Dell announced today a refresh of the PowerEdge M910 blade server based on the Intel Xeon E7 processor.  The M910 is a full-height blade that can hold 512GB of RAM across 32 DIMMs.   The refreshed M910 blade server will also feature Dell’s FlexMem bridge that enables users to use all 32 DIMM slots with only 2 CPUs.  You can read more about the M910 blade server in an earlier blog post of mine here.

According to the Dell  press release issued today, the systems being refreshed with the Intel Xeon E7 processor, which include the PowerEdge M910 blade server  as well as the PowerEdge R910 and R810, offer customers substantial performance gains including;

• Up to 38 percent improvement in Oracle application server and database performance over previous generation Xeon 7500 “Nehalem-EX” based servers *.
• Up to 18:1 server consolidation ratio over four socket dual core processor based systems offering up to 93 percent lower operation costs resulting in a one year return on investment**.
• Up to 34 percent improvement in SQL database virtualization performance and 49 percent higher performance per watt with the combination of Xeon E7 processors and new Low Voltage memory (LV RDIMM) offerings***.

According to Dell, the new Intel Xeon E7 CPUs will be shipping this week.

NIC Partitioning (NPAR)

Dell is also announcing the addition of a Converged Network Adapter (CNA) for the M710HD’s network daughter card, or NDC.  I wrote about the NDC in this post, but as a recap,  it’s a removable card that provides the blade server’s LOM (LAN on Motherboard) network adapter.  Previously Dell offered a 4 port Gigabit Ethernet card, but this 2nd offering opens the ability for users with the M710HD to upgrade to a fully converged 10Gb infrastructure.  This card is also the first adapter to offer the industry’s first “network partitioning” or NPAR scheme that makes it possible to split the 10GbE pipe with granularity free of any fabric vendor lock-in.  NPAR enables optimal use of physical network links allowing each 10GbE port to be carved up into multiple physical 1GbE NICs without the use of software and without any CPU overhead.  For example, each 10GbE port can be divided into up to four multiple physical NICs totaling 10Gb offering more flexibility.  The NPAR scheme is handled by the Unified Server Configurator, enabled by the Lifecycle Controller that is embedded on the server.  For more information on the Unified Server Configurator, check out Dell’s website:
http://content.dell.com/us/en/enterprise/d/solutions/unified-server-configurator.aspx

*Oracle: Based on Dell and Oracle testing performed in March 2011 running an industry-standard SPEC Java Enterprise benchmark . SPEC® is a registered trademark of the Standard Performance Evaluation Corporation.  Actual performance will vary based on configuration, usage and manufacturing variability.

**Consolidation: Up to 18:1 server consolidation performance with return on investment in about one year” claim estimated based on comparison between 4S MP Intel® Xeon® processor 7041 (dual-core with Intel® HyperThreading Technology, 4M cache, 3.00GHz, 800MHz FSB, formerly code named Paxville) and 4S Intel® Xeon® processor E7-4870 (30M cache, 2.40GHz, 6.4GT/s Intel® QPI, code named Westmere-EX) based servers. Calculation includes analysis based on performance, power, cooling, electricity rates, operating system annual license costs and estimated server costs. This assumes 42U racks, $0.10 per kWh, cooling costs are 2x the server power consumption costs, operating system license cost of $900/year per server, per server cost of $36,000 based on estimated list prices, and estimated server utilization rates. All dollar figures are approximate. Estimated SPECint*_rate_base2006 performance and power results are measured for Intel® Xeon® processor E7-4870 and estimated for Intel Xeon processor 7041 based servers. Platform power was measured during the steady state window of the benchmark run and at idle. Performance gain compared to baseline was 18x (truncated).

***Based on the DVD Store 2 benchmark testing  performed by Dell Labs in March 2011.  Actual performance and power draw will vary based on configuration, usage and manufacturing variability.

Updated 1/27/2011
Dell quietly announced the addition of a 10 Gigabit Ethernet (10GbE) switch module, known as the M8428-k.  This blade module advertises 600 ns low-latency, wire-speed,  10GbE performance, Fibre Channel over Ethernet (FCoE) switching, and low-latency 8 Gb Fibre Channel (FC) switching and connectivity.
In this era of virtualization in the datacenter, CPU and RAM capabilities on servers are no longer the bottleneck, but instead it is the servers’ I/O.  Technology vendors are now offering 10Gb Ethernet options which helps to reduce the traditional Ethernet I/O bottleneck.  In fact, many vendors have begun to offer a convergence of networking allowing both Ethernet and Storage commands to travel down a single converged pipe.  These devices, known as converged network switches have been around since Cisco first introduced their Nexus product a couple of years ago.  The design of the original converged switch took a single 10Gb cable to a Top of Rack (ToR) switch in which the Ethernet traffic and the fibre storage traffic was separated.

Over the past 9 months we’ve seen a surge of a new generation of converged switches, that does not require a ToR switch, but instead splits the Ethernet and Fibre traffic out of a single module within a blade chassis.  The advantages of this approach are: it frees up blade server mezzanine (expansion card) slots and it also requires fewer modules within a blade chassis therefore freeing up space for additional future fabrics.  In late December 2010, Dell announced their next generation converged module, called the M8428-k Converged 10GbE Switch Module.  The M8428-k has 28 ports: 16 internal and 12 external.  The 12 external ports are made up of:

*8 ports of 10 Gigabit Ethernet (GbE) speed and Data Center Bridging (DCB). 
*4 Fibre Channel ports: that support 2, 4, and 8 Gbps full duplex

The 16 internal ports (one per Dell M1000e Server bay) supports DCB and Fibre Channel over Ethernet (FCoE).  One last point to make, the module comes with all ports activated.  This is an advantage over some competitors that offer upgrade licenses to activate additional ports.

UPDATE – Currently the only compatible blade server mezzanine Converged Network Adapter (CNA) is the BR1741M-k.  This card provides FCoE connectivity from blade server to the M8428-k Converged Switch Module.  You can read more about this card on this Dell site.

For more information ont he Dell M8428-k Converged 10GbE Switch Module, visit Dell.com.

BL2x220c G5 (2 server "nodes" shown)

HP’s BladeSystem server offering is quite extensive – everything from a 4 CPU Intel blade to an Itanium CPU blade, however their most well hidden, secret blade is their BL2x220c blade server.  Starting at $6,129, this blade server is an awesome feet of design because it is not just 1 server, it is 2 serversin 1 blade case – in a clam shell design (see below).  This means that in a HP C7000 BladeSystem chassis you could have 32 servers!    That’s 64 CPUs, 256 CORES, 2TB of RAM all in a 10U rack space.  That’s pretty impressive.  Let me break it down for you.  Each “node” on a single 2 node BL2x220c G5 server contains:

• Up to two Quad-Core Intel® Xeon® 5400 sequence processors
• Up to 32 GB (4 x 8 GB) of memory, supported by (4) slots of PC2-5300 Registered DIMMs, 667 MHz
• 1 non-hot plug small form factor SATA or Solid State hard drive
• Embedded Dual-port NC326i Gigabit Server Adapter
• One (1) I/O expansion slots via mezzanine card
• One (1) internal USB 2.0 connector for security key devices and USB drive keys

You may have noticed that this server is a “G5″ version and currently has the older Intel 5400 series processors.  Based on HP’s current blade offering, expect to see HP refresh of this server to a “G6″ model that will contain the Intel® Xeon® 5500 series processors.  Once that happens, I expect for more memoryslots to come with it, since the Intel® Xeon® 5500 series processors have 3 memory channels.  I’m guessing 12 memory slots “per node” or 24 memory slots per BL2x220c G6.  Purely speculation on my part, but it would make sense.  

Why do I consider this server to be one of HP’s best hidden secrets?  Simply because with that amount of server density, server processing power and server memory, the BL2x220c could become a perfect virtualization server.   Now if they’d only make a converged network adapter (CNA)…

Other than Cisco’s UCS offering, IBM is currently the only blade vendor who offers a Converged Network Adapter (CNA) for the blade server.  The 2 port CNA sits on the server in a PCI express slot and is mapped to high speed bays with CNA port #1 going to High Speed Bay #7 and CNA port #2 going to High Speed Bay #9.  Here’s an overview of the IBM BladeCenter H I/O Architecture (click to open large image:)

Since the CNAs are only switched to I/O Bays 7 and 9, those are the only bays that require a “switch” for the converged traffic to leave the chassis.  At this time, the only option to get the converged traffic out of the IBM BladeCenter H is via a 10Gb “pass-thru” module.  A pass-thru module is not a switch – it just passes the signal through to the next layer, in this case the Cisco Nexus 5000. 

10 Gb Ethernet Pass-thru Module for IBM BladeCenter

The pass-thru module is relatively inexpensive, however it requires a connection to the Nexus 5000 for every server that has a CNA installed.  As a reminder, the IBM BladeCenter H can hold up to 14 servers with CNAs installed so that would require 14 of the 20 ports on a Nexus 5010.  This is a small cost to pay, however to gain the 80% efficiency that 10Gb Datacenter Ethernet (or Converged Enhanced Ethernet) offers.  The overall architecture for the IBM Blade Server with CNA + IBM BladeCenter H + Cisco Nexus 5000 would look like this (click to open larger image:)

Hopefully when IBM announces their Cisco Nexus 4000 switch for the IBM BladeCenter H later this month, it will provide connectivity to CNAs on the IBM Blade server and it will help consolidate the amount of connections required to the Cisco Nexus 5000 from 14 to perhaps 6 connections ;)