On The First Class of Service the Network Gave to Me…

By: John Shepler

One of the big impediments in getting networks to work well for all types of traffic is that different types of traffic have different needs. Treat all packets the same and you’ll enjoy excellent results in some cases and unacceptable results in others. What’s the solution? You need to add a touch of class to your network.

The idea behind class of service or CoS is that different types of traffic get different priorities on the network. For instance, do you really think that data backups are as sensitive to network characteristics such as latency and jitter as VoIP telephone conservations? Of course not. If your network has the capacity it may not matter. With infinite bandwidth all traffic will flow smoothly from source to destination. The problem is that we don’t have infinite bandwidth or infinitely large anything else.

This really shows up in MAN and WAN connections. These telecom links are almost always more limited than your LAN resources simply because of the cost involved. How many companies can afford 10 or 100 Gbps private lines cross-country? In fact, even 1 Gbps is a stretch for many IT departments. Many companies will have to get everything done with 10 Mbps Ethernet or 100 Mbps Fast Ethernet MAN and WAN service.

Can this work? It likely can if you carefully manage your traffic and don’t have so much that you saturate the network. The trick is in establishing multiple classes of service and assigning each type of traffic to one of those classes. Do it right and everything on the network will run smoothly.

The simplest example is to establish two classes of service. One is for latency sensitive applications. The other is for everything else. A good example is a SIP trunk that provides you with both VoIP business phone service and Internet access. This is becoming a popular setup for smaller business locations because of the cost savings involved in having only one outside trunk line. The VoIP traffic is extremely sensitive to latency and jitter so it goes into the class that has priority. Most of what you do on the Internet isn’t that sensitive, so it goes into the non-priority class.

You need this type of division because voice traffic is typically low bandwidth per conversation and will break up if there is any network congestion. Data transfers and web page downloads aren’t so sensitive to congestion and will take all the bandwidth they can get. If there isn't enough, they may slow down a bit but you won’t lose any data.

Class of Service keeps the big data and video downloads from pushing the more sensitive voice and two-way video conferences to the side. This can make your network perform much better overall without increasing bandwidth. Even so, there’s no free lunch. When the network gets congested enough, you are going to notice the slowdown on non-critical traffic and will have to increase bandwidth to make everything run as fast as it can.

Class of Service is one of the big differentiators between the Internet and private networks. There is no CoS on the Internet. In fact, it is a matter of pride that all packets are treated equally. This principle is called network neutrality. It makes the public Internet fair for all users, but inhibits performance on latency and bandwidth sensitive applications.

Is there a reason to create more than two classes of service? There is for more sophisticated operations. XO Communications is offering what they call Premium CoS with six different classes.

At the lowest level is standard class service for data and Internet traffic. The next step up is Priority class for data and Internet traffic that has a more urgent need to get through quickly, such as email messaging. Up from that is Priority Plus for traffic that has some sensitivity to delay and jitter such as FTP transfers.

Critical, the next class up, is for business applications such as SAP that will affect productivity if they are delayed. Critical Plus goes a step higher for applications such as Telnet Routing Protocols. The real time applications are at the top of the heap for priority. These include voice and two-way video. One way video can take a lower class because it is easily buffered to compensate for most jitter and latency issues.

Are you frustrated by the problems that have arisen as you’ve expanded the expectations of what types of traffic your network is expected to handle, especially for MAN and WAN connections? If so, a private network with two or more classes of service may get everything running smoothly again at a reasonable cost. Get more information about MAN and WAN class of service network solutions, along with pricing and availability.


Note: Photo of high speed traffic courtesy of Wikimedia Commons.

New Data Center Networking Solutions

Data management and storage requirements for enterprise level organizations continue to increase. In addition, a majority of enterprise IT departments now replicate data between data centers. This has created a need for new and more cost effective bandwidth services, especially at 10 Gbps and above. Level 3 is addressing these needs with an enhancement of its existing data center networking solutions.

What’s different about data center connectivity compared to other business requirements? Data centers are far more concentrated than other computing resources. This is especially true for storage, a prime driver for offsite data centers in the first place. What this does is increase the need for high bandwidth connections between the concentration of storage and the concentration of users. It also influences the connectivity required.

What you need for last mile connections to an MPLS network or private lines to link business locations can usually be handled by T-Carrier, SONET/SDH or Carrier Ethernet line services. Data centers need higher bandwidth connections, but may also need special SAN protocols including Fibre Channel, Infiniband, ESCON and FICON.

SAN stands for Storage Area Networks. These are large collections of disk drives that connect directly to the application servers. SAN networks have traditionally been very localized, with the disk arrays physically near the servers. But what do you do when you need to duplicate data between data centers?

Many companies operate two or more data centers, sometimes a half-dozen or more. Some of this is due to mergers and acquisitions of formerly independent organizations with their own IT infrastructures. Much is driven by a need for disaster protection and recovery. When your business is critically dependent on its electronic data, normal backup processes aren’t adequate. You can have multiple copies of your files locked up locally and still be put out of business by a fire, tornado, flood, earthquake, hurricane or other disaster that destroys whole buildings and everything in them.

This means that you want to have copies of your critical data dispersed geographically. Establishing independent data centers in two or more cities far apart means than a disaster that takes out one center probably won’t knock out everything. There’s also an advantage to content providers in having the servers and data content located close to the customers. Latency is reduced and performance is increased due to less network congestion. A challenge is how to make sure the data is replicated exactly at all locations.

This is where long distance SAN networking comes into play. You need high levels of bandwidth but also support for SAN protocols to keep everything synchronized. FCoE or Fiber Channel over Ethernet is a popular protocol for transmitting Fibre Channel frames over 10 Gbps Ethernet connections. WDM or Wavelength Division Multiplexing avoids having to layer protocols to connect facilities. Each wavelength in a fiber optic link can be assigned its own protocol, regardless of what is running on other wavelengths in the same beam.

Level 3 Communications has expanded its data center networking portfolio to include SAN fiber channel protocol, dark fiber and managed fiber solutions. This is in addition to dedicated bandwidth services from 1 Gbps to more than 40 Gbps, including 10 Gbps EVPL or Ethernet Virtual Private Line service. They offer low latency route guarantees and an international service footprint, including a presence at key public exchange facilities.

Are you concerned with managing multiple high performance data centers? Would you like to have more options and/or better pricing for all your telecom connectivity needs? If so, get complementary consulting help and pricing and availability for high bandwidth networking services.



Note: Photo of data center servers courtesy of WikimediaCommons