The final presenters for Tech Field Day 5 were from HP. HP presented on two different architectures that at first seemed to be somewhat unrelated. The first was their HP StoreOnce data deduplication appliances. The second was an overview of the technologies that comprise the HP Networking converged networking solutions. These two technologies are very intrinsic to the future of the datacenter solutions offered by HP.
After a short marketing overview about HP and their direction in the market, as well as reinforcement of their commitment to open standards (more on this later), we got our first tech presentation from Jeff DiCorpo. He talked to us about the HP StoreOnce deduplication appliances. These units are designed to sit inline with your storage and servers and deduplicate the data as it flies past. The idea of inline dedupe is quite appealing to those customer that have many remote branch offices and would prefer to reduce the amount of data being sent across the wire to a central backup location. By deduping the data in the branch before sending it along, the backup windows can be shorter and the costs associated with starving other applications with high data usage can be avoided. I haven’t really been delving into the backup solutions focused on the datacenter, but as I heard about what HP is doing with their line of appliances, it started to make a little more sense to me. The trend to me appears to be one where the data is being centralized again in one location, much like the old days of mainframe computing. For those locations that don’t have the ability or the need to centralize data in a large SAN environment, the HP StoreOnce appliances can shorten backup times for that critical remote site data. The appliances can even be used internal to your datacenter to dedupe the data before it is presented to the backup servers. The limits of the things that can be done with deduplication seem to be endless. My networking background tends to have me thinking about data in relatively small streams. But as I start encountering more and more backup data that needs priority treatment, the more I think that some kind of deduplication software or hardware is needed to reduce those large data streams. There was a lot of talk at Tech Field Data about dedupe, and the HP solution appears to be an interesting one for the datacenter.
Afterwards, Jay Mellman of HP Networking talked to us about the value proposition of HP Converged Networking. While not a pure marketing overview, there were the typical case studies and even a “G” word printed in the bottom corner of one slide. Once Jay was finished, I did ask a few questions about the position of HP Networking in regards to their number one competitor, Cisco. Jay admitted that HP is doing its best to force Cisco to change the way they do business. The Cisco quarterly results had been released while I was at TFD, and the fact that there was less revenue was not lost on HP. I asked Jay about the historical position of HP Network (formerly Procurve) and his stance that the idea of an edge-centric design was a better model than Cisco’s core-focused guidelines. Having worked with both sets of hardware and seen reference documentation for each vendor, I can say that there is most definitely disagreement. Cisco tends to focus its designs around strong cores of Catalyst 6500 or Nexus 7000 switches. The access layer tends to be simple port aggregation where few decisions are made. This is due to the historical advantage Cisco has enjoyed with its core products. HP has always maintained that keeping the intelligence of the network out in the edge, what Cisco would term the “access layer”, is what allows them to be very agile and keep the processing of network traffic closer to the intended target. I think part of this edge-centric focus has been because the historic core switching offerings from HP have been somewhat spartan compared to the Cisco offering. I think this situation was remedied with the acquisition of 3Com/H3C and their A-series chassis switches. This gives HP a great platform to launch back into the core. As such, I’ve seen a lot more designs from HP that are beginning to talk about core networking. Who’s right in all this? I can’t say. This is one of those OSPF – IS-IS kind of arguments. Each has their appeal and their deficiencies.
After Jay, we heard from Jeff about the tech specs of the A-series switches. He talked about the support HP has for the open standards in the datacenter. Casually mentioned was the support for standards such as TRILL and QCN, but not for Cisco FabricPath. As expected, Jeff made sure to point out that FabricPath was Cisco proprietary and wasn’t supported by the A-series. He did speak about Intelligent Resilient Framework (IRF), which is a technology used by HP to unify the control plane of a set of switches to make it appear as one unified fabric. To me, this sounds a lot like the VSS solution that Cisco uses on their core switches. HP is positioning this as an option to flatten the network by creating lots of trunked (Etherchanneled) connections between the devices in the datacenter. I specifically asked if they were using this as a placeholder until TRILL is ratified as a standard. The answer was ‘yes’. As IRF is a technology acquired from the H3C purchase, it only runs on the A-series switches. In addition, there are enhancements above and beyond those offered by TRILL that will ensure IRF will still be used even after TRILL is finalized and put into production. So, with all that in mind, allow me to take my turn at Johnny Carson’s magnificent Karnac routine:
The answer is: Cisco FabricPath OR HP IRF
The question? What is a proprietary technology used by a vendor in lieu of an open standard that allows a customer to flatten their datacenter today while still retaining several key features that will allow it to be useful even after ratification of the standard?
The presentation continued to talk about the trends and technolgy in the datacenter for enabling multi-hop Fiber Channel over Ethernet (FCoE) and the ability of the HP Flexfabric modules to support many different types of connectivity in the C7000 blade chassis. I think that this is where the Cisco/HP battle is going to be won or lost. By racing towards a fast and cost-effective multi-hop FCoE solution, HP and Cisco are hoping to have a large install base ready for the standards to become totally finalized. When that day comes, they will be able to work alongside the standard and enjoy the fruits of a hard-fought war. Time will tell whether or not this approach will work or who will come out on top, if anyone.
I think HP has some interesting arguments for their datacenter products. They’ve also been making servers for a long time and they have a very compelling solution set for customers that incorporates storage, which is something Cisco currently lacks without a partner like EMC. What I would like to see HP focus more on in their solution presentation is telling me what they can do and what the are about. Conversely, they should spend a little less time comparing themselves to Cisco and taking each opportunity to mention how Cisco doesn’t support standards and has no previous experience in the server market. To be honest, I don’t hear that from Cisco or IBM when I talk to them about servers or storage or networking. I hear what they have to offer. HP, if you can give me all the information I need to make my decision and your product is the one that fits my needs the best, you shouldn’t have to worry about what my opinion of your competitors is.
Tech Field Day Disclosure
HP was a sponsor of Tech Field Day 5, and as such was responsible for a portion of my airfare and hotel accommodations. In addition, HP provided their Executive Briefing Center in Cupertino, CA for the Friday presentations. They also served a great hot breakfast and allowed us unlimited use of their self-serve Starbucks coffee, espresso and chai machine. We returned the favor by running it out of steamed milk for use in the yummy Dirty Chai. HP also provided the delegates with a notepad and pen. At no time did HP ask for nor were they promised any kind of consideration in this article. Any and all analysis and opinions are mine and mine alone.