed interfaces in a fast and effective way.
For example, a 10 Gbps network appliance can be upgraded to support 40 Gbps by porting four instances of the network appliance software to four virtual machines running on a single server supported by a single 40 Gbps intelligent network adapter. Four logical ports are created to distribute the data between the four virtual machines, making sure that none of the virtual machines receive more than the expected 10 Gbps of data. Thus, a 10 Gbps network appliance becomes a 40 Gbps network appliance without having to re-haul the network appliance application software.
This approach also can be used to upgrade older network appliances supporting legacy operating systems or where resources to update the network appliance application software no longer are available.
Double protection
Virtualization has and continues to transform the economics of running data centers but it isn’t enough. Today’s cost, space and power demands require that network appliances are as effectively and efficiently utilized as their application server counterparts.
While many high-speed, high-performance network appliances already make optimal use of the server resources available to them, there are a number of opportunities for network appliance consolidation that can be exploited, especially as we move to higher network speeds and ever more powerful physical servers.
Don’t drown in the data explosion. Instead, get ready to ride the data wave.
Dan Joe Barry is VP of marketing at Napatech. He may be reached at [email protected].
