By David Mackey
We at VistaOne like to educate customers on available solutions to ensure through collaboration, we employ the best solution for the problem and the environment. VistaOne engineer Jim Dixon recently described the way several traffic shapers prioritize network data.
Blue Coat PacketShaper
PacketShaper allows the management of aggregate class traffic through the use of “partitions” or virtual pipes. Traffic in a partition can have both a guaranteed minimum and a capped maximum. Individual flows or connections can also be controlled with the use of policies. It’s also possible to shape or control on a per host basis with the use of “Dynamic Sub-Partitions”. PacketShaper’s shaping mechanism is TCP Rate Control and works by manipulating TCP rather than directly dropping packets.
Exinda also allows for the management of aggregate traffic via guaranteed minimums and capped maximums. With Exinda, this can be done either with a virtual circuit, which is similar to the PacketShaper partition, or on separate classifications of traffic called policies within a virtual circuit. Exinda also does per host shaping they call Dynamic Virtual Circuits.
In addition, it’s possible do shaping or control based on total accumulated bandwidth per host. For example, I could set a bandwidth limit of a gig a day per host and drop the connection rate for hosts who exceed that daily limit. Both policies and virtual circuits can be scheduled and thus different policy can be in effect during business and after hours. Exinda uses its own implementation of Hierarchical Token Bucket Queuing for control.
Procera has a unique approach of allowing an individual connection to be classified more than once. This allows for per host, or even per connection policy to be in effect at the same time a global policy is in effect on the same type of traffic. So it’s possible to limit each P2P connection to 1k and also limit all P2P to 1M, etc. With Procera, traffic can be classified by application attribute (for example, HTTP URL or server host name) and even behavior (streaming, download, random looking/encrypted).
Procera can classify connections based on their client or server roles. It’s possible to limit (or block) local hosts when they are acting as servers but allow client connections on those same hosts for the same application.
Procera can also do volume-based shaping or bandwidth budgeting for a given time period and drop back connection rates when hosts exceed their budget. Policy scheduling is also very easy and can include events as well as regular schedules. Procera uses its own sophisticated multiple-queue “Leaky Bucket” queuing algorithm for control.