Worthwhile afternoon at the Microsoft Mountain View campus
Notable startups -
Chalkable, a platform for apps for schools
ToshL, worldwide financial management (from Estonia)
Monogram - personalized shopping platform for iPad
StoryPanda - kids stories on iPad
The most interesting people in terms of velocity and traction were Tokyo Otaku Mode - 6 million people like its anime content. Will be interesting to see them monetize. There were a surprising number of startups focused on clothing and helping people to dress better.
Dave McLure was neither so manic nor so rude as he's been reported. Anyone considering emulating his schtick (he's known for what Americans call, quaintly, "f-bombs") shouldn't - it doesn't scale up well.
Led by Jason Seats, with lots of help from Nicole Glaros, TechStars Cloud ran from January - April in San Antonio, Texas. Here's a first time mentor viewpoint.
The point of the Cloud variation on TechStars' usual model was to focus on infrastructure - companies building the underpinings of the ecosystem which supports consumer oriented apps, and companies building functionality for enterprises. It has the usual 13 week format. There was also an intent to bring startup companies at least temporarily to San Antonio to see if any of them wanted to stay put afterwards.
On the Riverwalk, where it meets E Pecan St. Techstars had office space in the Weston Centre opposite.
The sucessful applicants were companies building a mixture of tools for creating and manipulating APIs, for data analytics, for video editing, and for understanding your cloud spending. Cloudablity was the most advanced, having taken a seed round after its launch in June 2011 - some of the other groups had just formed companies in order to take initial $118,000 investment which is offerred to each succesful applicants.
In the second week of the accelerator, there's a sort of speed dating day, where mentors go through 30 - 45 minutes session with each of the companies whom they think are interesting and which have asked to see them. At the end of the 4th week companies pick which of the mentors they want to work with - mentors are expected to pick one company and focus on it. Cloudability picked me. I'd expected a team to form among the mentors, but mostly the companies dealt with one mentor at a time - they learned to pick and choose advice. Towards the end of the 13 weeks, more of the companies I'd met contacted me, so I ended up advising four of the eleven, mostly by email, with occasional conference calls. During the run up to and on the day of the final review I was in England, so missed going back to San Antonio to attend Demo Day in person. I did review the videos of the pitches: there was a huge improvement over the three months in both content and presentation style from all the companies.
At this point (July 2012) almost every company has additional finance in place, and a much better idea about their target market and prospective customers than they had in January. I learned a lot about how a good accelerator is run (this was the first TechStars Cloud, but it benefitted hugely from the about 15 programs which preceded it).
Now I'm mentoring at Alchemist, much closer to home in Menlo Park, and planning to mentor another Cloud next January.
Having been involved with assorted PR for small and large companies, here's a model I use for thinking about how and when to do what.
First, there is a lot of background noise, and if you go to the trouble of crafting a story about something you'd like strangers as well as your friends to read, then you need enough signal to stick out of the noise.
Expand this a little to a 3D surface metaphor, over which you and your audience is travelling in time. Events are hills - some of these are predictable, like trade shows and holidays, some are not, like weather events and acquisitions. Looking back, they perturb the landscape, with the size of the hill and the slope of its sides representing the volume and trend rate of the commentary about and around the event.
image from Matlab
Each of your audiences is starting from a different place on the surface, whether existing customers, prospects, existing employees, future candidates, existing and potential investors, or politicians. Tailoring your content, the distribution of that content, and the release timing is an effort to get your story to form a noticeable hill in the path of that audience. Each of those audiences will also look back in time (will search for relevant material); including the right terms improves your visibility from some future time looking back.
While planning for a release, remember to account for internal review time - to improve the quality, to check for legal issues, to include other parties mentioned or likely to be influenced by the release.
Lastly, proof read! People won't notice correct spelling and good grammar, but they will mark you down heavily for uncorrected mistakes.
Several weeks back, I talked to a friend of a friend about getting started in the Valley. He has a software development company in Northen Italy, and was able to take a couple of months to stay in San Francisco to find out what it would take for him to move his company there. Here are some of the suggestions, generalized a little.
Mountain View taken from Airship Eureka (2010)
At least a couple of months before you arrive, find one or more contacts who are themselves well connected to startups and developers already in the Valley. In this case, the mutual contact was a family friend who moved from Italy 20 years ago. Ask them to make introductions for you so that you can exchange email ahead of time and set up meetings with some of them.
Look up the list of trade shows going on in Moscone Center - you don't have to sign up, but many thousands of people do, and they hang out in bars and restaurants in that area of town, so you can get a sense for what's going on in that business. If you are developing for Apple users, for example, there's WWDC; and a few days later there's a meetup for iOS developers.
If you can get into one of the accelerator programs, you get some money, but more importantly you get access to shared resources for most of the things that are essential, time consuming to understand, and not really your focus when building a company - accountancy, legal, sometimes office space,and mutual support from peers. See below for lists of accelerators - each of them describe their schedules (most often 13 weeks, several times a year) and competitive criteria for acceptance.
Read about Lean Startups - the google group for discussion is https://groups.google.com/forum/?fromgroups#!forum/lean-startup-circle
Look for blogs from Eric Ries, Steve Blank, and Sean Murphy
Ask for a frank opinion about the quality of your English. It doesn't need to be perfect, but Americans are bad, in general, at listening to non-native speakers - the better your English is, the more effective you can be. One thing which will help is if you can write a blog; aim to post every day, even if it's only a few sentences at a time. Your audience is people like you, who wish they could spend a summer in San Francisco learning how to get a company started here.
Your blog, your Twitter presence, and your LinkedIn page are important - get them organized ahead of time, so that the people to whom your contacts are introducing you will get a positive sense that you are doing interesting stuff from which they might learn (and so whether they will take time to drink coffee and share experiences).
Twitter people - here's a selection from the people I follow, with a little bit of explanation. Look at whom they follow, too, and pick some of them (use what they post and their profiles to guide your choices). Reply to posts if you have anything constructive to say - this builds your own publicly available profile.
@Venturehacks - Angel List founder
@dorkitude founder Keen.io They just completed Techstars Cloud and raised a seed round.
@stormental co-founder @cloudability Techstars Cloud
@rodolfor founder at @Storybricks
@kotikan - mobile app development agency in Edinburgh
@500 500 Startups (Dave McClure)
@springboardnews - accelerator program in London and Cambridge, England
@techstars - @davetisch, @nglaros
@timburks - runs iOS developers meetup in the Valley
@scobleizer - high volume, interviews interesting new startups (works for Rackspace)
@bfeld - founder Techstars, VC at Foundry
@avc - Fred Wilson, VC at Union Square
@skmurphy - see Lean Startups, above
@ycombinator - accelerator
@gezbrady - supports startups for Silicon Valley Bank
@borthwick - CEO Betaworks. Very interesting business investing in media properties including bit.ly
@sgblank - Steve Blank
@martenmickos - Finn moved to Silicon Valley
@msuster - blog Both sides of the table - VC at GRP Partners - read the blog too
Silicon Valley Bank put together an event reviewing companies with less than $5m in revenue, in the big data and analytics space. In-Q-Tel were the sponsor; the audience included senior people from Siemens, Intuit, HP, Autonomy (an HP company), Northrop Grumman, GE, Standard Chartered Bank, IBM, Splunk, Google, NEA, Hummer Winblad, USVP and Jafco Ventures.
After the introductions, the format was 3 sets of 7 companies, presenting for 7 minutes each. After each set there was a break, at which the audience was encouraged to find the presenters to ask questions. The quality of the presentations wasn't uniform, but it was at least as good as the IBM global smartcamp event I mentored earlier in the year - which says volumes about the pre event coaching which must have been applied by Gerald Brady (SVB) and his team.
Most worth attention: NuoDB for the 'elastic' promise; Kaggle for the game approach; Full Contact for people background; Cloudant for database virtualization.
Dave Feinleib's graphic sorts topics; most of these people could actually do most of the other things, but this does reflect their current postitioning.
Notes on the companies
Sumo Logic Log management - think syslog for everything that emits status, in private or public cloud, with reporting and analytics front end.
$15M in Series B funding in Jan 2012 Mountain View, CA
Platfora Hadoop for non specialists in large enterprises - making big data useable without Map Reduce.
$7.2m Series A Oct 2011 San Mateo, CA
Placed Location analytics anchored on lat/long positions. Reminded me of Palmap (Shanghai).
$3.4m seed round April 2012 Seattle, WA NuoDB Grand claims for a new database architecture - scalable,geo-distributed, elastic, SQL/ACID -compliant, shared nothing, asynchronous, peer-to-peer .. this is Jim Starkey's (Datatrieve, InterBase) latest vehicle for improving relational databases.
$10m venture round, April 2012 Cambridge, MA
Metamarkets Data science as a service -trend analysis, spotting anomalies in big data volumes, at speed. Their blog gives a better idea of what they do than their CEO did. http://metamarkets.com/blog/
$15m Series B April 2012 San Francisco, CA
Lucid Imagination Private or public cloud search - 24x7 support for open source Apache
Lucene/Solr search, analytics, big data
$16m Venture, October 2010 Redwood City, CA
Tempo DB Techstars Cloud alumni. Customized time series database as a service for sensor data.
$750K seed May 2012 Chicago, IL
Kaggle Host competitions for data scientists - make existing data available in public to find and qualify scientist, then in private to specific scentists. Competition shows stages of improvement in application of algorithms.
$11m series A Nov 2011 San Francisco, CA
Integrate Advertizing support, ad spend optimization
$11m March 2011 Scotsdale, AZ
InsightsOne Predictive analytics. Real time multichannel user offer matching.
$4.3m venture March 2012 Santa Clara, CA
Huddle Enterprise and government content collaboration platform - big wins in UK, just launching into US.
$24m May 2012 London and San Francisco, CA
Hadapt Add RDBMS to Hadoop nodes so existing SQL tools work. Adaptive analytics platform.
$9.5m series A Oct 2011 Cambridge, MA
Full Contact Impressive, and scary if they can do what they claim. Take partial information about an individual, clean it up and correlate by collecting what they've put out on the net over time. Social profiles, too.
$1.5m series A Denver, CO
Delphix Support developers where lots of copies of the same DB get used. Database virtualization. This is a real problem, and they cite real numbers for money saved and productivity improved.
$25m series C Palo Alto, CA
DataStax Enterprise services and platform using Apache Cassandra open source database.
$11m series B San Mateo, CA
Cloudant Low latency on small low power devices, by getting big data on or close to its consumers. Has 18 data centres worldwide. Data replication and sync using NoSQL database monthly service.
$2m venture round Dec 11 Boston, MA
Bottlenose Low cost real time analytics - see what's trending - process 3000 messages/second - browser frontend.
seed investment (no numbers given) Raising round in 2012
Birst Automating supply chain for low volume high value data. Easy to use business anayltics.
$26m May 2011 series D San Francisco, CA
Apigee API technology and services, for making data available and understanding how it is used.
$14.1m Jan 2010 Palo Alto, CA
Agilone Predictive marketing intelligence. "maximize customer lifetime value at the optimal marketing spend"
Sequoia Capital invested, no numbers given. Mountain View, CA
10gen Commercial support and development for open source MongoDB. Used as an Oracle replacement, gain scability, agility, productivity for developers.
$42m series E May 2012 New York, NY and Palo Alto, CA
Several people were interested in Techstars; TempoDB just emerged from the Techstars Cloud program, which I mentored. Links below for more background - there's another blog post to come on that topic, too.
Last week Silicon Valley Bank saw a lot of me - two afternoons in a row at events they sponsored.
The 'Technology on the move' event in their Tasman Drive offices was worth the time. Good reviews of the silicon systems in cars from Simon Segars, ARM; and Taner Ozcelik, NVIDIA. Their components are used not just for entertainment and navigation, but also for instrument cluster replacement (the 'glass cockpit') and crash and engine simulation systems. NVIDIA's parts are in more than 100 models of car, and ARM's in very many more.
A panel discussion about electric vehicles focused on what it would take to have EVs be the primary household vehicle. Better battery life and higher petrol prices are part of the story - so is an incentive for drivers to understand their actual day to day requirements, rather than needing to own a vehicle which is only used to its full capacity during a long summer vacation drive.
Mark Platshon, who set up BMW's venture fund, ran through a future scenario where his calender was connected to his car, enabling advance reservation of parking spots close to meeting locations and suggestions about proximity of colleagues, friends, and restaurants, to illustrate the sort of things in which i-Ventures is investing.
This wasn't discussed at the meeting - what I really want isn't so much a particular car as Transportation as a Service - being able to drive myself, or be driven, or take public transport, or have things delivered, as something for which I pay a subscription. All the technoloy to do this exists, but the APIs don't. It would take capital, and interaction with very disjoint transport providers - the Bay Area can't even operate a one ticket system for the different (and very poor) rail services. Expect something like this to emerge in one of the Scandinavian cities first.
Didn't attend in person, did listen and watch most of the broadcast material. The quality of the broadcast received was noticably better than for previous Nanogs.
Moore's Law and Networking
Andy Bechtolscheim did the first morning keynote - he's good at presenting a lot of reasonably sophisticated technical material without obviously promoting Arista. If slides are posted, they will be on the agenda page.
Moore's Law will continue to hold through 2024 at least, and probably through 2031. It will soon be possible to buy flash memory chips with 1 terrabyte of capacity. Networking speeds have not kept up with either the increasing density of memory nor the increasing capability of CPUs. The bottleneck is in the offchip communications - throughput per device has increased about 10 times in 12 years.
Manufacturers have been using ASIC workflow for networking chip development - now there's sufficient demand for 10G ethernet components to justify custom design flows, laid out with a specific clock target. Expect reasonable pricing and volumes of 40GE components in 4 years, and 100GE in 8 years.
The other issue holding back mass adoption of 10GE has been the cost of the external optical connections to the fiber used for connecting top of rack switches to the aggregating routers. (Connections from servers to the ToR as short enough for copper). The original IEEE standards didn't support an inexpensive standard for the 100m - 300m distance. Silicon photonics on parallel single mode fiber is changing that - 12 duplex channels in 4.5mm, MTP/MPO multifiber connectors - expect parts commercially available in 2014.
For the 'best' performance, have to decide if the goal is minimising latency, or minimsing packet drops, and consider the details of the TCP stack(s) in use or available. Switching paths through single chips using SRAM minimise latency; chips using less expensive off chip DRAM can have much bigger buffers.
Microsoft datacenter design
Using BGP as an IGP allows for much wider scale out, more reliable operation, than using L2 MLAG. They have some requests for vendor support for specific BGP features :
AS_PATH Multipath Relax
Allow AS In
Fast eBGP Fall-over
Remove Private AS
Building a sustainable broadband network in the city of San Francisco
Tim Pozar described some of the politics and technology involved in having a non-profit build network using City fiber with City oversight. It's still work in progress and they'd like more help.
Software defined WAN at Google
As one of the lightning talks, Ed Crabbe, who talked about Google's use of Openflow at the last Nanog, presented a short form of the talk Urs Hoelzle gave at ONS. They use their own hardware routers, run a minimal embedded OS on them, including BGP and ISIS from Quagga. There is a global broker/scheduler, non standard transport protocols, demand is tailored to be deterministic.
There was a lighting talk, then a full presentation, about the possible consequences of the ITU deliberation due to take place in December 2012. The Internet as it currently operates has many cross border collaborations in place. Proposals from country governments for much more UN regulation are likely to interfere.
Content Delivery Networks
Most of the contributions from this panel were worthwhile listening to. Comcast has a proposal for a new address family for BGP in a single AS, since content caches perform better if they know something about the network they are in. Dave Tempkin from Netflix described the CDN they are making available.
The talk on BGP route origin verification using reverse DNS had the most people passionate about it - it could have done with a lot more time for questions and discussion.
Energy Science 100G network
Substantial network with 100G wide area network links across the US. Good diagrams of the network, and explaining 100GBase-LR4 . Advanced Networking Initiative, ANI. They built deliberate routing loops for testing.
Notable in the last batch of talks were:
Jim Martin from ISC describing progress to date on BIND 10. the first try was "a miserable failure"; it's been tested some more since and is back in production on AS112. Will be ready for user testing in October 2012.
Patrick Gilmore from Akamai with statistics from Akamai's monitoring showing recent (an hour ago) and the gradual build up of IPv6 traffic on World IPv6 Day.
9 Aug 2012 Editing to add - video was posted on the agenda some weeks back - look for the .wmv files.
Heavy Reading are selling a write up on photonic integration - there's a useful list of system and component vendors available without paying for the full report.
Having been to the last 3 years of Engage Invest Exploit, the event run by Informatics Ventures (University of Edinburgh) where the startups which they support present to investors, I couldn't make this year's event. Michael Hayes from RookieOven posted his summary. Since there isn't a handy list of exhibitors with their URLs, I've made one. We've talked to some of these companies on previous visits; we'll review the others and add some of them to the meeting list next time we're in town.
Now that the list for the most recent entrants into the 500 Startups accelerator has been published, it's time to offer congratulations to Scott Allison, CEO and Founder at Teamly. He and I met (via Twitter) when he was still in Glasgow, contemplating moving to London, and he's achieved a lot since then.
He's also contributing regular articles to Forbes - the latest on what it takes to get accepted into the 500 startups accelerator, for nine of the twenty seven new entrants.
Silicon Valley is still the best place to grow a 'eyeball' business - one which has potentially millions of customers, whether consumers or enterprise. You can start that business somewhere else, and probably should, since somewhere else will be cheaper - but if you want to scale it up and to compete with everyone else selling to the English speaking world, the Valley is the place to do that. It has the investors, it has the accelerators, it has the lawyers, it has the people who've done it before, all accessible.
To get to the Valley, the company needs to show some success , and have some resources with which to do that - 500 startups, Techstars,and Ycombinator, are accelerators - you have to have something to validate their investment of money and time in you, and you are competing with a tough, global, crowd.
Well done to Scott and Teamly for having made it this far !
Be awesome at work; stay focused, collaborate better, and celebrate your achievements.
One of the more interesting infrastructure themes currently getting attention and investment is Software Defined Networks. The ambition is for there to be a simple network control plane, with clearly defined APIs, enabling changes in network operations depending on the priorities of the applications currently using the network.
The post collects my notes.
Updated 3 May 2012 to add summary from Google keynote at the Open Networking Summit in April.
IBM white paper (May 2011) describes IBM and NEC Openflow demo at Interop 2011.
"The OpenFlow Protocol allows entries in the Flow Table to be deﬁned by a server external to the switch, which creates the potential to unify server and network device management. For example, a ﬂow could be a TCP connection, all the packets from a particular MAC or IP address, or all packets with the same VLAN tag. Each ﬂow table entry has a speciﬁc action associated with a particular ﬂow, such as forwarding the ﬂow to a given switch port (at line rate), encapsulating and forwarding the ﬂow to a controller for processing, or dropping a ﬂow’s packets (for example, to help prevent denial of service attacks)."
Nanog 54 - the most interesting panel at San Diego (6 Feb 2012) discussed Openflow, standards work in progress, and one approach to implementing SDN.
Quoting Ed Crabbe (Google) on why Google is interested in SDN - it sees an opportunity to improve cost control for their infrastructure while maintaining performance.
common threads: partition resources and control within network elements
minimize network element local control plane
offline control of forwarding state
offline control of network element resource allocation
minimize complexity of software local to network elements
SDN has been around as a concept for a long time. Ipsilon GSMP, 1996; Cambridge's The Tempest, 1998 and so on down the years. Flowspec in 2008 looks like openflow, too.
IETF PCE, 2004, based on having RSVP-TE deployed, ISIS, etc. From network element software side, it's more complex, but it's deployable today.
But why SDN?
comes down to cost. Most of what we do is cost optimization with bounds for performance
make efficient use of resources
network element CPU and memories
underlying network capacity
move heaviest workloads off expensive, relatively slow embedded hardware to fast, cheap, commodity hardware
reduce network complexity and thus operational overhead and outage time
simplify policy composition
enforce correctness constraints and invariants
reduce inter-dependencies between routers, between protocols
reduce complexity of distributed control system software on network elements
implement innovative new techniques (Heller's Elastic Trees, etc.)
Innovation velocity - desirable to speed feature implementation and deployment
Do new protocol development fast; right now, takes too long to get new protocols into devices.
IBM distributed virtual switch
IBM announced a distributed virtual switch (5000V) 14 Feb 2012 supporting SPAN, ERSPAN, Ethernet Virtual bridge, load balancing, ACLS, etc, integrated into VMware. Alex Bennet (Battery Ventures) "This is an important development because as the first tier of network switching moves into the server, the virtual switch becomes extremely strategic real-estate and control point for emerging SDN architectures."
Summary and notes from Urs Hoelzle's keynote at ONS, titled 'Openflow @ Google'
"OpenFlow has helped us improve backbone performance and reduce backbone complexity and cost"
For the WAN, aim is to have cost per bit/second go down with scale, as for CPU and storage; without a lot of systems engineering, that's not what actually happens, since network complexity goes up; also, a 100G bps interface costs a lot more than 10 x 10 gbs interfaces, or 100 x 1G bps.
Google are operating the WAN backbone carrying traffic between data centers with centralized traffic engineering.
Multiple switch chassis in each domain, custom hardware running Linux; Quagga BGP stack, ISIS/IBGP for internal connectivity
They have a simulation environment for testing "with the complete system, end to end. Everything is real except the hardware".
Building a software defined WAN results in higher performance, better fault tolerance, and lower costs.
Conclusion : Openflow and SDN are ready for real world use.