With globalization, organizations are now not confined to a single place but
spread across large geographical areas. This means they have a sizeable
workforce that's always on the move. As these employees need continuous and fast
connectivity to critical applications like ERP, products such as WAN
accelerators are the need of the hour. We received a number of WAN accelerators
from Riverbed, which we deployed under various scenarios to see how they can
enhance enterprise WAN connectivity.
Steelhead appliances for WAN acceleration are powered by a software platform
named RiOS 5.5 (Riverbed Optimization System). At the core of RiOS lie four
types of optimization techniques. Here's a brief explanation of each:
- Data streaming is achieved by reducing transmission of redundant bytes and
depending on organizational needs, prioritizing data. This leads to a
considerable reduction of bandwidth requirements (up to 95%). - Transport streaming reduces TCP traffic by reducing the number of TCP
packets required to transfer data thus optimizing bandwidth usage. - Application Streaming is specific to an organization's applications, and
one can optimize a number of protocols like CIFS, NFS, HTTP, HTTPS, etc. - Finally, Management Streaming helps if organization is using multiple
Steelhead appliances, as management of these devices can be done centrally.
This reduces overhead and saves man-hours.
Some applications accelerated by Steelhead appliances include Microsoft's
Office, Lotus Notes and Oracle 11i. RiOS 5.5 also supports virtualization,
giving a fascinating dimension to Steelhead appliances. It means now a single
Steelhead appliance can run five virtual machines (VMware-based). So, an
organization can virtually run firewalls, mail server, etc.
Test setup
We used two standard machines acting as client and server with a WAN emulator
between them. We emulated the WAN using 'Network Nightmare' appliance, which can
simulate the WAN and one can easily configure bandwidth and latency on this
device. We simulated the WAN by setting the bandwidth to 128 Kbps with a latency
of 100 milliseconds. We calculated the time taken to transfer files from the
client to server. Here's the list of file transfer scenarios carried out on each
set of WAN accelerators:
- Copying a 10 MB Highly Compressible Text File.
- Copying a Highly Compressed (with bzip2 -9) 11MB file.
- Copying a 10 MB Highly Compressible Text File WITHOUT WAN accelerator.
- Copying a Highly Compressed (with bzip2 -9) 11MB file WITHOUT accelerator.
To check QoS we prioritized HTTP trafic on the WAN accelerators while
simulaneously transferring HTTP and CIFS data. This was accomplished by running
Apache server on one end of test setup. Finally, to test caching and referencing
we retransmitted files and calculated the time of retransmission and compared
this time with the actual transmutation time.
This diagram shows the test setup for testing WAN accelerators from different vendors. We used Network Nightmare for simulating WAN conditions. |
Steelhead Appliance 1020
This appliance is meant for a mid-size organization. We received two units and
configured them according to our test setup (explained above). The setup was
quick and easy. Just attach keyboard and monitor to the appliance and it will
boot into Steelheads OS where one can configure it using commands similar to
Cisco router commands. We gave an IP address to this device and then configured
it using its elaborate and easy to use web interface. Reporting done by this
appliance is very impressive.
We started by transferring a 10 MB compressible text file from one machine to
the other, without Steelhead appliances, and it took 720 seconds. While a highly
compressed 11 MB file took 780 seconds. When we used Steelhead appliances on
both ends with RiOS version 5.5, the compressible text file took only 11 seconds
and a highly compressed file took 741 seconds.
After getting these results, we retransmitted both these files and it further
reduced the transfer time to 3 seconds for a 10 MB text file and 6 seconds for a
highly compressed 11 MB file. Finally, to check QoS we simultaneously
transferred a highly compressed 11 MB file via CIFS and HTTP prioritizing CIFS;
HTTP transfer took 22 seconds less than CIFS transfer.
This fig illustrates the test setup for Steelhead Mobile controller. SMC lies beneath the network firewall controlling mobile devices with Steelhead mobile software. |
Steelhead Mobile 2.0 with Controller
This product is specifically meant for the mobile workforce of an organization
and needs a central controller (Steelhead Mobile Controller). Once an
organization purchases licenses for mobile workstations, using SMC,
administrators can install Mobile software on remote devices. Steelhead Mobile
Software interacts with Steelhead appliance sitting behind SMC in same fashion
as two Steelhead appliances interact with each other. We received a unit of SMC
and found that configuring it was simple. Simply, connect a key board and
monitor to the SMC and boot into Steelhead Controller.
Graphical representation of the results obtained from symmetrical setup of Steelhead appliance 1020. |
It will ask for some basic settings including IP address. Once the IP address
has been set, connect SMC to a network and open configuration pages using
browser. From this interface, one can configure controller and mobile devices
attached to the controller.
Once the controller is up and running, the next setup is configuration of
policies for mobile devices. There are two kinds of policies for endpoints:
'Endpoint policies' and 'Acceleration policies.' Along with these policies, one
needs to create and assign Deployment IDs (DID) that link endpoints to both
these policies.
Here, the Endpoint policies are like templates for configuring a group of
mobile clients and it determines computer specific software setting for mobile
clients, i.e. data store size and Mobile controller device it is connected to.
On similar lines, acceleration policies act as a template for a group of mobile
clients with similar performance requirements. This policy includes protocol
setting, setting optimization rules, etc. Both these policies are necessary for
optimization of traffic. In our setup we used default Endpoint and Acceleration
policy named 'Initial' with default settings. Now how do you install these
policies along with required software on mobile devices? To accomplish this
task, we used default MSI package that contains the 'Initial' policy. To locate
this file from SMC's home page, go to 'Manage Endpoints' tab and click on
'Packages.' Here one can find Default MSI package. Install this package on
mobile clients. Finally, one has to buy licenses for their mobile clients. SMC
assigns these licenses to mobile clients and the software on a mobile client
acts as the Steelhead appliance.
Graphical representation of the results obtained from Steelhead mobile 2.0. Performance is similar to Steelhead 1020. |
We used similar tests on SMC and mobile software. We transferred a
compressible 10 MB text file and it took 720 seconds for this file to move from
one system to the other. The highly compressed 11 MB file took 780 seconds. When
we used Steelhead appliances and Steelhead mobile devices with RiOS version 5.5,
the compressible text file took only 12 seconds and the highly compressed file
took 741 seconds.
The interface of mobile software after we installed Default MSI package from SMC. The optimization status should be 'Healthy' and the controller connection status should be 'Connected:Licensed.' |
After getting these results we retransmitted both files and found that the
transfer time was further reduced to 3 seconds for the 10 MB text file and 8
seconds for the highly compressed file. QoS tests gave similar results to those
with two appliances. These results are almost similar to the test setup where we
used two Steelhead appliances, which means the performance or optimization
remains the same with mobile software version 2.0.