Riverbed has two types of technologies for WAN optimization:
The Steelhead technology, which performs the WAN optimization.
Note that the Steelhead Mobile, Cloud Steelhead, Steelhead Cloud Accelerator and Virtual Steelhead all have the same technology as the Steelhead appliance but are implemented in different scenarios:
Steelhead Mobile runs on desktop and laptop computers running the Microsoft Windows and Apple OS X operating system and optimizes traffic from the single computer it is running on.
Cloud Steelhead runs in cloud environments such as Amazon EC2 and optimizes traffic towards hosts in the cloud.
Steelhead Cloud Accelerator runs in the Akamai network and optimizes traffic towards Software-as-a-Service applications like Microsoft Office365 and Salesforce.
Virtual Steelhead runs as a virtual machine under VMware ESXi.
The Interceptor technology, which acts as a redirector of traffic towards Steelhead appliances.
Unlike other traffic redirection protocols like PBR or WCCP which are agnostic of the device it is forwarding to, the Interceptor only works with Steelhead appliances and takes advantage of the knowledge of the status reported by the Steelhead appliances.
There are two related appliances developed by Riverbed, which do not perform any WAN optimization but are used to manage and monitor the Steelhead appliances in the network:
The Steelhead Mobile Controller (SMC), a configuration and reporting appliance to manage Steelhead Mobile deployments.
The Central Management Console (CMC), a configuration and reporting appliance to manage configurations and monitor the behaviour of Steelhead appliances, Interceptor appliances and Steelhead Mobile Controllers.
From the outside, a Steelhead appliance has the following typical features:
A serial console port, to be used to do the initial configuration of the device and a way to access the device in case it isn't reachable anymore via the IP addresses configured on the primary interface. Its speed is 9600 bps, 8 data bits and one stop bit.
Two network interfaces, primary and auxiliary. These are the base interfaces used for management. The primary interface is normally the management port, which should be used to access the device via the network. The auxiliary interface is commonly used for data store synchronization but can also be used for management.
One or more by-pass cards, each with a pair of network interfaces labeled LAN and WAN. These are the interfaces which are connected to the WAN router and the LAN switch. The by-pass cards can be integrated on the chassis or installed via additional PCI cards.
Depending on the model, the following features may be available:
LED alarm lights on the front. They show the operational status of the device.
Hot-swappable hard disks on the front.
One or more hot-swappable power supplies.
On the inside there are several other important features:
The machine has two hardware watchdogs.
The hardware watchdog, which will reboot the appliance if a user-land process has not been communicated with the hardware watchdog for 30 seconds. This prevents the Steelhead appliance to hang indefinitely and gives it a chance to recover.
The network watchdog, which will put the by-pass card into fail-to-wire or fail-to-block mode if the optimization service has not communicated with the network watchdog for 7 seconds or more. This will remove the Steelhead appliance from the path in the network if there are problems with the optimization service.
If the machine is in normal operation, the LAN and WAN ports of the by-pass card are in operational mode and the optimization service is intercepting the packets. If the appliance gets powered off, the by-pass card either goes in fail-to-wire or fail-to-block mode. In the former mode, the LAN and WAN port are directly cross-connected with each other so that the LAN switch and WAN router are on Ethernet level directly connected to each other. In the latter mode, the LAN and WAN port will be disconnected from each other and no Ethernet link will be established.
As mentioned before, the primary and auxiliary interfaces are for the management of the port. Despite being two different interfaces, they cannot have an IP address on the same IP subnet as they share the same IP routing table.
Each physical LAN and WAN interface is combined into a virtual in-path interface in the configuration of the Steelhead appliance. The IP subnet on the in-path interface can overlap with the IP subnet of the primary and auxiliary interfaces and with the other in-path interfaces. Each in-path interface has its own routing table and must have an IP address and a default gateway configured before optimization is possible via this in-path interface.
Each Steelhead appliance has a data store to store its dictionary and every data store has a unique data store ID to identify the data store with other Steelhead appliances in the network. The exceptions for these are Steelhead appliances which are in data store synchronization cluster, where the data store ID of the slave is the same value as the master.
Some Steelhead appliances have only one hard disk, others have multiple hard disks in a RAID0, RAID1 or RAID10 setup so that for the operation system it looks like one giant hard disk. The operating system, the optimization service and the data store are all located on various partitions on this one giant hard disk.
Devices with a Fault Tolerant Segstore (FTS) have their operating system and optimization service separated from the data store: These devices have a normal RAID1 implementation for the operating system and the optimization service, while the data store is spread over the remaining disks. The contents of the data store are not protected for disk-failure but will be lost when one of the disks of the FTS fails. The advantage of the FTS is that there is no RAID overhead, the full capacity of the disks installed in the Steelhead appliance is used and there is no RAID rebuild overhead when a disk gets replaced.
The Steelhead appliance has flash memory storage where it boots from and where it keeps a copy of the configuration. It allows models without the RAID1 or RAID10 capability to rebuild themselves in the field if the hard disk gets replaced.
The Steelhead appliance has two partitions which it can boot into, one with the current version of RiOS and one with the version installed prior to the last upgrade. When a console is connected during boot-up of the appliance, you will have the option to select the partition to boot from.
When a serial console is attached to the serial port of the Steelhead appliance, you can follow the start-up sequence of the Steelhead appliance.
The Power On Self Test (POST) of the machine, during whic hthe BIOS is accessible and the RAID controller gets initialized.
The boot-loader, which selects the RiOS version to boot from.
The start-up of the Linux kernel and loading of various device drivers.
The mounting, and if needed the checking, of file systems.
If a software upgrade or downgrade needs to be done, this happens now.
Other upgrades such as a network card firmware and RAID Kit installation will also occur at this stage.
Start of the process pm. This process is the RiOS Process Manager which controls all the RiOS specific processes, such as the statistics process, the optimization service, the management process, and the SNMP service related processes to name a few.
From now on you will be able to login to the appliance via the network and via serial console.
Once the Process Manager is started, the configuration of the network interfaces will be applied, but the by-pass card will not go out of fail-to-wire or fail-to-block yet: This happens only when the optimization service has completed the initialization and has become operational.
Besides the optimization service, there are other processes to
support the operation of the Steelhead appliance. The following
lists contain the processes known on a Steelhead appliance running
RiOS version 7.0, but some processes are only run when certain
features are enabled. This list can be obtained with the commands
show pm process ?
on the CLI of the Steelhead appliance.
This is a list of the most important process on the Steelhead appliance:
acp, the Akamai tunnel process.
alarmd, the Alarm manager
cli, the CLI of the Steelhead appliance.
cmcfc, the CMC auto-registration.
httpd, the web server for the GUI.
mgmtd, the central management process.
pm, the Process Manager which keeps track of all Steelhead related processes.
qosd, the interface towards the QoS service.
rcud, related to CIFS pre-population.
rgp, CMC management related process.
rpgd, CMC management related process.
rspd, RSP watchdog
shark, for Cascade Pilot integration
sched, the job scheduler.
sport, the optimization service.
statsd, the statistics collection process.
virt_wrapperd, RSP related.
wdt, the user-land process to keep the watchdog happy.
webasd, related to the web server for the GUI.
winbind, for Active Directory integration.