Convergence in the Data Center’s Future
Unni NarayananTuesday, February 2nd, 2010
We are at the forefront of an exciting period for data center design and innovation driven by what is occurring in three separate-but-fast-converging market segments: co-location businesses, managed services vendors and cloud providers.
Our view today is that the future, and fortunes, of each segment is inextricably linked and that the relationship of key conditions in each segment — capacity for co-lo, profit margin for managed services and technology innovation for cloud providers — will meaningfully define the opportunities, challenges and performance of what we see as a truly amalgamated marketplace in the next 10 years. Briefly, here is what PGR’s network is reporting:
Capacity — Enterprises suffer a data center capacity shortage brought on by a dearth of viable real estate, a state of affairs that PGR’s network consistently characterizes as fundamental rather than temporary. The capacity squeeze is made worse by growing demand from SMBs among others that want to get out of the data center business with its acute heat/power problems.
Margin — Right now, SMBs and enterprises are willing to pay a premium for the value added by managed services providers. We believe this will result in continued high margins for managed services providers over the foreseeable future.
Technology innovation — While it is clear that elements like SAAS are accelerating, all of the technologies related to bill-back, security and private clouds and back-end support for performance enhancement among others are immature. A lot of heads are “in the clouds” right now and speculation about what is coming and what will occur as a result is intense.
Looking at the three segment reports above, it is clear that developments and conditions in one segment weigh strongly on the others. Thus, factors that impact co-location vendors inevitably affect the prices enterprises are willing to pay managed services providers and create opportunities for cloud providers to innovate to solve IT challenges.
More to come.
![[Post to Twitter]](http://www.pg-research.com/blog/wp-content/plugins/tweet-this/icons/tt-twitter.png){kind=icon}
Tags: Co-location, Enterprise computing, managed services, SAAS, SMBs
Posted in Cloud Computing, Data Center, Unni Narayanan | No Comments »
You Heard It Here First: BRCD a Winner in 2010
Unni NarayananTuesday, January 26th, 2010
Any number of pundits will tell you BRCD is on the ropes, hanging by a fingernail, ready to succumb to a combination of 1.) flat growth rates in FC and, consequently, no TAM to fuel BRCD’s expansion in the data center; 2.) technology obsolescence at the hands of FCoE and big price and margin erosion in core storage products due to insufficient differentiation; then 3.) a plunge into the really deep end with the Foundry acquisition, which the punditry avers will destroy BRCD’s very foundation.
Sounds pretty bad. But wait! Our checks show BRCD is far from rolling over, taking it on the chin or moving to the Jersey Shore. In fact, PGR’s network accentuates the positive noting BRCD style and panache that should stand it very well this year. Check it out:
FC demand is robust because of refreshes and green-field opportunities. If someone counted the total number of FC cable, HBAS and switches in 2010 vs. 2009, they might not see a big difference but it seems no one is counting the huge “rip and replace” opportunity in the ongoing storage and data center consolidation build out.
There is a huge co-location capacity squeeze. New co-lo space is expected to come on line throughout 2010. And, although SANs by and large do not saturate BRCD 8Gb/sec Director class switches, it is only an INCREMENTAL cost to refresh the SAN with high-performance products as part of an upgrade in the co-los. Furthermore, BRCD is well positioned to handle FCoE when it becomes real. That and CSCO has stopped attempting to “buy accounts” with the largely unimpressive Andiamo legacy products. Hence, we do not expect accelerating product erosion.
Everyone likes to say that if BRCD could “do it over again,” they would have given the Foundry acquisition a second thought. However, the reality is that BRCD controls its own destiny if it can manage the integration of the Foundry products well. In that scenario, it is quite possible for BRCD to “pick off” some incremental share from the likes of JNPR. And, with sufficient effort, they can move up in the networking hierarchy. HPQ’s acquisition of COMS is the end of any possible merger with BRCD because the war in the data center between CSCO and HPQ will center around the winner of the storage battle. All that and BRCD is a free agent that can join either side.
Not so bad after all, we say.
Tags: BRCD, Brocade, CSCO, FCoE, Foundry, HPQ
Posted in Data Center, Networking, Unni Narayanan | No Comments »
BRCD Survey: Switches and HBAs? Selling to both sides…
Unni NarayananThursday, October 29th, 2009
It is PGR’s view that BRCD has solidified its hold on the FC switch market and is maintaining solid margins in a macro environment favorable to Greenfield storage deployments. Additionally, our network continues to report BRCD is picking up HBA share but the question remains: at who’s expense?
For different reasons, our network is positive on BRCD. Here’s a summary:
1.) All agree BRCD is really the only game in town in core switching products and margins are robust;
2.) But the value of BRCD’s solid market position is offset somewhat by questions about the fundamental viability of the Fibre Channel market itself. Many of our experts see lackluster growth over the next year, the result of better traction in FCoE;
3.) Right now, our network does not think BRCD poses an immediate threat to QLGC or ELX in HBA but their position gives customers a look at how BRCD can work both sides of the equation. The question remains whether BRCD can get enough out of HBA to drive switch sales. The jury’s still out on that.
Clearly, storage is a key battlefield in the war for the data center. And no clear victor can emerge from the likes of CSCO, HPQ or IBM until a major player focuses on the idiosyncrasies of areas such as SAN switching. So, in the meantime, BRCD is secure in its role as a profiteer selling to both sides.
Tags: BRCD, CSCO, ELX, FCoE, HPQ, IBM, QLGC
Posted in Data Center, Storage, Unni Narayanan | No Comments »
Evolutionary versus Revolutionary Data Centers
Unni NarayananMonday, August 17th, 2009
Microsoft’s and Google’s contrasting data center strategies reflect the differences between evolutionary and revolutionary approaches in technology implementation. History has proven that each paradigm has its own risks and rewards.
MSFT is riding the well known trends toward the usage of data center containers. This natural evolutionary approach requires a minimum investment in surrounding building infrastructure and a reliance on broadly available commodity hardware. The benefits here are a complete negation of risks related to specific vendor exposure and wasted dollars on potentially fruitless R&D efforts. In a sense MSFT is waiting for the risk/reward pareto point to clearly emerge from natural market forces. Historically this is consistent with MSFT’s development efforts (e.g., their DOS relationship with IBM or their relatively late but successful offering with Windows as compared to Apple’s products).
In contrast GOOG believes it already understands the nature of that elusive optimal data center solution. GOOG’s teams rightfully view themselves as technology visionaries and want to take a revolutionary approach in design. Hence, GOOG needs to develop custom hardware solutions. This strategy has not always paid off. For example, GOOG vacillated between purchasing servers from vendors such as RACK and designing their own compute platform solutions. And yet, our checks at Primary Global Research indicate that for the time being GOOG’s data center approach is a standard that many wish to emulate.
And, while the data center design battle rages, in the backdrop state tax laws are in flux. This is an issue that transcends both the evolutionary and revolutionary approaches. Neither MSFT nor GOOG can predict the political whims of the taxpayer – especially, when data center consolidation drives lower TCO (totally cost of ownership) and that savings is largely a result of reduced headcount. Once the legal ambiguities dissipate, the battle lines will be more cleanly drawn and, as Ashlee Vance writes in his recent NY Times Bits blog column, the evolutionists and revolutionists will be fighting in a neighborhood near you!
Tags: Data Center, GOOG, MSFT
Posted in Data Center, Unni Narayanan | No Comments »
Co-Lo Mojo Less So
Unni NarayananMonday, August 10th, 2009
Q2 of 2009, PGR started a program of quarterly co-location checks. Our network has consistently reported a shortage of available co-location space with price-points increasing. This does not appear to be a short lived phenomenon. Rather, there are fundamental drivers making available capacity quite tight, including: 1) Advances in cloud computing, 2) Maturity and widespread adoption of virtualization technologies, 3) An inherent shortage of viable real estate, 4) Momentum behind municipalities to remain green, which further exacerbates demand problems, and 5) Increased power and cooling challenges that are too difficult for in house IT personnel to handle.
Look to see this thesis updated and enhanced as we continue our quarterly checks.
Tags: Co-location
Posted in Cloud Computing, Data Center, Unni Narayanan | No Comments »
HPQ: Network Rookie
Unni NarayananMonday, August 10th, 2009
On paper at least, HPQ looked like a “shoo-in” to join the networking big time with its Procurve offering. And I must confess, I have been cheering for the rookie if only to give CSCO, the salty veteran of interconnectivity, some competition.
However, PGR’s talent scouts (HPQ/CSCO experts) report a completely different story. Specifically: 1) HPQ has a poor track record with networking hardware, 2) The “go to market” strategy around Procurve is nonexistent, 3) Organizational changes at HPQ don’t appear to favor good execution around Procurve, and 4) CSCO just has a better sales force. Still, important questions remain in terms of which player will ultimately win the data center. For example, CSCO’s blade strategy has momentum largely because CSCO’s name is on it (not because there’s a real product).
As they say in sports, come game time, both rookie and veteran have to show up and play. HPQ’s presence on the server side is real. The question is whether HPQ can leverage the talent to raise its game.
Tags: CSCO, HPQ
Posted in Cloud Computing, Data Center, Networking, Unni Narayanan | No Comments »
Wall Street’s Embrace a Precursor of Enterprise Acceptance
Unni NarayananThursday, July 2nd, 2009
Wall Street’s adoption of cutting edge technology has always been a precursor to broader mainstream enterprise acceptance. The reason is that financial firms — and trading companies, particularly — are comfortable with a higher degree of risk. As a result, Wall Street seems to gravitate to technologies at precisely the instant these technologies start to have favorable risk/reward profiles. There are numerous examples, ranging from Salomon Brothers’ aggressive utilization of higher end workstations like Sun in the 1980s to the Nasdaq’s implementation of its Supermontage system.
In this context, Ben Worthen’s June 29 story in the Wall Street Journal, “Remaking the NYSE’s Data Center” highlights a couple of interesting points: 1.) NYSE CIO Steve Rubinow states the exchange is utilizing a best-of-breed approach. This is consistent with where the data center design is headed. For example our checks at Primary Global Research indicate no single vendor is going to “own” a vision for cloud computing. It is more likely that cloud computing architectures will organically unfold as IT managers (as in the case with the NYSE) piece together an optimal solution; 2.) That the vendor list includes Juniper, Ciena, and Voltaire is notable for several reasons. First, there is the obvious non-mention of Cisco, which is part of a broader statement that competitors are closing the gap. But, more interesting is the revelation that NYSE’s design uses optical networking products (Ciena) and will implement Infiniband (Voltaire). Infiniband has been relegated to the high performance computing niche (think Livermore Labs, etc.) for years. Perhaps these technologies are at inflection points where we might see more widespread usage.
Although many enterprises do not push the envelope in a manner that requires the “renting of space” for proximity reasons in the data center, one thing remains clear: demand for performance and bandwidth remains unfulfilled and numerous emerging technology companies are finally in an attractive sweet spot to address these challenges and reap the rewards.
Tags: Ciena, Cisco, Juniper, Voltaire
Posted in Data Center, Technology, Unni Narayanan | No Comments »
Cisco’s Blade Server Strategy
Rajan VaradarajanWednesday, July 1st, 2009
Cisco already has had dominant market share in its core businesses. In looking for new sources of revenue and continued growth, the company saw the datacenter virtualization trend as a great opportunity. Perhaps it was a challenge as well, to enter a server market dominated by its traditional partners such as HP, Dell and IBM and, hence, antagonize them. This is inevitable, because wherever Cisco goes for growth, it is going to step on somebody’s toes.
Datacenter design is changing rapidly to meet the demands of massive amounts of data being driven largely by video applications. With the benefits of virtualization at hand, designers are beginning to take a top-down view of the datacenter. Cisco has teamed up with the leader in the space, VMWare, and storage behemoth EMC to just do that. In addition, Cisco also has partnered with NetApp in its grand vision of the USC- Unified Computing System.
The basic question may be, why does Cisco really care about server virtualization?
One possible answer: In a datacenter, each virtual machine mimics a physical server. Obviously, Cisco cares about any technology that increases bandwidth usage and thereby helps sell its mainstay routers and switches. Some examples are video like TelePresence, WebEx, and now Flip Video. Cisco wants to accelerate the adoption of 10Gb/s port usage so it can start building the 40Gb/s and 100Gb/s gears of the future. Datacenters are the first places that can happen. However HP, DELL and IBM are neither in a hurry nor motivated to provide their server blades with 10Gb/s ports yet. They appear to be content with the existing 1Gb/s ports. So, Cisco saw an opportunity to push its 10Gb/s adoption in server virtualization. Instead of waiting for other companies, Cisco is now building blade servers for the datacenter that are customized for server virtualization. These blades provide up to 1Gb/s pipe to each virtual machine, which would mean more 10Gb/s ports at the backend of the datacenter, helping in Cisco’s 10Gb/s push.
An alternate possible answer as to why Cisco cares about server virtualization: HP has recently stepped up its efforts in building networking gear, stepping on Cisco’s toes. HP’s Procurve product line seems to be gaining traction among its enterprise customers. In addition, with the acquisition of one of Cisco’s largest customers, EDS, HP is beginning to challenge Cisco’s dominance in the market. By some accounts, HP is the second-largest networking gear vendor.
Perhaps the reason is a combination of both the answers or there may be a third one. However, Cisco definitely has the wherewithal - money, technical prowess, innovative spirit, and great leadership in John Chambers - to deliver on the promise of its vision - UCS.
Conclusion: Admittedly, this is a very big vision. In hard times, Cisco re-invents itself as it has done many times in the past. I believe it has the fundamentals, savvy management and competitive spirit to pull off this one as well.
Tags: data center virtualization
Posted in Data Center, Networking, Rajan Varadarajan, Semiconductor, Software, Solar, Technology | No Comments »
Ntap/Data Domain: duplicate product?
Unni NarayananTuesday, June 9th, 2009
At first glance, NetApp may now appear to have “duplicate” products since it already offers a free de-duplication solution. But, the dynamics of the marketplace make the arranged marriage of NTAP and Data Domain a must. The reality is threefold: (more…)
Tags: data canter, ntap
Posted in Data Center, Technology, Unni Narayanan | No Comments »