Network Engines, Inc. - Double the CPUs for Web Serving

  • Written By:
  • Published:

Vendor Genesis

Network Engines, Inc. (NEI) was formed in 1997 as an IP content delivery-focused vendor. The founder was Lawrence Genovesi, who continues to run the company as President. The original product was a multi-processor server targeted at streaming media providers. In 1998, the company shifted its focus to clustered server products, based on the belief that Internet growth would lead to the need for scalable, cost-effective servers.

NEI's revenues come almost exclusively (95%) from product sales and licensing agreements. Presently, the company farms out most of its support activities, although it does have an in-house support group that answers customer questions. Network Engines presently focuses on the high-end ISP and ASP market, primarily on those customers who need the power and flexibility of clustered server appliances, buy significant quantities of them (>50 per order) and are willing to pay $5000+ for a system. This is in contrast to its closest competitor Cobalt Networks, which focuses on lower-end vendors and prices its systems accordingly -- ~$1000-$2000 each. Network Engines is presently considered a niche player, due to both its market focus and size.

NEI is privately held, but we estimate revenues in the $20M range for the preceding 12 months. The bulk of this revenue is from sales/licensing of their WebEngine product line. When sold directly to end users, this product usually includes NEI's maintenance processor and management software. Licensed systems are delivered both with and without maintenance processors, so revenue from licensing is mixed vis--vis the maintenance processor.

Vendor Strategy and Trajectory

NEI presently focuses on the higher-end web serving marketplace. Its current strategy is to provide functionality, rather than low price (Note: the sub-$10K price tag for a well-featured system is not high, but it is at the middle-to-upper end of the server appliance market.)The next logical step will be either to go after the "next-lower" market segment, i.e. the mid-range ISPs/ASPs, or the low-end market, where Cobalt currently focuses. If NEI chooses to expand its focus, we believe the mid-range market is where they will attack first - the product changes required will be less drastic than those required for an attack on Cobalt's territory. The company will also likely move into other high-end horizontal markets, such as Network Attached Storage (NAS) and E-commerce, as well as developing partnerships with companies having complementary products.

NEI is currently a niche player, but has the potential to become a market leader in the server appliance arena because of their combination of powerful hardware (2 CPUs in a 1U-high rack chassis) and their maintenance and clustering software. It should also be noted that IBM has chosen the WebEngine as the hardware platform for their Netfinity 4000R, whose high-density offering we wrote about in early September. (See TEC News Analysis "IBM Announces Netfinity 4000R Super-Thin Server", 9/10/99)

NEI will use its high-CPU-density hardware, distribution through IBM, and maintenance processor and software to increase revenue and market share for the near term. For the long-term, expansion into other hardware areas and market segments (both horizontal and vertical) is expected. In addition, NEI will focus on building its internal resources, with the goal of increasing its product offerings and product quality.

Vendor Strengths

System density: Network Engines is strong in two key areas: system density; and maintenance software. As mentioned previously, the WebEngine sets the new CPU density standard, at two CPUs per one rack height unit (1U). In a market where compute density is the Holy Grail, this sets a new standard. The next-densest current products are the eight-CPU servers (Compaq, Dell, HP) packaged in a 7U chassis - the WebEngine increases that density by 75%. NEI's offering is so compelling that IBM, not known for its densely-packaged Wintel servers, is using the WebEngine system as the basis for its Netfinity 4000R.

Maintenance Processor Software: On the software side, their maintenance processor/software is designed to allow remote monitoring and control of their servers, and will provide status even when the server is not powered up. This is a significant market differentiator - we believe NEI's software is the only one of its type currently available in the marketplace.

Clustering: Network Engines' clustering software will support up to 255 systems (plus one for administration of the cluster). This is in contrast to Windows NT 4.0, which supports a maximum of two servers. One caveat: the load balancing and replication "modules" are reasonable, but not best in class. NEI understands that its management software represents a competitive advantage in the long term. NEI's ability to execute on its vision remains to be seen.

Network Engines can use its dense hardware to leverage licensing deals with other big-name server hardware vendors who may not want to invest in developing their own products. Likely candidates include Dell and HP (or Intel, who typically supplies server hardware to Dell and HP). Compaq tends to keep their development in-house.

Vendor Challenges

Performance: Although the vendor has been shipping the WebEngine since June, 1999, there are still no audited/validated performance figures for the product. One of the key benchmarks for web servers is requests per second. Although NEI's product literature publishes figures which would indicate good performance (1400 hits/second for a single-processor model), Network Engines cannot provide an audited report, similar to the Web Polygraph report by Data Communications magazine. In addition, although NEI can theoretically cluster up to 256 WebEngines, they have no data on the scalability "percentage factor". [Definition of "percentage factor": if one WebEngine can handle "Z" requests per second, then "N" servers should be able to handle a total of P x N x Z requests, where "P" is the percentage factor. A perfectly scalable cluster - not possible in the "real world" - would have P = 1.00.]

System serviceability: We believe the system should be made more serviceable, both in user access and ability to change components. Although thin web server vendors (including Network Engines) generally operate from the "box upgrade" point of view, there are times when customers would prefer not having to swap out a $5000 FRU (Field Replaceable Unit), but instead just deal with a power supply or hard disk drive. Additionally, customers may want the hard drive hot-swappability that TruSOLUTIONS (a manufacturer of a 1U-high general-purpose server, as differentiated from a web server or server appliance), a leading Network Engines competitor provides.

Price: At a base price of approximately $5000, Network Engines clearly intends to target the higher end segment of the thin web server market. However, customers may question whether the $3000 premium (vs. offerings from Cobalt and TruSOLUTIONS, two other 1U-high server vendors) is justified by the extra-CPU capability and maintenance processor.

Company size: As with most technology markets, big fish eat little fish. Consolidation has already begun in the cache serving market, consolidation will continue in the other server appliance sub-markets. Although Network Engines plans to follow in the IPO footsteps of Cobalt and CacheFlow, their size makes them vulnerable for buyout. However, with the backing of Egan-Managed Capital (and its unofficial link to EMC Corp.), this is not as great a challenge to NEI.

The first two issues can be ameliorated by Network Engines adding more staff in the design and performance engineering areas. The price issue can be addressed by Network Engines offering a lower-end version.

We do not see company size as a serious long-term problem. We do not see Network Engines acquiring anyone in the next 1-2 years, but believe it will continue to hire, and leverage relationships with partner(s) as the means to plug any strategic gaps. Its current partnership with IBM provides it with a "ready-made" Quality Assurance department in the form of IBM's testing staff. However, this relationship will only carry NEI so far, and they need to quickly build their own group.

Vendor Predictions

We believe this vendor will grow rapidly over the next three-to-five years. NEI has a good product set, and their obstacles are surmountable. We believe their strategy of focusing on functionality will be successful (65% probability), and they will continue to grow at a rate comparable to the server appliance market, which we estimate to be 100% CAGR.

We believe Network Engines will start to compete more directly against Cobalt Networks and TruSOLUTIONS, since they are the only other 1U-high server vendors of any significant volume. A third vendor, StarBox Netsystems, has announced plans to ship a 1U-high system in January 2000, but we do not see them as a major player in this market.

Vendor Recommendations

To establish and maintain a leading market position, Network Engines should do the following:

Diversify: The market for $5000+ web servers is large, but will be a low-growth area in 3-5 years (55% probability). Therefore, NEI should focus on lower-price products, to compete in the same area as Cobalt and TruSOLUTIONS.

Add serviceability: As mentioned earlier, Network Engines needs a system with more serviceable components (e.g. hot-swap drives) to provide an additional, compelling message to customers. We believe this can be done easily within their normal design process.

More OEM agreements: Network Engines should try to repeat its IBM success with other major vendors such as Dell or HP. A deal with Compaq is unlikely because it tends to keep development in-house. As the field becomes more crowded, it will be harder for Network Engines to differentiate themselves. Arranging OEM deals in the next 3-6 months will help NEI keep ahead of this trend.

User Recommendations

Users who are trying to build their web serving infrastructure and want maximum CPU density should seriously consider Network Engines. Although the price appears high, we believe its maintenance processor software and clustering ability give NEI an functional advantage. However, users wishing for best-in-class load balancing and replication software should consider offerings from other vendors, such as those from RADWARE or F5 Networks.

Users looking for low-end web servers will find Cobalt Networks's RaQ series a better fit for their budgets. In addition, customers interested in purchasing a relatively small number of servers (less than 25) may find that Cobalt or one of the other larger-box web server vendors may be more appropriate. As discussed earlier, users should research and review the company's long term viability, especially as it relates to company size. ANetwork Engines IPO will alleviate some of the viability concerns.

comments powered by Disqus