Bull updates its server platform architecture, reinforces HPC positioning

These systems come in two models and are the first Bull products implementing a new platform architecture called Mesca (Multiple Environments in SCaleable Environments). Fundamentally, this architecture provides a single platform that can either accomodate (Nehalem) Xeon (i.e. X86) or (Tukwila) Itanium processors and enables quasi-linear scaleability up to16 processors per machine (i.e. 128 nodes) using Intel's QuickPath Interconnect (QPI) and a critical, Bull-developed high performance switch called BCS. This technology, which can be compared to IBM's also recently announced eX5 architecture, greatly expands the memory addressing capacity of x86 servers (in Bull's case, up to 2 TB). Implementing a Symmetric MultiProcessing (SMP) architecture, which enables applications to see the entire memory as a single addressable space, these servers avoid complex application design requirements and are well-suited to the current needs of what Bull calls “Extreme Computing”.

The BullX Supernodes implement a 3U form factor, bringing a density that remains lower than that of blade-based systems. However, the achieved density is largely sufficient for many high performance environments where cooling capacity and energy availability are a constraint, limiting anyway the achievable density. In addition, the new systems bring connectivity capabilities that are well-suited to HPC environments.

Finally, Bull also paid attention to operations continuity as well as energy optimisation, notably with the introduction of « super-capacitors » on the motherboards. These devices enable the servers to weather electrical micro-interruptions by providing up to 250 ms on-board electrical autonomy. This allows Bull to implement dual power supplies which operate on an active/passive (stand-by) mode rather than on the traditional mode with both power supplies active. This is more efficient from an energy consumption standpoint. In addition, it will allow data center designers to implement lower cost and more energy effective uninterruptible power supplies (since these will have an extended delay to « take over », and will not have to handle micro-interruptions).

Duquesne Group believes this announcement will allow Bull to reinforce its relatively recently acquired leading position in the High Performance Computing (HPC) market. This is a highly competitive and innovative market. The permanent quest for performance has always been a key characteristic of this market, but (as HPC expands in scope) is now compounded by economic, availability, reliability and useability requirements that impose constraints very similar to those encountered in « conventional » IT markets. This market context enables Bull to leverage its technical capabilities as well as its long-acquired knowledge of IT critical production environments in the mainframe and large Unix systems world.

Obviously, the MESCA platform technology will be applied by Bull in other system environments: supporting Itanium chips means it can be implemented in Bull GCOS machines; and the characteristics of the platform (notably the performance and the SMP architecture) make it a suitable all-purpose base for high-end servers. For a midsized manufacturer like Bull, the potential versatility of the new platform is a major asset, allowing R&D leverage across a broader product line. This will also make the technology potentially attractive to OEMs.

HPC is a traditionally innovative IT market segment; many technologies have initially found their way into this segment prior to mainstream market adoption. With economic and operational constraints now becoming also prominent in HPC, we believe it will be even more in the vanguard for IT users. Large IT organisations will benefit from watching this space to identify opportunities to deliver value cost-effectively to business users.

Hardware, system and network innovations implemented in HPC environments (e.g. energy-optimised data center designs or Infiniband connexions applied to large server clusters) will continue to percolate into the business IT market. For example, HPC requires very large memory management capacities, but this also applies to a growing number of business applications. To cite an example, as the SMP design eliminates the application code and data organisation complexities associated with share-nothing, massively parallel designs, it enables the fast deployment of very large databases, as well as large, full in-memory databases. MESCA-based systems will also be well-suited to the implementation and management of very large virtualised environments.

With this announcement, Bull confirms and expands its ability to compete in the HPC market; it also acquires a largely expanded capacity to leverage its technology into other IT market segments. Beyond the pure technological, engineering and manufacturing capabilities of the company, we believe Bull's strong customer intimacy (exemplified by Bull's close cooperation with the French Commissariat à l'Energie Atomique (CEA) while it was developing MESCA and Bullx Supernodes) and the resulting ability to exactly address user requirements is key to its success in segments such as HPC or the legacy GCOS base.

As Bull remains a mid-sized, « niche-focused » vendor, the challenge will be to develop a market share on the innovation-demanding segment of the large server market. This market opportunity does exist and is likely to expand with the generalisation of virtualisation and the emergence of cloud computing. Bull has an opportunity there as long as it remains capable of surfing on the forefront of the innovation wave, staying ahead of the commoditised mainstream IT market. The challenge will be to spot and reach the next waves, combining a strong involvement in the technology ecosystem with excellent customer intimacy, in a number of cases, up to the level of true “co-innovation”.