HP Shows its Next Generation Blade and Converged Infrastructure – No Revolution, but Strong Evolution
With the next major spin of Intel server CPUs due later this year, HP’s customers have been waiting for HP’s next iteration of its core c-Class BladeSystem, which has been on the market for almost 7 years without any major changes to its basic architecture. IBM made a major enhancement to its BladeCenter architecture, replacing it with the new Pure Systems, and Cisco’s offering is new enough that it should last for at least another three years without a major architectural refresh, leaving HP customers to wonder when HP was going to introduce its next blade enclosure, and whether it would be compatible with current products.
At their partner conference this week, HP announced a range of enhancements to its blade product line that on combination represent a strong evolution of the current product while maintaining compatibility with current investments. This positioning is similar to what IBM did with its BladeCenter to BladeCenter-H upgrade, preserving current customer investment and extending the life of the current server and peripheral modules for several more years.
Tech Stuff – What Was Announced
Among the goodies announced on February 19 was an assortment of performance and functionality enhancements, including:
- Platinum enclosure — The centerpiece of the announcement was the new c7000 Platinum enclosure, which boosts the speed of the midplane signal paths from 10 GHz to 14GHz, for an increase of 40% in raw bandwidth of the critical midplane, across which all of the enclosure I/O travels. In addition to the increased bandwidth midplane, the new enclosure incorporates location aware sensors and also doubles the available storage bandwidth.
- 40 Gb Ethernet switch — An industry first, HP also announced a 40Gb Ethernet switch for the c7000 enclosure that can provide full 40 Gb/port bandwidth to the blades. Each blade can have up to two 40 Gb mezzanine adapters, which when combined with the embedded 2×10 Gb FlexNICs can provide up to 100 Gb of connectivity to a single blade. While we don’t expect HP to be alone for long, this is a major advance for bandwidth-hungry applications. Additionally, the switch latency is a claimed 250 ns, extremely low for an Ethernet switch, which will make the high-velocity trading crowd salivate.[i]
- Enhancements to Virtual Connect (VC) – HP’s Virtual Connect was a highly disruptive technology when it was first shipped in 2007. Since then IBM has made major strides, and Cisco’s UCS has provided a credible challenge on a number of fronts, including enhanced QoS. HP’s latest set of enhancements is a direct response to the challenges offered by Cisco, particularly in managing end-to-end QOS. VC release 4 now includes much more comprehensive QoS capabilities as well as the ability to handle multi-hop FCoE traffic, allowing multiple VC modules to connect to a TOR switch (most likely a Cisco Nexus 5000) before breaking out the FCoE into its component FC and Ethernet traffic. Additional enhancements include automatic configuration assists for configuring VC with VMware virtual machines, a labor cost saving and error-reducing improvement.
- HP SmartMemory — An ingenious 24 GB three rank RDIMM memory module that offers both improved density per slot over 16 GB DIMMs as well as a claimed 25% performance improvement over conventional DIMM configurations of the same capacity. In addition to the performance boost, the 24 Gb SmartMemory will help to counter the availability of 24 DIMM slot offerings from Cisco and IBM.
In addition to these core enhancements, HP also announced enhanced storage offerings and a new workstation blade with up to 8 GPUs targeted at large-scale VDI deployments.
What It Means For HP Customers
For HP customers, this announcement is unmitigated good news, offering:
- Roadmap clarity — It removes any uncertainty about the next several years for the c7000 product life (although it does look like I have been wrong in my predictions that HP would make fundamental architectural changes this year or next — painful, but needs to be said).
- A bridge to the next generation of Ethernet technology — The improved midplane performance will support the next generation of 40 Gb Ethernet, along with 16 Gb FC, ensuring that c7000 customers can smoothly move to the next performance plateau without disrupting their current investments.
- A continued commitment to Virtual Connect — The improvements in VC release 4 are significant, and will make many aspects of configuring and managing VC, especially in VMware environments, which are probably the majority deployment scenario for the c7000 today and in the future.
- A clear sign that the server group has not lost its focus — In earlier posts on the impact of HP’s management and financial fiascos, I stated that the core product groups were still competent and focused. I view this announcement as a clear token of this focus. While it is impossible to tell whether or not the decision to enhance rather than replace was driven purely by pragmatic product planning or partially influenced by budget constraints, it looks to me to be a very carefully focused set of enhancements delivered by a group of people who have kept their heads down and focused on their customers for the last 1 to 2 years, regardless of what their executive management has been up to.
For HP Competitors
HP’s primary competitors in blades/converged infrastructure are IBM and Cisco. From a competitive standpoint, HP’s improvements shore up some critical shortcomings and will make it that much more difficult for competitors to displace HP on a technology feature basis. When I used to fly small airplanes, the acronym “EFC” meant “Expect Further Clearance” from air traffic control in the event of inclement weather, congestion or other irregularities. If I were forecasting the competitive weather in the core converged infrastructure and blade segment (Wow, is this stretching a figure of speech or what?), I would steal the FAAs acronym and say that EFC now means “Expect Further Competition” — with three dominant world-class system suppliers (and enough second-tier competition to avoid any chance of oligopolistic behavior) competing for the fastest growing segment of the x86 market, it will remain intensely competitive and richly rewarding to customers for the foreseeable future.