The Riverbed Blog (testing)

A blog in search of a tagline

Riverbed’s optimizations for Citrix ICA and other real-time traffic types

Posted by riverbedtest on September 9, 2010

There has been some confusion lately regarding Riverbed's capabilities to optimize thin-client and other real-time traffic types, so in this blog I'd like to summarize all of the different ways that Riverbed enhances the performance of Citrix ICA, RDP, VoIP, and other similar traffic types.

Numerous Riverbed customers are using Riverbed to deliver significant bandwidth reduction and performance improvements for Citrix ICA and RDP traffic.  Typical data reduction results range from 20% to 60%, while response time improvements vary from 10% to 40%.  Actual results depend on various environmental factors; better data reduction results are generally achieved when a greater number of thin clients access data from a given remote site.  Response time improvements are generally more noticeable when there is pre-existing network congestion and performance issues prior to enabling the Riverbed optimizations.

Riverbed data deduplication capabilities are particularly effective when applied to thin client and other traffic types characterized by small packet sizes because of the granularity of the Steelhead algorithms.  On average, Riverbed is able to identify byte-level commonalities of only 100 bytes.  In contrast, many of Riverbed's competitors use algorithms that can only identify byte-level commonalities of 4KB to 32KB–a huge difference.  In addition, Riverbed Steelheads are able to implement their algorithms entirely in memory when applied to real-time traffic types such as RDP and Citrix ICA, in order to minimize processing latency and avoid adding jitter and latency to the real-time traffic.  Most of Riverbed's competitors lack this capability.

For Citrix ICA traffic in particular, Riverbed's optimizations are specialized and powerful.  The Riverbed Steelheads are actually able to communicate with the Citrix XenApp server's API's and disable its default compression on a per-flow basis.  The Steelheads can then apply their algorithms on the original uncompressed Citrix ICA data, allowing the Citrix connection to receive the full benefit of Riverbed's byte-level data reduction technology.  Generally, most competitive vendors require that compression settings be globally disabled on the Citrix server.  This has the drawback that all Citrix users–those local in the LAN as well as those at remote sites who do not have WAN optimization devices–receive uncompressed Citrix ICA traffic, which makes those users more vulnerable to bandwidth constraints and network congestion issues.  With Riverbed, only those flows being optimized by the Steelhead appliances will have their default compression disabled.

Riverbed Steelheads also have the ability to decode the ICA Priority Packet Tagging that identifies the virtual channel from which each Citrix ICA packet originated.  As part of this capability, Riverbed specifically developed a packet-order queuing discipline that respects the ordering of ICA packets within a flow, even when different packets from a given flow are classified by Citrix into different ICA virtual channels.  This allows the Steelhead to deliver very granular Quality of Service (QoS) enforcement based on the virtual channel in which the ICA data is transmitted.  Most importantly, this feature prevents any possibility of out-of-order packet delivery as a result of Riverbed's QoS enforcement; out-of-order packet delivery would cause significant degradation in performance and responsiveness for the Citrix ICA user.  Riverbed's packet-order queuing capability is patent-pending, and not available from any other WAN optimization vendor.

With regard to optimizing VoIP traffic, here again Riverbed has specialized capabilities.  Since the VoIP data is already efficiently compressed by the voice codec, it therefore is not possible to achieve further compression of the VoIP data (you should be wary of any vendor that tells you otherwise).  However, it is possible to apply very granular QoS enforcement for VoIP traffic to ensure the voice call remains unaffected by network congestion that may occur in the WAN.  Toward this end, Riverbed's Hierarchical Fair Service Curve (HFSC) scheduling mechanisms in its QoS enforcement capabilities can guarantee high-priority forwarding service for any type of traffic, including traffic types such as VoIP that consume a small amount of bandwidth. Traditional systems that use Packet Fair Queuing (PFQ) mechanisms, including those used by Riverbed's competitors, don't work as well in these settings.  The problem with PFQ-based systems is that low-latency forwarding service is only provided to traffic classes that have been allocated large amounts of bandwidth.  As a result, PFQ-based systems do not perform as well when handling high-priority traffic that consume small amounts of bandwidth, including traffic types such as VoIP, Citrix ICA, RDP, telnet, etc.

Here is a link to a real Riverbed customer who achieved 61% data reduction for Citrix ICA, saving 14GB of Citrix data per month that didn't have to be sent over the WAN. 

2 Responses to “Riverbed’s optimizations for Citrix ICA and other real-time traffic types”

  1. ben said

    Hello, useful track, the link does not work, can you update it ? thanks

  2. Josh Tseng said

    Strange…the link seems to work fine for me. Here’s a URL to the page:
    http://community.riverbed.com/t5/Performance-Hall-of-Fame-2010/14Gb-of-CITRIX-Traffic-Saved-Per-Month/td-p/3766
    Here’s another link to a different Riverbed customer using Steelheads to optimize Citrix ICA:
    http://community.riverbed.com/t5/Performance-Hall-of-Fame-2010/500GB-of-iSCSI-replication-removed-PER-WEEK/td-p/3784

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

 
%d bloggers like this: