Squeezing 10 pounds of data into an existing 5-pound connection can save a lot of money. Network data compression is becoming so commonplace that router vendors have begun including limited compression features - such as TCP/IP header compression - in most models. However, for maximum compression performance that can be tailored to fit your specific traffic patterns, you should consider dedicated network compression devices.
We tested five systems - from BoostWorks, Expand Networks, ITWorx, Packeteer and Peribit Networks - to see how they compressed data through a limited network connection and what the deployment and management costs amount to.
All products go well beyond shrinking packet headers by also significantly compressing the packet payload. In our FTP testing, we achieved as much as a 400 percent performance gain, which represents a best-case compression scenario.
The performance gains from these products heavily depend on the traffic types and flows running across a network. Your mileage definitely will vary.
Within the broad category of "network compression," the vendors found many ways to squeeze data into a standard pipe. The major differences lie in how compression is defined and what traffic is compressed.
The algorithms these products use depend, broadly, on one of two techniques: redundant string compaction and replacement dictionary lookup. Redundant string compaction replaces strings of repeated characters, or strings with regular patterns, with smaller replacements and reconstruction instructions. Dictionary lookup also targets repeating strings and patterns, but instead of inserting compacted replacement strings, lookup keys pointing to dictionary entries are inserted into the traffic stream.
Compression algorithms have a significant effect on performance, but larger differences center on whether compression is applied to all traffic or only traffic from a particular source, and how much of the overall network traffic stream already is compressed.
All products except BoostWorks' attempt to compress all network traffic, unless a setup parameter has exempted a particular type of traffic or traffic from a particular address.
The BoostWorks product attempts to compress only particular types of traffic, such as HTTP, Simple Mail Transfer Protocol (SMTP) or specific application transactions. Its designers are banking on the assumption that most enterprise traffic falls into one of the targeted patterns.
It also makes a difference on complexity whether compression is unidirectional (from server to client, for example, where a network compression device acts as a server and compresses specific types of traffic, and decompression occurs on the client) or bidirectional (end-to-end, where a compression device sits on both ends of the network segment to handle compression). All products in this review support bidirectional compression. BoostWorks also supports unidirectional compression.
Most devices pay close attention to the TCP port from which the traffic originates, although Packeteer also considers information from the packet header.
Most products - except those from BoostWorks and ITWorx, which don't compress User Datagram Protocol (UDP) traffic - could compress voice-over-IP (VoIP) traffic we threw at them by 20 percent to 25 percent. In our FTP tests, a 146M-byte uncompressed copy of the Linux kernel took 39 minutes to transfer. When we used compression devices, transfer times shrank to between less than 10 minutes up to just over 19 minutes.
If data passing across the network is repetitive, caching effectively can reduce the amount of traffic passing through the network - Expand and Peribit support caching. When we threw a pre-compressed 2M-byte file at Peribit's box, the first and second times yielded a transfer time of 34 seconds. The third time around, the same file took two seconds. Expand's box exhibited similar behavior. This sort of caching/reduction would benefit an enterprise link, where copies of the same file are being sent around through e-mail or FTP transfers.
Packet aggregation effectively joins small LAN packets into jumbo packets to reduce header overhead costs. This can be important on satellite links, where network latency on one side of the connection is a factor. While packet aggregation had a significant effect in our UDP performance tests (changing packet aggregation timing values increased performance greatly), network managers should consider whether to apply packet aggregation to UDP traffic because the process can have an adverse effect on latency. Most of the products tested support packet aggregation. Packeteer and Expand let you configure their devices to support this feature while ITWorx and Peribit support it transparently to the user.
Where the other systems in our tests are designed to compress traffic moving between segments of enterprise networks, making the most of large leased data connections, BoostWorks also can work on a more common data connection - between the server and remote client. Rather than creating tunnels between appliances, BoostWorks' BoostEdge system can be deployed using one appliance together with a software client on the remote system. In this configuration, the BoostEdge is a unidirectional system, compressing FTP, HTTP, SMTP and Secure Sockets Layer (SSL) traffic sent from the server. The system also can be deployed without a software client, but in this mode only HTTP traffic is compressed.
By focusing on specific traffic types, BoostWorks achieved solid results, posting the second highest results we saw in transferring a large, uncompressed file via FTP. Getting to these results was simple with the BoostWorks system - taking the system out of the box to have it yield functional acceleration took only a matter of a minutes. The client loaded easily onto a laptop system and operated essentially invisibly. The FTP test was the only test we could complete because of the nature of the system.
Depending on the model, the BoostEdge can support up to 80,000 established connections and an input bandwidth of 42M bit/sec.
BoostWorks has a Web-based management interface that provides statistics on traffic and compression, and a mechanism for device setup, although BoostWorks' system requires less in the way of configuration than systems that depend on dedicated tunnels between endpoint devices.
BoostWorks doesn't support quality-of-service functions, although QoS is promised in future releases after the company's acquisition of StreamCore.
If a company is concerned about maximizing productivity of remote workers and reducing total bandwidth from, for example, an intranet Web server, BoostWorks is the only product we tested that takes network acceleration on the road. There are more complete systems for accelerating links between major network segments, but in its niche, BoostWorks stands alone.
Product name: BoostWorks BoostEdge BE500A
Pros: Supports uni- and bidirectional modes.
Cons: Application-specific compression, user needs to be running client-side software to take advantage of non-HTTP compression.
Expand Accelerator 4000 and 4800
Expand's Accelerator 4000 and 4800 models are similar products for different architectures. The Accelerator 4000 is a WAN-based product that connects directly to a router's serial ports. The Accelerator 4800 is a LAN-based product that connects to the network through a 10/100 Base-T Ethernet interface. Both provide high-bandwidth dataflow (up to 6M bit/sec in the 4800 and up to 2M bit/sec in the 4000 model) and strong performance, but with a cost to be paid in terms of setup and administration complexity.
Expand refers to its performance boost as "acceleration" rather than simple compression. The results obtained are based on traffic stream compression, header compression, data caching and error recovery. Getting to the good acceleration results took effort, in two areas in particular.
Twice during our testing, lengthy problem-solving sessions boiled down to routing issues through the boxes. In these cases, we saw compression results that didn't make sense based on the traffic used for the test and the design of the network. In both cases, traffic flowing in one direction was being routed around, rather than through, Expand's device. Because of the devices' implementation, setting up Expand's systems requires the administrator to be very explicit about the accelerator's place in the network and its relationship to key routers.
While Expand's systems produced good results when deployed with default compression, performance improved considerably after experimentation with various error-correction features. For packet aggregation to have a significant effect, we had to adjust timing values.
The default performance of the unit was quite good, especially in testing with VoIP traffic. Typical compression of VoIP traffic using devices in this review resulted in a 20 percent to 25 percent reduction rate - Expand's WAN accelerators compressed that traffic by 5 percent to 10 percent more. The TCP and UDP traffic tests showed good levels of compression. Expand also did well with the QoS test, performing on par with the other systems tested.
Expand's Accelerators provide encryption and compression to traffic, act as authentication servers, and support numerous router- and application-specific protocols. They do not, however, support the basic security functionality of Secure Shell (SSH), a significant lack in a system that provides other security features.
The administration command-line interface (CLI) will be familiar to anyone who knows Cisco's IOS, and the Web-based GUI provides plenty of performance graphs, charts and tables, although the GUI was neither as polished nor as easy to navigate as those found on other systems tested.
Product: Accelerator 4000 and 4800 4810
Cost: Range of bandwidth models for both with 512KB starting at $9450 (for both)
Pros: LAN and WAN products available; strong compression performance, although tweaking required for best performance.
Cons: No SSH access; somewhat complex setup and deployment.
ITWorx's NetCelera is a LAN-based network compression product. It includes two 10/100M bit/sec ports - a local one connected to the switch and a second, labeled remote, that is connected to the router. NetCelera scales to speeds of up to 45M bit/sec. NetCelera boasts a quick installation process and made a very strong performance showing in our tests, particularly with TCP/IP traffic.
Full-blown configuration and ongoing management are accomplished using a Web-based GUI, and a CLI is also available. Both methods require secure connections. Access lists might be employed to prevent unauthorized connections to the device, but we did not test this feature. Logging is accomplished via SNMP and syslog support.
ITWorx says NetCelera works at the session layer of the Open Systems Interconnection model - this device inspects flows at Layer 5 (before they are merged at Layer 3 into one IP stream). The other devices we tested can do what is termed packet aggregation: Rather than compress and pass each packet along with an individual IP header, these devices can wait a number of milliseconds, compress each packet and then pass an aggregated packet with one header, thereby saving space. Because NetCelera "understands" TCP, there is no need for it to implement this mechanism.
But we have to note that NetCelera does not support compression of UDP traffic and that simply passes it through the device unchanged. Also, pass-through (no compression) of UDP traffic meant that the VoIP test performed on other devices was not conducted with this device because the G.711-based VoIP calls are UDP-based.
We were disappointed that the system did not include a central management system or QoS features.
ITWorx said it will add features that address these concerns within the next six months. The company says its central management system, which will include reporting of top talkers, response times and realized savings, and will let a network manager prioritize traffic by type and IP address.
Pros: LED front panel makes configuration a snap; very strong TCP performance.
Cons: UDP passthrough, no central management software.
Packeteer PacketShaper Xpress 2500 and 6500
Packeteer's PacketShaper is a LAN-based QoS product, but with the purchase of Xpress software, compression capabilities are added to the device. The PacketShaper 2500 and 6500 each include two 10/100M bit/sec ports, one labeled "Inside," the other labeled "Outside." The inside port is connected to the switch, while the outside port is connected to the router. The 2500 supports configurations ranging from 2M to 10M bit/sec, while the 6500 scales to 100M bit/sec. If you're looking for the best mix of compression and QoS features, this is your box.
The PacketShaper Xpress software is the only product tested that lets the operator choose which compression algorithm to use. By default, the system uses Predictive 2, a two-pass predictive algorithm. It compresses only traffic deemed capable of incurring a significant compression ratio.
Ongoing configuration and management is performed through a GUI, but a CLI is also available. In our testing, PacketShaper had good performance with TCP and UDP traffic, and VoIP.
Because the PacketShaper automatically selects the best TCP parameters, we did not have to spend a significant amount of time tweaking TCP parameters to obtain optimal results. While the focus of this review is not on QoS, what really separates Packeteer from the rest of the pack is the maturity of its QoS features - they had no problem protecting the VoIP call by placing it into a prioritized queue.
The only problem we ran into during testing was the same problem we saw during our QoS testing with the device last year (DocFinder: 6931) - it might take a few minutes for low flow-count traffic to be recognized by the PacketShaper's classification engine.
Packeteer also has introduced the AppCelera ICX, a unidirectional (server to client) device that can work separately or together with PacketShaper, and provides Web-based compression and SSL offloading, complete with detailed historical reporting. We did not test this box, but mention it because it is comparable to the BoostWorks option.
Product: PacketShaper Xpress 2500 and 6500
Cost: US$2,500 to US$39,000, depending on link speed; shaping license US$750 to US$8,000 depending on link speed; AppCelera functionality starts at US$8,000. Pricing varies in AUD$. Check with distributor.
Pros: Very mature QoS functionality; good compression performance.
Cons: Very low flow-count traffic might not be classified.
Peribit's SR-50 is a LAN-based network compression product. It includes two 10/100M bit/sec ports. The local port is connected to the LAN switch, and the remote port is connected to the WAN router. The SR-50 is capable of handling up to 45M bit/sec of traffic. Installation can be completed in less than 10 minutes, and performance in all areas is very strong. Peribit says it has applied DNA pattern matching technology to its algorithm, which is known as molecular sequence reduction.
Ongoing configuration and management are accomplished using a Web-based GUI, although a CLI is also available. Both methods require secure connections. Access lists might be employed to prevent unauthorized connections to the device. You also can create a read-only user account so that reports might be viewed.
For skeptical users, a profile mode is available. You can profile the traffic on your network and see what kind of performance gains you might achieve after placing the box inline. Logging is accomplished via SNMP and syslog support. Other benefits include a central management system, the ability to preserve and set type of service /Differentiated Services values, and interoperability with Routing Information Protocol, Open Shortest Path First and Border Gateway Protocol, which means you don't have to manually specify each subnet you want to turn compression on for, as the list is provided via your routing table.
The SR-50 performed equally well on TCP and UDP traffic without incurring significant latency. In our FTP testing, while a significant reduction of data occurred on the first transmission, the pattern matched to what is in memory and recalled from the cache in subsequent transfers to further lower latency.
A VoIP call placed over the network showed a moderate level of compression. A denial-of-service attack was launched to disrupt the VoIP call, but the attack was easily defended against by placing the VoIP traffic in a prioritized queue.
Cost: SR-50, including application acceleration, from $7000 (depending on units and bandwidth ordered).
Pros: LED front panel makes configuration a breeze; routing protocol interoperability; solid performance.
Cons: QoS features not as rich as Packeteer.