From "Cisco Networkers" to "Cisco Live"

I just returned from Cisco Live Milan last week.  The previous time I attended the conference was in summer 2007 when I was representing VMware and the conference was still called "Cisco Networkers".

There has been a sea change in the conference:

• 6 years ago at the solution exchange floor, most I saw were switches/routers and SNMP based network management software; 6 years later I saw some complete data center management stack, including end to end compute, network, storage, and management stack.
• 6 years ago there was a first ever session on "virtualization", which was given by Jacob Jensen, then VMware's networking product manager who is currently running Nexus 9000 product; 6 years later, virtualization and cloud are part of the DNA of this conference.  Among many other things, the InterCloud product announcement showcased Cisco's ability to be thought leader in the cloud era.  

Networking is still an element of the conference obviously but it is no longer "THE" element.  Now I can see why the conference was renamed from "Cisco Networkers" to "Cisco Live".

OpenStack Hong Kong 2013

I attended OpenStack conference for 2 days this week.  I have to say I was really impressed by the momentum behind OpenStack: 3,000 energetic people from 50 countries came all the way to Hong Kong! It smells like VMworld in the early days. :)

Based on my experiences with OpenStack code base and various discussions with folks in HK, my assessment is OpenStack is still far away from the "prime time", despite so many companies went on stage touting their production deployment.   In my opinion, a lot of people use OpenStack because it is the only game in town for what they want to do -- I don't believe there is an equivalent commercial product. 

People often treat vSphere as the commercial alternative, but I look at OpenStack to vSphere just like NoSQL to Oracle -- OpenStack is solving a fairly different use case and targeting very different audience.  Just like Oracle was never designed for scale-out database, vSphere was not designed for the level of scale-out current OpenStack users want.

For those who put OpenStack in any serious deployment, they must have invested tremendous Dev cycles for the deployment.  So in a way, if you add up the Dev investment, it is not saving money, not today!  They use OpenStack as a framework not a turnkey solution, and build/modify things according to what their own needs.  The base distribution is not quite "complete" at this point, for instance, among other things, the it does not have an installer today and the TripleO project is targeted for 2014.

Finally, I believe the availability of the source code, the relatively meritocracy based community with lots of passion, and the momentum/sexiness of the brand name do give people a lot of imagination on the potential. I suspect by the end of 2014, we will see OpenStack distribution with enough bells and whistles for those who cannot afford to invest their own "Dev" cycles.

Software Defined Job Move

News always travels fast on SDNCentral and virtualization.info.  I started at Cisco this week and will manage all of Cisco's Cloud Networking and Services Group engineering team and execution.

I'm tremendously excited about this new journey for two reasons:

1. Network virtualization is getting mainstream and Cisco is one of the key players in this transition.  I see Cisco offer some unique technical domain expertise, deep physical and virtual integration, broad product portfolios and ecosystem partnership, and solid presence with the customer.  
2. I'm joining a team that I have collaborated closely from my VMware tenure and I've known a lot of folks here well, from engineers to executives.  I have a lot of confidence in what this team can accomplish and I also anticipate to learn a lot out of this journey.

Thanks to an extremely talented team at Big Switch Networks where I spent two years and learned a great deal from each person there.  Thanks to good wishes sent from former co-workers at BSN and VMware and other industry friends.   Networking industry is at an interesting time and our passion in bettering it is absolutely worthwhile.

I'm looking forward to contributing my energy and experiences at Cisco, learning new things everyday, and collaborating with folks both inside and outside Cisco on this worthy cause.

-Howie

PS: Opinions on this blog does NOT represent my current (or former) employer.

OpenDaylight

On April 15,  Gartner released a report titled "OpenDaylight Project Casts Cloud Over SDN Deployment".   Mark Fabbi and Joe Skorupa highlighted the following: "The OpenDaylight Project will likely delay the adoption of enterprise software-defined networking solutions and stifle innovation from smaller, SDN-focused startups with open SDN-based applications and related capabilities."

One month later on May 15, Cisco called out its "leadership in the OpenDaylight project" in the quarterly earning call.

Some people compared OpenDaylight with OpenStack.  Was Gartner telling us that OpenStack was formed to combat incumbents while OpenDaylight was formed to combat start-ups?  Interesting time for both incumbents and start-ups.

Traditional Network Vendors Advancing Switching

In the past few days, I've seen a lot of new product announcement from the switch vendors.  I am pasting a few Network World article links below as they cover the news quite well.  My overall impression is as follows:

• Progress is impressive, but still very much a "speeds and feeds" game.  Density is ever more important in the "flat network" era.  Overall, the big differentiation is on the HDN not SDN.
• When it comes to SDN, OpenFlow and OpenStack are important part of their story
• HP finally has a virtual switch story with its 5900v after Cisco Nexus 1000V and IBM 5000v roll-out.  This is about 4-5 years behind the first mover.

Happy to hear your thoughts too?  Have a nice weekend,

-howie

Arista: http://www.networkworld.com/news/2013/050113-arista-269279.html
    . "The 7500E is a 30Tbps, 14 billion packets/sec non-blocking switch housed in the same 11 RU chassis as the 7500. Densities per chassis are 1,152 10G, 288 40G or 96 100G Ethernet."
    . $10k for each 100G!

HP: http://www.networkworld.com/news/2013/043013-hp-sdn-269254.html
    . "HP also unveiled a virtual switch to run under server hypervisors. The FlexFabric Virtual Switch 5900v supports the IEEE's Ethernet Virtual Bridging -- a.k.a. "VEPA" -- standard for offloading some switching capabilities from the server to a physical switch. The 5900v is optimized for VMware environments and supports mobility and management of network policies for virtual machines
configured through VMware hypervisors."
    . FlexFabric 5,900 ToR, added VEPA
    . FlexFabric 11,908 aggregation switch, 64 40G, OF1.3, 768 10G and
256 40G, TRILL/SPB
    . FlexFabric 12,900 core

Brocade: http://www.networkworld.com/news/2013/043013-brocade-data-center-269231.html
    . vRouter from Vyatte and OpenStack Plug-in

Juniper: http://www.networkworld.com/news/2013/032513-juniper-sdn-268073.html
    . EX9200 based on MX technology/asic, Juniper's 40/100G play (likely will limit Juniper's QFabric data center switch to 10G-only)
    . Separately they announced JunosV based on Contrail technology, leveraging XMPP.  Juniper is pushing XMPP into Open Daylight as well.

Cisco's Insieme Business Unit (no public announcement yet):
   . a fixed 36x40 Gigabit Ethernet (GbE) device;
   . a modular devices supporting 288 40G;
   . a modular devices supporting 576 40G;
   . a SND controller supporting those fabric switches

SDN Career

In the last 3 months alone, about 1,000 engineers sent us their resumes.  As the VP Engineer, it is a humbling experience with many world class talents interested in changing the network world with us together.

I was at Open Network Summit exhibit last week.  I ran into quite some Cisco or Juniper engineers who literally took vacation and paid the conference out of their own pockets just to observe the SDN industry first hand and evaluate their career options.

So what's the ground truth I'm telling the candidates and what should SDN job candidates out there want to know whichever company they have interests?

1) SDN: Hype or Real?

It is a complete hype for those who think SDN is curing cancer; and it is completely real for those who have come to realization the status quo networking technology is a mainframe technology.  For others, it is somewhere in between.

Honestly don't even waste your time in agonizing on "whether" SDN will make it, it is going to happen no matter what.  One may argue which particular established or start-up company is going to lead it.


2) VMware's acquisition of Nicira: good or bad?

It is both a blessing and curse frankly.  It is a blessing because customers started questioning HDN way more than before.  It is a curse because the huge price tag artificially fast forwarded the customer expectation on SDN industry maturity by 2-3 years.  VMware was able to develop and mature its technology a decade ago without much distraction from the crazy press and blogsphere world, whereas everyone is watching closely the "VMware for the Networking" companies.

For now, the "fast forwarding" has not helped the industry in general.  It created misleading perception on the stage the industry is in, how mature the economic model is, how soon the liquidation event shall happen, and when customers are ready to adopt it in mass.


3) Open Daylight Consortium: good or bad?

The "Open" part is good, assuming the consortium lives up to its meritocracy promise, as it gives Big Switch Networks and everyone an opportunity.  The "Consortium" is naturally suspicious as historically there have been more failure than not in 90% of the "Consortium".  Finally the "light" naming must be perfect, especially after Big Switch has set the de facto standard by launching two SDN "lights" earlier ("Floodlight" and "Switch Light"). :-)

Overall it is not yet the time to make any judgment.  My colleague Jason Matlof said recently, "We are cautiously optimistic, with an emphasis on cautious.”


4) Who are the players in this space?


Mostly importantly, we are at the beginning of the SDN journey, the "lead" by anyone today is not meaningful yet.  The team that has the best talent, approach, and execution focus will earn the winning in the future.  That said, a few thoughts on some of the obvious players:

• Cisco is the indisputable leader in networking technology but now has to react to SDN.  Dave Ward shared Cisco's perspectives on OpenFlow and SDN in a fairly technical and thoughtful way lately.  Then there is also Insieme.  Given the proven track record of MPLS team, Insieme will get both internal mind share and external market share.   Insieme is not a start-up and it is a Cisco BU with special purpose and special treatment.  The employees sit in the same Cisco cubicles, drink the same tap water, have access to the same code, employee directory, and bug system, have the same customer account.  They will deliver a differentiated HDN solution with a nice SDN angle.
• VMware pioneered in overlay networking in the past decade (Akimbi's Network Fencing, Nicira's NVP, and VMware's DVS, VXLAN, vCNI) and they fundamentally want to absorb as much networking functions to the server land as possible.  It is a very logical strategy given VMware is fundamentally a server  infrastructure company and the key thing to watch out is how much they leverage from the legacy networking partners.  Too much and too little won't accomplish their mission.
• The aspiration Big Switch has is to bridge the server and networking domains and keep CCIEs relevant but in a different way.  Few SDN pure players have the similar ambition to own an end to end stack from applications, to SDN Controller, all the way to virtual and physical switching stack.

SDN Myth? My take ...

As usual, I enjoy reading my friend Matthew Palmer's thought provoking blog quite a bit!  Matt listed 5 myths and here is my take:

1) In theory everyone should have a controller, yet it is a bit like saying HP/IBM/Dell etc server companies in theory need an OS too.  In practice, customers look for a controller that is not the single point of failure, not a single point of performance bottleneck, and providing additional/clear value than the status quo physical switch embedded stack.

2) At the end of the day, what matters is how the Open Source project is governed not just how it is licensed.  People are always willing to pay for value but what they don't want is a vendor lock-in scenario where the dominant player can literally name the price.

3) I agree to an extent, at least for the ones Matt and I often talked about. :)  I think we should set the expectation appropriately at the first place: network virtualization use case in particular is an exciting journey for both customers and vendors but we are at the very beginning of it as an industry.

4) The goal for a start-up is to have a solid product with indisputable value prop, VC should take care of cash flow issue for them, so positive or not should not be a concern at all for those who have a good product.  I don't know whether big companies want cash flow positive for their SDN P&L? Probably not either, it is too early.

5) I think the most important thing is we need to look at the big picture of why people are moving. Talent shift from one generation of hi-tech companies to another for technology transition.
Just like people left SGI/DEC for Google/VMware at the turn of the century, we are seeing people leaving from previously top engineering organization to the "next Google/VMware".  Big Switch Networks benefited tremendously from this talent migration shift given the networking industry needs a transformation desperately and many top engineers are hungry to make a big impact in this SDN journey.