News & Analysis
Comment
Bert22306
Yes indeed, good for Quanta, bad for Dell and HP, eh? None of this is surprising ...
escher
HP and Dell have created their own Frankenstein monster, which now has the ...
Taiwanese server makers outflanking traditional OEMs
Rick Merritt
2/14/2013 3:01 AM EST
The software challenge--and opportunity
The new business model brings challenges. Taiwan’s ODMs have long experience building low cost Wintel systems. But data center racks typically run Linux and other low-level open source code that needs to be carefully tweaked for best performance and lowest power.
So Quanta is now hiring software engineers at about twice the rate as hardware people.
“Power consumption highly depends on the specific workload of the data center,” said Yang. “I need to have people with expertise in BIOS, firmware and rack management software."
Yang will also invest this year in creating a software team to explore the OpenFlow protocol and software-defined networking. Google has championed the technology and other big data centers are showing interest in it as a way to simplify their increasingly large, complex networks.
“I believe big data centers will develop their own OpenFlow efforts, so I’d like to set up a team to help them,” said Yang, who aims to sell a complete package of data center hardware and software.
Quanta designs all its own top-of-rack switches, typically using Broadcom chips. Thus, it sees a natural progression to adopting OpenFlow. In this way, the Quantas of Taiwan will not only take on Dell and HP, but the Ciscos, Junipers and Aristas of the world as well.
Unlike typical business computer users, the big data centers are mainly focused on fast layer 2 networking with minimal complexity in layer three protocols, Yang said. “They have closed environments, understand the features they need and can control what they want to do,” he added.
Next: Keep the silicon simple
The new business model brings challenges. Taiwan’s ODMs have long experience building low cost Wintel systems. But data center racks typically run Linux and other low-level open source code that needs to be carefully tweaked for best performance and lowest power.
So Quanta is now hiring software engineers at about twice the rate as hardware people.
“Power consumption highly depends on the specific workload of the data center,” said Yang. “I need to have people with expertise in BIOS, firmware and rack management software."
Yang will also invest this year in creating a software team to explore the OpenFlow protocol and software-defined networking. Google has championed the technology and other big data centers are showing interest in it as a way to simplify their increasingly large, complex networks.
“I believe big data centers will develop their own OpenFlow efforts, so I’d like to set up a team to help them,” said Yang, who aims to sell a complete package of data center hardware and software.
Quanta designs all its own top-of-rack switches, typically using Broadcom chips. Thus, it sees a natural progression to adopting OpenFlow. In this way, the Quantas of Taiwan will not only take on Dell and HP, but the Ciscos, Junipers and Aristas of the world as well.
Unlike typical business computer users, the big data centers are mainly focused on fast layer 2 networking with minimal complexity in layer three protocols, Yang said. “They have closed environments, understand the features they need and can control what they want to do,” he added.
Next: Keep the silicon simple
|
|
|
|
|
Navigate to related information
rick.merritt
2/14/2013 11:09 AM EST
How do you see the server business shifting in the cloud era?
Sign in to Reply
marcopol
2/15/2013 7:43 AM EST
To me this seems to be a general drawback of outsourcing the manufacturing and the know-how to an ODM. In long term, you rise a competitor.
Sign in to Reply
rick.merritt
2/15/2013 11:19 AM EST
Indeed contacts at HP bewail the fact the few systems engineering experts they have in servers are approaching retirement age.
Sign in to Reply
FuturePlus
2/15/2013 3:57 PM EST
All's I know is that memory DIMM failures for these big cloud servers is the #2 failure in the data center. Not only do they need simplier IC's but more reliable memory subsystems.
Sign in to Reply
rick.merritt
2/15/2013 4:41 PM EST
Whats the problem with the DIMM failures as far as you have heard?
Sign in to Reply
przemek
2/15/2013 5:18 PM EST
Funny because there was a recent article on Japanese DRAM industry, failing because they insisted on high-overhead manufacturing and testing for quality, and being undercut by competition which adopted shortcuts. I guess the chicken are coming home to roost.
Sign in to Reply
escher
2/17/2013 8:20 PM EST
HP and Dell have created their own Frankenstein monster, which now has the advantage of knowing more about making their products than the OEMs do (ref: Ricks comment about HP systems experts retiring with no one to take their place).
Sign in to Reply
Bert22306
2/18/2013 2:01 AM EST
Yes indeed, good for Quanta, bad for Dell and HP, eh? None of this is surprising though, right?
Now, how about Foxconn/Apple? I that the next article we'll be reading along these lines?
Sign in to Reply