Bechtolsheim: The server is not the network
Whatever Cisco says
HPC on Wall Street Andy Bechtolsheim knows a thing or two about servers, storage, and networking. He co-founded workstation and server maker Sun Microsystems as well as two networking companies: one that he sold to Cisco and became the basis of its Gigabit Ethernet biz, and another that he recently started and runs while working one day a week at Sun.
And when he gives one of the keynote addresses at today's HPC on Wall Street event in New York, one of the themes will be that there is lots of room for innovation among stand-alone network equipment providers.
Prior to the event, Bechtolsheim took some time to yack on the phone with El Reg about his company, Arista Networks, which is carving out a niche for itself as a supplier of 10 Gigabit Ethernet products that have the low latency and low cost that supercomputer and other high-performance computing shops - particularly the financial services firms running trading and market data systems where every microsecond counts.
Bechtolsheim has a longer and deeper history in networking than he does in servers at this point, and he's well aware of how the two play off each other in various computing environments. He was working as a PhD student on a project to integrate networking interfaces with microprocessors while at Stanford University when he was tapped by Sun's other co-founders, Scott McNealy and Vinod Khosla, to be the upstart's first chief technology officer.
He stuck around Sun as CTO until 1995, when he left to start a company called Granite Systems, which made Gigabit Ethernet switches and which was flipped a little more than a year later when Bechtolsheim sold it to Cisco Systems. The Granite Systems products eventually evolved into the Catalyst 4000 series of switches at Cisco, and Bechtolsheim was the general manager of this product line at Cisco.
In 2001, Bechtolsheim caught the entrepreneurial bug again, and he saw the InfiniBand switched fabric as the next place to make some money and have a technological impact. And so he founded another startup called Kealia, which created the "Galaxy" line of Opteron-based blade servers, the "Magnum" monster InfiniBand switch, and the "Thumper" X4500 storage servers that came to market individually after Sun bought Kealia in early 2004 and made Bechtolsheim CTO of its server biz. The original vision that Kealia had, of course, was for an integrated blade and storage platform with an InfiniBand backbone, something Sun is selling as the Constellation System to HPC and media streaming customers.
While Sun was busily peddling the Constellation System and open sourcing its entire software stack, including Solaris and Java, Bechtolsheim caught the entrepreneurial bug yet again, and this time was smitten by 10 Gigabit Ethernet. As El Reg reported last October when Arista came out of startup mode, while Bechtolsheim was still at Sun - and with its permission - he started up a 10GE switching company originally called Arastra. Last fall, the company changed its name to Arista Networks just as he hired Jayshree Ullal, formerly senior vice president of the datacenter, switching, and services group at Cisco, to be president and chief executive officer with Bechtolsheim reprising his role as CTO and founder.
Next page: 10GE goes mainstream
How many servers could fill 10Gb link?I suppose not all servers could (or you would even want them to).
However I can see in some arcatectures may need the enhanced speed. Where for example you have servers doing front end and load balancing, cacheing and other things. I suspect where you have massive databases or video feeds where highspeed storage is held (ram Drives for example) you will start to see these sort of volumes of trafic being thrown about.
Anyway, In general you never want to use 100% of your bandwith, so you have some room for errors and issues with cable quality and so on.
I know when I was last lurking around in a server room, they were already shotgunning 1Gb to get higher threwput, so I presume 10Gb is just the next step.
I also noted the trend towards lots of smaller servers doing the same task over a single large one for cost effective and reduncancy reasons. Not to mention the separation of storage from servers. All these devices need to connect and communicate lots of diffrent data to lots of diffrent servers (and externaly).
In a situateion where you have 3 networks per Server (One to link all like servers to each other to maintain coheasion, one that links servers to clients, and one that links servers to data storage or similar) it would be much easyer to have a single 10Gb link and a few virtual networks, rather than 3x1Gb networks. Anyway, 10Gb switches become really handy when joining multiple 1Gb networks togeather.
Not to mention that chaps argument for 10Gb uptake is that, even for servers that may not need 10Gb, they will recive that upgrade because the cost diffrence between 1Gb and 10Gb will shrink to near nowt. In the short term it will be for connecting clusters, and specific specalist installs such as supper computers, and storage arrays.
"The server is the network"
I seem to remember it was Novell that coined the expression "The server is the network" - causing some confusion amongst the less informed.
Given that network and systems tend to run in different teams with separate budgets and different preferred suppliers, Cisco's bid to muscle into the server world looks a bit ambitious to me.
10GE: I had to come up with the networking for some rack systems recently and my boss was pushing 10GE. When I showed him that just the 10GE upgrades for our Cisco kit cost more than all the server kit put together, he was less keen. We went with multiple 1G instead.
How many servers could actually fill a 10G link?
External versus internal network
Converging the external network with the internal (storage/compute interconnect) one is something any risk averse manger would regard with horror. They are very different needs. Just about everything above the need to move packets of bytes around is different. Especially the security and reliability issues. But also the patterns of data and provisioning needs are different. So not only would you want an air gap between the networks, almost everything inside the switches is going to focussed on different issues.
One suspects that the reason FCoE has any future is because it provides FC, and everything that entails. It is something people can have the warm fuzzies about, and thus satisfy their risk averse nature. Infiniband makes a great deal more sense if you were scratch starting system design. Or more pertinant; building atop RDMA is a good place to start. Especially with SSD coming into play. But since the vast majority of customers will want stories about seamless upgrade paths from where they are, with understandable risk, FC is not going away any time soon.