Egads, a new column? Actually, the return of an old one.
What you’re reading is a “revival” of Yours Truly’s “Nitty Gritty” hardware and fault resilient computing-for-telecom column that ran from 1999 until 2003 in Computer Telephony magazine, later called Communications Convergence.
As I wrote in that first column: “So be it. I am now in the (presumably) enviable position of serving as the Walter Winchell of the products that really make computer telephony go, and go with great reliability. So, all ye wayward fundamental CT products, come forth and be deconstructed here!”
For that statement to be applicable to this new column, of course, we should substitute the words “Computer Telephony” with “IP Communications.” Time, after all, marches on.
So, I hope you all now have a good idea of what this column is about. We don’t cover Broadway openings, tasty new recipes, or ripping yarns about any of my former bosses — but if you’ve got a particularly juicy one, by all means send it to me! (Just kidding.)
We begin this second Nitty Gritty era by taking a look at a new member of IBM’s BladeCenter family. Bruce Anthony, Distinguished Engineer and CTO of IBM’s Telecommunication Server Line, says: “This is an extension of the BladeCenter family, taking the BladeCenter H product and making applicable to the telecom core network. The new device is called the BladeCenter HT.”
The BladeCenter is IBM’s blade-and-server offering that’s been on the market since 2002. Hundreds of thousands of them have shipped, and Anthony says that IBM is currently the number one market share leader for blade servers in the world. An IBM BladeCenter essentially comprises a chassis holding a common shared infrastructure of elements for such things as power supplies, cooling, etc. It also allows you to plug in servers as one would slide books into a bookcase.
“The servers are self-contained systems,” says Anthony, “with their own memory, processors, operating systems, and so on. They all connect together through an Ethernet backplane and then to the outside world through network switches.”
“So think of the BladeCenter as an enterprise version of that design,” says Anthony. “The first generation of BladeCenter was based on 1 Gbps networking technology, where each blade had four network ports into the backplane and there was a set of switches that aggregated those ports together and gave you an uplink that switches traffic off the chassis to additional systems. The new BladeCenter H that we introduced earlier this year added 10 Gbps networking, so you can have four 1 Gbps networks and four 10 Gbps networks off of each blade going to a total of eight switches in the chassis, and then off-chassis from there.”
“And now, BladeCenter HT is a telcom-optimized version of that chassis that’s geared toward the central office [CO] environment,” says Anthony. “So we add NEBS [Network Equipment Building System] Level 3 and ETSI [European Telecommunications Standards Institute] compliance, DC power supplies, all kinds central office alarming, and all of the ruggedness that you need to be able to place a computer system into a central office environment.”
As for the BladeCenter’s form factor, it’s neither CompactPCI nor AdvancedTCA (both used in telecom). “It’s actually a third form factor that’s an open specification to which anybody can build,” says Anthony. “We’ve opened the specs for BladeCenter and over 750 companies that have downloaded those specs and are working on capabilities for our product.”
“If you think about AdvancedTCA [ATCA],” says Anthony, “it was really a design that was optimized for I/O networking capabilities, so it has a great deal of ‘tailgate space’ on the front for lots of connectors, as well as lots of connectors on the back. These devices take airflow in the bottom and blow it out the top. BladeCenter, on the other hand, was designed for computing in the network. So it takes in air right through the front of the system and blows it out the back, which yields the highest performance cooling specs that we can achieve. Having dealt with the heat, that allows us to focus on building a processor-centric environment, as opposed to the kind of cable-centric environment that ATCA has.”
BladeCenter offers a number of options, says Anthony. “You can put daughtercards within a blade in its slot, to add features such as FibreChannel, 10 gigabit networking, or additional storage capabilities,” he says. “We also have adopted something from the ATCA space, using their AMC [ATCA Mezzanine Card] adapters, so we’re also introducing as an additional announcement to the BladeCenter HT, what we call our AMC carrier. This allows us to plug in four AMCs into a single blade and then slide that into the BladeCenter system and utilize all of the capabilities of the AMC ecosystem.”
“Each of the backplane slots is connected to a switch,” says Anthony, “just like in the earlier PICMG 2.16 Ethernet backplanes. Any of the servers in the system, through the Ethernet, can talk to the slot that has the carrier board upon which are plugged in the AMCs. The carrier itself has a switch on it between the AMCs, to enable them to communicate with each other as well as connecting to the backplane. So, you’re essentially able to have processor-based AMCs and line cards in a mix-and-match format to allow the generation of all kinds of different gateway functions, and then have that connect into the backplane to interact with the rest of the system.”
The topology of the signals connecting the blades is an 8- star configuration: Four 1 Gbps stars and four 10 Gbps stars. The new HT chassis is 19 inches wide can house 12 blades and is 12 U (12 “Rack Units” = 21 inches) in height. As for storage, each blade has a number of options that involve storage. “Typically, the server blades have one or two disk drive slots on them for small form factor SAS [Serial Attached SCSI] drives,” says Anthony. “You may just want to boot your operating system from it and so on, which is what many people use it for. We also have the ability to put in solid-state flash-based drives, if you want to get away from the spinning media. Then we have networked-based storage with fiber channel adapters and fiber channel switches that you can install, so you can get to any of the fiber channel arrays that are out there. We also have iSCSI so you can do Ethernetbased storage as well.”
“We designed the new BladeCenter H and HT chassis to support four wide-lane connections to the switches,” says Anthony, “and one switch offering currently that we have is Infiniband 4X, which we see being used quite a bit in the high performance computing marketplace. We think that in the telecom marketplace, 10 Gbps Ethernet will be a more prevalent substantiation of how the chassis is used, and so to that end a new 10 Gbps Nortel switch is being introduced based upon the BladeCenter HT as well. That will have 10 Gbps connections internally to all the blades, and six 10 Gbps uplinks off the chassis.”
All in all, IBM (quote - news - alert) looks like it has a winner with its flexible, super-powerful BladeCenter HT for telecom. The next Nortel switch you buy may be based on it!
Richard Grigonis is Executive Editor of TMC’s IP Communications Group.
If you are interested in purchasing reprints of this article (in either print or PDF format), please visit Reprint Management Services online at http://www.reprintbuyer.com or contact a representative via e-mail at firstname.lastname@example.org or by phone at 800-290-5460.