The Fort Lauderdale, Florida-based company recently rounded out its portfolio of app delivery products with the $52m purchase of Orbital Data Corp, a privately-held developer of technology for optimizing large file transfers over any protocol on a WAN link, said Greg Smith, director of product marketing for Citrix.
Technology for branch offices
The Orbital technology was developed originally for branch office environments and is symmetrical, i.e. it requires devices both at head office (i.e. at the head end of the WAN link) and out in the branch. Its initial claim to fame, and what it made its name on, was its approach to TCP flow control, which differs from competitors like Peribit (now part of Juniper) and Swan Labs (now F5).
TCP is a notoriously inefficient protocol for WAN links, so optimization vendors develop ways of stripping out some of the inefficiency, for instance in the infamous three-way handshake that TCP comms normally require between sender and receiver.
However, most of the others do it by hard-forming tunnels between the two boxes and changing packet headers, which results in a loss of transparency, such that no router, firewall, monitoring or QoS device has visibility into the traffic, said Smith. We, on the other hand, do it transparently to the network, with no changes to headers.
CIFS and multi-level compression
Above and beyond its TCP flow control prowess, San Mateo, California-based Orbital had been filling out its offering in the months prior to acquisition, adding first the ability to optimize CIFS, the remote file access protocol for Windows environments, and, most recently, multi-level TCP compression.
The former took Orbital squarely into competition with wide-area file systems (WAFS) vendors such as Actona (now part of Cisco), Tacit (acquired earlier this year by Packeteer), DiskSites (which was bought by Expand) and, the last remaining independent, Riverbed. Citrix is known to have considered the acquisition of Riverbed, but opted instead for Orbital, and while the technology was doubtless one factor in play, there may also have been a price issue: Riverbed is also said to have rebuffed various approaches from Cisco with a price too high for the prospective buyer.
Multi-level compression is, in essence, the ability to detect repeated patterns and suppress them, after their first occurrence, with a token, such that they can be re-introduced at the receiving end. Orbital had already been able to do that in a short history fashion, i.e. intra-session compression with the patterns held in dynamic memory, but the multi-level functionality meant it also began to do the same for long history scenarios, namely from one session to another, which of course meant the introduction of disk drives on the boxes for longer-term storage of the patterns.
Laptop client
The most recent development at Orbital, however, was of the client, which takes the technology beyond the branch office and into remote access, be it on a fixed connection (e.g. for the home worker on a DSL link) or an unwired one (e.g. a laptop with WiFi in a hotel).
The entire Orbital product set is to be renamed as the WANScaler portfolio, associating with the Layer-7 app acceleration Citrix acquired last year when it bought NetScaler, whose products are eponymous, and the laptop client, which was provisionally known as Orbital Edge, is now to be called the WANScaler Client. It recently went into beta, and, said Smith, is equally applicable on a wireless link as a wired one, since the data will still be going over TCP/IP.
Indeed, since data sessions over cellular rather than WiFi connections, since they also traverse TCP/IP links, would also be suitable cases for treatment with the technology, opening up the perspective of optimization of traffic to connected PDAs and smart phones. Indeed, the main obstacle to extension of the client to such environments would appear to be support for mobile device OSes such as Windows Mobile 5.0 and Symbian, and Smith commented that Citrix has broad support for other platforms, and such a move is part of the conversations we’re holding about product evolution.
Other developments
Beyond that, he cited further improvements to Citrix thin client optimization as a result of the Orbital acquisition and, down the road, improvements to the kind of QoS WANScaler can offer, for both Citrix and non-Citrix environments.
On the Citrix thin client front specifically, Smith said Orbital had already been optimizing Presentation Server comms before it was bought, and in fact one of the reasons we chose them was because they performed better than other competitors in Citrix environments. He explained, however, that the proprietary protocol that underpins the Presentation Server, called ICA, has virtual channels between client and server for different functions, so our deeper understanding of the protocol means even better optimization.
As for QoS, the Orbital technology right now has the ability to control bandwidths available to individual apps based on their class of service (CoS) setting, but we’re not yet doing shaping, which can be done by queuing or by window sizing, he explained, adding that the dedicated QoS vendors out there use one of both of these techniques.
These dedicated vendors are the likes of Packeteer, Allot and Citara (now Converged Access), who were really the first generation of WAN optimization developers. To be fair, they have all sought to increase their offering into flow control and compression, and indeed the vendors who came into the market from the compression side of the business, such as Juniper/Peribit and Expand, have beefed up their QoS capabilities.
Smith said Citrix expects to have made some decisions as to how it will do traffic shaping, as well as the other platforms it will want to support with the WANScaler Client, within the next couple of months.
