Silver Peak NX-3500's caching algorithms shine

2 devices perform admirably, according to tests.

Silver Peak Systems submitted two NX-3500 WAN-optimization devices for testing in our Alabama labs. The devices performed admirably, and we suspect that 2-year-old Silver Peak will become a major contender in the WAN-optimization market.

The NX-3500 was reliable, had excellent protocol support, and was amazingly easy to set up and configure. We were concerned with its ability to scale, but Silver Peak says the company plans to extend the NX product line in the future.

To continue reading, register here and become an Insider. You'll get free access to premium content from CIO, Computerworld, CSO, InfoWorld, and Network World. See more Insider content or sign in.

Silver Peak Systems submitted two NX-3500 WAN-optimization devices for testing in our Alabama labs. The devices performed admirably, and we suspect that 2-year-old Silver Peak will become a major contender in the WAN-optimization market.

Click to see: Clear Choice Roundup logo

The NX-3500 was reliable, had excellent protocol support, and was amazingly easy to set up and configure. We were concerned with its ability to scale, but Silver Peak says the company plans to extend the NX product line in the future.

Intended for regional hubs and small data centers, the NX-3500 appliance ($18,000) works on a peer-to-peer basis when connected via WAN link to other NX-3500s; aggregates multiple session data from as many as 50 model NX-2500s (a $10,000 entry-level machine with less capacity than the NX-3500) at remote sites; and, when used with the high-end Silver Peak NX-7500, ties a hub or small data center to a major data center at a central site.

The NX-7500, the vendor says, handles 155Mbps WAN speeds, has 2TB of disk space and is priced at $65,000. We think the huge price and performance gap between the NX-3500 and NX-7500 cries out for at least one midrange WAN-optimization device to help capacity planners choose the right tool for the right job.

Caching strength

The NX-3500 did a good job of compressing data and accelerating protocol streams. Its biggest strength, however, was its effective caching algorithms. In our FTP tests, we transferred large files from remote branches to a central site. Between transfers, we made slight changes to the files to emulate a typical day's business activity against the files.

For every transfer after the first, the NX-3500 units transported just the compressed changes (known as compressed deltas) over the WAN link. A transfer that consumed 19 minutes without WAN optimization took about 20 to 35 seconds with the NX-3500 and yielded an average bandwidth-increase factor of 28.5.

The NX-3500's caching approach was intelligent and practical. The appliance copied inbound and outbound traffic onto its internal hard disk. While this data store is application-independent, the NX-3500 remembered the operation (for example, FTP) requesting the data, in addition to the identity of the session partner making the request.

The NX-3500 also tracked the number of requests for the data. Least-used data was discarded when the cache became full, but daily or weekly use of the data will keep it in the cache. When the appliance detected changes to the cached data, it computed deltas to represent those changes. Any session partner (for example, another NX-3500) requesting the data repeatedly received the deltas (compressed), reassembled the data at the destination site and supplied the full, correct data set to, for example, an FTP client asking for the data.

WAN OPTIMIZATION NX-3500 Silver Peak Systems

4.0

Price: $18,000. Pro: Excellent caching feature. Con: Product line needs a midrange option. The breakdown Performance 30% 5 Scoring Key: 5: Exceptional 4: Very good 3: Average 2: Below average 1: Subpar or not available Protocol support 20% 4 Ease of use 20% 4 Scalability 20% 3 Documentation/installation 10% 3 TOTAL SCORE 4.0

Click to see:

Beyond file transfers, the NX-3500s perform Common Internet File System- and Common Internet File System-based file copy operations, as well as backup and restore operations, that benefit from caching to a degree similar to that of the FTP operations. For e-mail and Web (HTTP) traffic, we saw an average bandwidth-increase factor of 8.0 to 12.4. As expected, pure text traffic gave us the higher number, compared to messages and Web pages that contained graphic images. For Citrix and VoIP traffic, we experienced an average bandwidth-increase factor of 3.5. To our delight, the NX-3500 devices added virtually no latency to our WAN links, and can support WAN speeds as much as 10Mbps.