Satellite networks come in all manner of shapes and sizes and use a multitude of technologies to transmit information. Each type of network has its own idiosyncrasies and set of variables, and regardless of network type, satellite engineers strive to create optimal designs which effectively compete with wireless and terrestrial alternatives and provide reliability, affordability and provide an excellent user experience. As improvements in technology come along, engineers seek to optimize new and existing network designs.

Optimization involves weighing a number of variables and making careful choices in order to optimize the overall function to be improved. Engineers tasked with optimizing a satellite network must juggle multiple variables to get the best overall result. Many of the basic design considerations involve the RF link, antenna size, satellite frequencies and satellite modems, but as satellite networks increasingly are interconnected with IP-based networks, network optimization includes both wide area network concerns as well as RF considerations. What can be done to optimize a satellite network on the other side of the RF chain?

Compression and Multi-Destination Routing

“When you think about optimization, it is like a toolkit,” says Marc Nadon, president of LinkSat. “You rarely get huge savings in just one area, so you must make small, incremental gains on multiple levels. The individual effect on any one thing may be small, but the cumulative effect is large, which is the goal.” Earlier this year, LinkSat introduced its LinkIP router, which incorporates the company’s routed multi-destination network architecture and proprietary compression algorithms known as LinkShrink. The company believes in a layered approach to bandwidth efficiency, optimizing traffic several different ways.

“The inherent benefit of a satellite Earth station is its broadcast properties,” Nadon says. “We use this to our advantage by having each Earth station broadcast a single carrier that every other Earth station in the network hears. This creates a stat mux effect, resulting in savings of 10 percent to 30 percent. Normally, when dynamic routers receive packets that aren’t addressed for them they try to forward the packets. Obviously, in the architecture I just described, the routers would keep forwarding packets and create a broadcast storm. LinkSat has developed a highly efficient method of filtering packets in the LinkIP router that lets packets through to their intended destination and discards the rest. Since every node can communicate with all of the other nodes in the network, you have a very elegant mesh network which minimizes double hops and reduces latency. Each Earth station broadcasts just one carrier. This allows service providers to aggregate outbound traffic and eliminate guard bands between channels, which realizes yet more savings on space segment.

LinkShrink is a proprietary compression algorithm embedded into every LinkIP router. As packets enter the router they are checked for tags indicating source, destination, protocol, port or dozens of other variables. Certain types of traffic, such as MPEG video and JPEG images, are virtually impossible to compress, while other types of traffic can be optimized. VoIP traffic has traditionally fallen into the former camp, but LinkSat’s engineering team has found a way to further compact VOIP data even after it has been compressed. Working with a client in the South Pacific operating a SCPC network, LinkSat installed their LinkIP routers between the satellite modems and two different types of VoIP optimizers, one a media gateway and the other an A-Bis optimizer. LinkShrink was able to further compress the data without any loss of quality and managed to improve compression by 32 percent on the output of the A-Bis optimizer and 40 percent on the media gateway.

Adding LinkSat’s SCAN (Scanned Channel Assignment Networking) bandwidth management option to the LinkIP router provides clients a bandwidth-on-demand solution. SCAN includes a spectrum analyzer the company builds, enabling SCAN to make bandwidth acquisition decisions on the basis of monitored spectral activity. One possible application of SCAN allocates separate satellite channels for on-demand applications that require large amounts of bandwidth, such as video. When extra throughput is needed at a particular Earth station, its SCAN module grabs available bandwidth and then returns it to the pool after the transmission is finished. “When you add up all of the efficiency gains at the end of the day, it is quite impressive. LinkIP routers work with any brand of SCPC modems, and it gives carriers a cost-effective way to improve performance and reduce costs without having to replace entire racks of modems. The return on investment is typically about seven months.” 

Reducing Chattiness

TCP was developed with a blind eye to satellite networks, but the protocol has a few interesting quirks. Since it was developed to run on a wire, the protocol developers assumed that: one, there would be virtually zero latency between the time a packet was transmitted and the time it was received, and two any delay between sending and receiving was caused by congestion on the network.

When you think about optimization, it is like a toolkit. You rarely get huge savings in just one area, so you must make small, incremental gains on multiple levels. The individual effect on any one thing may be small, but the cumulative effect is large, which is the goal.
— Nadon, LinkSat.

TCP protocol requires that three acknowledgements be sent between the sender and receiver to determine the bandwidth that is available on a network before any payload data is sent. If this exchange occurs at the speed of light over an Ethernet cable, the acknowledgements do not add any perceptible delay, but if a satellite link is substituted for the cable, six satellite hops are required just to get the payload data flowing to its intended destination. If there is any idle time between transmissions, three more acknowledgements must be sent again. As noted earlier, the developers of TCP assumed that delay was entirely the fault of network congestion and integrated a throttling mechanism into the protocol to deal with this issue. Unfortunately, packet delays caused by satellite latency are interpreted as network congestion, and the protocol scales back the size of the packets being sent to deal with the “congestion problem.” Although there is plenty of bandwidth available to send large blocks of data, the protocol effectively limits how much data can be sent, which creates an expensive problem. Further, the developers of TCP assumed that since bandwidth was plentiful, if any packets were lost in transmission the sender would backtrack to the lost packet then resend everything from that point forward.

As the importance of the Internet mushroomed and Web browsing became inherent in daily business activities, Web browsing over broadband connections became commonplace. A good user experience is expected if satellite is to compete with terrestrial alternatives. Fortunately, there are a lot of clever software developers in the satellite industry that have devised ways to deal with TCP’s unique tendencies. Protocol spoofing was an early development which basically fools both ends of the TCP connection, eliminating the multiple hops required to get payload data flowing. TCP Fast Start is a widely adopted spoofing solution which allows Web pages to begin downloading faster.

Pre-fetch is another technique which enhances the end-user experience. A cache engine logs sites routinely visited by end users and then downloads new content during non-peak hours. The next time the end users visits one of their favorite sites, much, or all, of the content already is stored in the cache engine, and it is quickly displayed on the user’s screen. While caching speeds the loading of Web sites, it also reduces the amount of traffic on a satellite network during peak hours.

Global Protocols is a protocol development house which commercialized the Space Communications Protocol Standards (SCPS is pronounced “skips”). The software, called SkipWare, is a robust feature set and series of extensions to the standard Internet protocol. These extensions added much-needed functionality to the protocol, significantly improving its behavior over a satellite link. SkipWare not only is germane to satellite networks, it is used in other wireless applications that also face latency hurdles. In a satellite network, SkipWare resides in network devices at either end of a satellite circuit. The extensions to the protocol calculate the available bandwidth on a link and the round trip latency and SkipWare prevents the maximum transmission units (MTU) from being automatically scaled back in size. As discussed earlier, if a TCP packet is lost, the protocol goes back to the missing packet and begins resending all of the packets.

Global Protocols developed an acknowledgement scheme within SkipWare known as selective negative acknowledgement (SNACK). Rather than resending all of the packets, SNACK resends only the missing packet. In the past, network engineers often limited the size of the MTUs for fear of having to retransmit hundreds of large packets when an error occurred, thereby negatively impacting performance. Since SNACK sends only the missing packet, engineers are much more inclined to increase the size of MTUs, thereby improving link utilization. The net effect on performance is dramatic, with improvements in link utilization ranging from 25 percent to near 90 percent. The satellite industry has taken notice, and Global Protocols licenses SkipWare to many well-known hardware manufacturers for use in their products. Comtech EF Data has made SkipWare the standard acceleration technology powering their Performance Enhancing Proxies. The DataPath IP Accelerator is a SkipWare-enabled tactical accelerator, designed primarily for military field communications. 

Transport Streamlining

Riverbed Technologies is another company which incorporates SkipWare into its products. Riverbed, which manufactures the Steelhead line of WAN optimization products, uses three separate but concurrent techniques to optimize networks. The company’s sales are not limited to satellite networks, but it is in this market segment where they perhaps have the greatest impact. Riverbed’s appliances are installed at either end of a circuit and provide transport, data, and application streamlining. Transport streamlining harnesses the power of SkipWare and reduces chattiness on the network.

Data and application streamlining both reduce bandwidth crossing the optimized link which further increases efficiency. Inside the Riverbed appliances, a patented data algorithm analyzes strings of binary data and looks for repetitive patterns. Blocks of data sequences are characterized by a 16-byte data reference and stored on hard drives in each appliance. The data and its unique characterization are then stored at each end of the satellite circuit. When the original string of data needs to be sent again, only the data reference traverses the link, thereby significantly reducing network traffic. Riverbed can stack data references up to four levels deep and hundreds of megabits of data, even gigabits, can be characterized with a very short data string. Since only the data characterization crosses the satellite link, both network traffic and latency are significantly reduced.

TCP is known as a chatty protocol but chattiness is not limited to communication protocols. Many applications require a lot of back and forth communications between client and server, such as Microsoft Exchange, Maximo, Impact, Documentum, PeopleSoft and SharePoint, to name a few. Chattiness and satellite are not a match made in heaven, and Riverbed’s system solves this problem through application streamlining. The end result is better performance for customers and the delivery of certain applications which have never been able to run across a satellite link before. A vastly improved user experience often will eliminate the need for remote servers, allowing consolidation and fewer licenses.

Riverbed’s hardware is being adopted by a growing number of satellite providers, such as CapRock Communications. Rohit Chhabra, director of product management and marketing for CapRock, outlines some of the benefits associated with its WAN Optimization service. “Our customers can expect a fully managed solution which includes customized network design, equipment installation, management, maintenance and 24/7 monitoring. This provides increased simplicity and flexibility, allowing the customer to focus on their core business and not be burdened with the performance of their communications. Further, having the solution available as a managed service also allows it to easily scale as our customer’s business grows, minimizing the need to install and manage new equipment. Plus, it allows for a single point of contact for many key network services.”