10 Gigabit Ethernet still too expensive on servers
Another reason why x86 server sales 'underwhelm'
SaaS data loss: The problem you didn’t know you had
It may not be entirely clear what is holding back the x86 server racket in the most recent quarter – round up the usual suspects – but networking market watcher Dell'Oro says it knows what isn't helping as it was expected: 10 Gigabit Ethernet.
In a report casing the 10GE server adapter market it is a bit hard to figure out which came first: the Intel "Romley" Xeon E5 server platform chicken or the 10GE LAN-on-motherboard egg for the servers based on this combination of "Sandy Bridge" CPUs and "Patsburg" chipsets. What is clear to the analysts at Dell'Oro is that 10GE adapter sales were almost flat sequentially in Q3, and that was not what the server makers or Dell'Oro were expecting.
"We see two key factors constraining server migration to 10 Gbps network connectivity," explained Sameh Boujelbene, senior analyst for the controller and adapter market at Dell'Oro, in the statement. "First, slow migration to Romley-based server platforms, and second, users continue to opt for 1 Gbps – the price premium for 10 Gbps versus 1 Gbps is too wide of a gap."
The ramp for 10GE was supposed to follow a pattern similar to Fast Ethernet and Gigabit Ethernet before it. In those cases, companies installed some faster switches at a premium and use more expensive PCI-based adapter cards for those applications where the higher bandwidth and lower latency is necessary and justifies the higher cost.
Then, a year or two later, the volume ramp for that latest Ethernet technology takes off skyward when the cost (and now power consumption) of the chips for implementing that technology come down and are integrated on the motherboard with a perceived zero cost to the server buyer.
To Dell'Oro, it looks like the Romley ramp is taking longer than expected, and that is slowing down the 10GE server ramp.
"The Romley server transition is looking to be a four-to-five quarter transition rather than a two-to-three quarter transition for many of the major server manufacturers," adds Boujelbene. "Third quarter 2012 was a critical quarter as we expected some of the large server manufacturers to be shipping mostly Romley-based servers, but this did not manifest."
Part of the problem might be that the Romley launch itself was staged, with pre-launch machines shipping in August and September last year around. Intel was widely expected to launch the initial Xeon E5-2600 processors for two-socket servers in the third quarter of 2011, and then in the fourth quarter.
But issues with the SAS disk controller in the chipset (it was supposed to run at 6Gb/sec but was backstepped to 3Gb/sec) and other factors in the market compelled Intel to push the Xeon E5-2600 and Romley platform launch to March and to hold back the Xeon E5-2400 (for cheaper two-socket boxes) and the E5-4600 (for four-socket boxes) until May. Part of the problem, it would seem, is that the Romley ramp itself took some time and was later than expected.
The other factor, of course, is that the global economy has been more jittery than a drunk who hasn't had a shot for a day or two, and the uncertainty around the US presidential election and what the winner might do for (or more precisely to) the economy also caused companies to pull back a bit.
Uncertainties in the fiscal cliffs in both Europe and the US ripple around the globe, affecting all nations in this finite and yet unbounded spherical economy. (The world is not flat, but it is a rather thin veneer on a globe, a transition layer between molten rock and empty space that has most of the amenities we find convenient. Except when there is nowhere else to colonize and therefore extract unfair economic advantage.)
Something is clearly not going as well as it could be in the x86 server racket, that's also for sure, and it doesn’t look like Dell is the culprit. Dell got out of the gate with Xeon E5 server shipments ahead of most of its competitors with the revamping of its PowerEdge server line with its 12G systems.
In the most recent quarter ended on November 2, Dell reported that Xeon E5 machines represented two-third of its revenues. And hyperscale systems built by its Data Center Solutions unit grew 26 per cent.
Dell's networking business grew 40 per cent, but Dell does not yet break server and networking sales apart. Together, they comprised $2.32bn and rose 11 per cent year-on-year in the most recent quarter, which was the company's third quarter of fiscal 2013.
Force10 Networks is the bulk of Dell's networking revenues and had an estimated $200m annual run rate when Dell acquired it in July 2011. Dell doesn't have a Romley server ramp problem, but because the LAN-on-motherboard is a snap in card and does not come for free, customers may be balking just the same at moving on up to 10GE.
That suggests that the problem is not the Romley platform per se, but rather the fact that 10GE is a priced option on a lot of Romley systems and Gigabit Ethernet is still the default.
Here's an example. On that PowerEdge R620 two-socket, 1U rack server, a Broadcom Gigabit Ethernet daughter card is the default LOM snap-in, and it costs $159 to put in the quad-port Intel Gigabit Ethernet card. The Broadcom two-port 10GE plus two-port Gigabit daughter card that snaps in to the mobo costs $629, and an Intel daughter card that has two 10GE ports and a single Gigabit port costs $659. This is on a server that has a base price of $1,559.
The fact that the x86 server market as a whole is not doing well is not helping the 10GE cause on the server side, either. In the quarter ended on October 31, Hewlett-Packard's ProLiant, BladeSystem, and Scalable System revenues in aggregate fell by 7 per cent to $3.14bn, with particular softness in x86 server sales in EMEA.
IBM doesn't provide revenue figures by server product line, but did say during its call going over its September quarter that its sales were off 5 percent for its System x and BladeCenter lines.
The box counters at IDC and Gartner should be releasing their third quarter report cards for the server racket soon, and we will perhaps get some more insight on what is going on. ®
COMMENTS
Kind of forgetting the cost of the switches there aren't they
It's no good having cheap 10G Ethernet cards in your servers if the switches still cost a bloody fortune.
£400 10Gig adapters aren't the issue. They cost roughly three times as much as high quality Gigabit adapters. Spending that much on organic tech growth is pretty much discretionary.
By contrast the switches cost around eight times as much as decent Gigabit switches; they're all managed affairs (which isn't particularly desirable in a lot of backbone networks) and an £8k barrier to entry is not an easy pill for a lot of smaller businesses to swallow. The market is not going to take off until this becomes more balanced.
I am working on a education VDI deployment
on a large greenfield campus. Their preferential vendor was Dell and their core switching is Brocade.
Pricing up the Dell Blades was a strange one because the M620 comes with 10G on the LOM. Additional 10GB Broadcom cards were the same price as 1GB cards and that is before the insane discounts that the education sector gets.
The M1000e was kitted out with 6x IO Aggregators which have QSFP+ on board for a 40GB to uplink. Unless you start spending silly money trying to convert the QSFP+ to optical, you need a SFP+ or QSFP+ capable switch. The one that made sense was the Force10 S2410, but it doesn't exist any more so we had to go for the S4810 which comes in at $30k and then there is the price of the cables. Again education prices made these dramatically cheaper.
Where it got really expensive was the Brocade MLXe 24 port SFP+ modules. Unfortunately the Brocade MLXe doesn't support 40GB QSFP+, but their top of rack switches, the ICX 6615 does. Also Dell were being really fussy about what they would support, so this became a no go.
So really, this reiterates what others have said. Switch prices and options are the problem. I am struggling to justify having 96 SFP+ ports on a two-switch redundancy design, but it was really the only option given to us. When the environment is fully kitted out with SANs, uplinks, connections to the Blade Chassis, we will only need to use 4 per blade chassis, 1 per iscsi SAN and maybe 4 to the core.
Datacentres have a different problem, but environments that are smaller really need some smaller switch options to justify 10GB or a reduced price point.

IT infrastructure monitoring strategies
Agentless Backup is Not a Myth
Top 10 SIEM implementer’s checklist
Steps to Take Before Choosing a Business Continuity Partner
Enabling efficient data center monitoring