HP On Right Track With Two Network OSes

When HP completed the 3Com acquisition, it seemed obvious that the ProCurve product line was due for termination. In the last two years, HP has steadfastly continued with ProCurve and its custom ASIC development, and maintains that both the A-Series and the ProCurve occupy different spaces in the market. Maintaining two different product lines makes sense for HP and its customers.

This approach is not the expected behavior. Traditionally, after an acquisition like 3Com we would expect that all product lines would be merged or discontinued to follow a single production process or a single operating system. By maintaining both Comware and ProCurve, HP is committed to developing and maintain two network operating systems. At Interop in Las Vegas this year, I questioned HP Executive VP and General Manager, Enterprise Servers, Dave Donatelli about this practice. I received a firm look and terse comment that HP Networking is committed to both platforms, but no further questions were permitted.

The A-Series switches rely primarily on merchant silicon, and the core product value is in the Comware software that runs on the platform. The platform has a wide range of features and is mostly comparable with competitor products. The A10500 is a recently released mid-range FlexCampus switch using Broadcom silicon (the same silicon as used in the Cisco Nexus 3000).

The E-Series switches are the old ProCurve line. The core silicon and software for these products is developed by the HP ProCurve division in the United States. Prior to the purchase of 3Com, the ProCurve was developing a business that was competitive with Cisco C29xx and C37xx switches that are used in large volumes for connecting user desktops. Many large companies continued to use Cisco C4500 and C6500 devices for core campus switching and ProCurve stackable switches in the access layer to connect user desktops. The S-Series is a value product line for very low-cost requirements that also uses merchant silicon.In a recent blog post, ProVision Inside, some more more insight was offered on what drives this:

So what do we think we can bring to this market that makes us want to go through the pain and cost of doing our own ASICs? A lower cost switch family with functionality and flexibility that targets this specific market space.

The engineering control we have with ProVision allows us to make carefully balanced engineering decisions concerning features and on-chip resources. The features are right-sized and focused. For example, routing tables are sized to be reasonable but not too large. Resources usually found outside of the switch ASIC, such as the module CPU itself, and packet and ACL (TCAM) memory, are appropriately sized, based on our knowledge of our customer needs, and integrated onto the ASIC. It also gives us ASICs that supports true chassis products, with the high performance and modular connection flexibility that that brings, for the edge. You won't find chassis in this space in other vendor product lines.

And in the comments, Dan Montesanto, switch product manager for the Networking Unit, says:

In a nutshell, we think custom ASICs make a lot of sense in the 'middle of the market'. On the low-end, where features don't matter, nearly all the vendors use merchant ASICs. Likewise, in the high-end all the investment/value is in software--so vendors gladly use merchant ASICs so they can focus on switch firmware. Given that the A12500/A9500/A7500 products have been mostly designed to tackle data center and large enterprise campus LAN core deployments--where the longest datasheet feature list wins--I think it makes sense that those platforms are built using merchant ASICs, leaving the vendors to focus precious R&D on software.

In the middle--mid-size businesses or at the edge of large enterprise campus LAN environments-- customers need more than what low-end products can provide but don't want the complexity or cost that would come with products using full-featured ASICs really built for those feature-driven deployments.

In 2000 HP first recognized that there was an opportunity to built custom ASICs to enable us to build products that could strike this middle-ground feature-value balance. The result: simplified/lower cost hardware (through integration driving the need for fewer components), hardware flexibility (ASIC-level feature programmability) and ASIC-integrated security (ASIC-level DoS protection, memory error detection). Customers got products with balanced feature sets and a fair bit of investment protection at attractive price points.The claim is that HP Engineering is able to make cheaper and better chips than the open market when targeted at a specific point in the market. There is evidence to support this--HP manufactures a vast range of products, from printers to servers to tablets, and has extensive manufacturing expertise that provides market leadership in those categories. Consider that HP is able to successfully compete against Asian manufacturers such as Lenovo and Acer, and it's a reasonable claim.

Since the ProCurve product includes a range of chassis-based switches such as the E8200/E5400/E3500, the product is targeted at the mid-size business with relatively simple networking needs. The passive optical backplane that was unveiled at Interop has the primary purpose of reducing the cost of ProCurve switches. The optical backplane uses standard plastic molding and laser etching to reduce the manufacturing cost of the backplane itself to a fraction of conventional backplanes.

The A-Series products have enterprise features and capabilities, and is targeted at complex and demanding networks. The E-series product is targeted at low-complexity and price-sensitive customers. There are two operating systems, with different features and capabilities. The ProCurve operating system has been plagued by bugs and has a poor reputation for software reliability, and many engineers are reluctant to embrace ProCurve on these experiences.

HP seems to be applying manufacturing expertise for design, and operational capacity for volume, to produce the E-series at a price point that provides a competitive edge against the other vendors. Vendors, such as Arista, will be using merchant silicon and have only their own operating systems to differentiate themselves from other players. Their hardware pricing will be determined by merchant silicon. On the other hand, Cisco's recent woes with long development times on products such the Supervisor 2T for the Catalyst 6500 suggest that using your own silicon is also fraught with challenges.

It must also be challenging within HP Networking to have two competing product lines. At the point where the product lines overlap, there must be significant friction between divisions that want their products to "go to customer." Over the years, Cisco internal sales teams have struggled to position the C4500 and C6500 products to customers, often breaking into public view of customers as ugly little spats. Customers find the overlap confusing, and it takes a great deal of time to to research each product. Many customers talk of simpler purchasing decisions as a away of saving money, but the mantra of "choice" seems to continue.

One thing seems clear: HP Networking hasn't convinced the wider market that both Comware and ProCurve operating systems are necessary, and most network architects expect HP to migrate its product line to Comware. On the other hand, the companies that will buy the ProCurve products aren't likely to have network architects who can evaluate these choices. Many businesses need low-cost and low-complexity switches that have chassis features for reliability and performance, and these customers may be well-served by ProCurve.