Laserlike

Every so often a new technology emerges that changes everything.  In the world of storage, the last major shift was the move from tape to spinning media (“disk”).  While tape is still around today primarily as a target for backups and archiving (it’s cheap, durable, and portable for offsite storage), disk owns the vast majority of primary storage.

From a capital expense perspective (“CAPEX”, the cost of acquiring a unit of storage), disk is more expensive than tape.  Tape is also cheaper from an operating expense perspective (“OPEX”, the maintenance cost including power, space, and labor) — check out this recent article on “Tape is Still the Greenest Storage Around.”  But from a cost per unit of input/output (“I/O”), there is no contest.  Anything that requires fairly frequent and low latency access to storage has moved to disk.  Thanks to spinning media, a new set of applications was enabled.  Cost per I/O won.

Today we find ourselves on the cusp of a transition in storage as massive as the move from tape to disk — the move from spinning disk to solid state disk (SSD, also popularly known as flash memory).  While it may appear that the transition from disk to SSD will mean higher CAPEX on a cost per unit of storage basis, that may be misleading.  Amdahl’s Law helps us to understand the maximum improvement to a system when only part of the system is improved.  Mainframe’s leveraged Amdahl’s Law to maintain balance in a mainframe system between the various components (compute, storage, and so on) to optimize overall system performance.  Throughout the evolution of open systems this sense of balance has been lost.  While Moore’s Law has been applicable to storage in terms of a cost per unit of storage, storage I/O has not kept pace with compute I/O.  The system is out of balance.  Solid state disk will help us regain balance between storage and CPU, so the overall CAPEX per I/O may well decrease substantially.

The cost of OPEX will also decrease as SSD replaces disk.  Power consumption is lower with SSD than spinning media, so electricity costs per I/O will also drop.  Better I/O will reduce the number of systems that need to be managed and no moving parts should increase reliability, so labor costs related to maintenance should decrease.

Just as with the last storage discontinuity, cost per unit of I/O will be much lower for SSD relative to spinning media.  Consequently I believe that, in the coming 5-10 years, SSD will replace disk across the board.  Cost per I/O will win again.

And like the last move, a new set of applications will be enabled.  Already consumer devices like laptops, iPods, and the iPhone have put SSD to good use.  It’s just the tip of the iceberg.  There are a few huge opportunities that come to mind.  I would love to hear your thoughts on these as well as others I have left out.

1.  Enterprise storage.

The next EMC and NetApp will likely be built off the back of the move to SSD.  For the past year I’ve been looking at anything and everything in this space.  The incumbents will have difficulty changing their model to address the new world — they are introducing hybrid solutions into the market, which is what you would expect.  But their software and systems were designed with assumptions that are no longer appropriate in the new world.  There is room for at least one massive pure-play SSD enterprise storage vendor.

2.  Enterprise applications.

Wall Street notoriously had firms that spent upwards of $1 billion per year in IT — each!  While much of that expense was labor, many enterprise IT startups were funded by Wall Street IT buying as much as by venture capital.  And much of that was for extremely fast, super reliable systems.  In-memory databases, low latency messaging systems (TIBCO), and super fast storage software (VERITAS, EMC) were, and I suspect still are, a nice chunk of those IT dollars.

No doubt that financial services firms will be early SSD adopters.  Even more interesting opportunities will exist at the next tier of applications and firms that will be able to do things previously unimaginable.  Will SSD empower real-time analytics, supply chain management and financial reporting?  How will dramatically better I/O change the way CRM applications work?

On a more pedestrian note, why do companies have tens and sometimes hundreds of Exchange servers?  With dramatic improvements in I/O thanks to SSD, might companies be able to reduce their mail servers from many to just a few?  Will SSD empower MySQL and others to compete with Oracle at higher and higher levels of performance thanks simply to much better storage I/O?

3.  Consumer applications.

The largest consumer internet players spend vast sums of capital on in-memory solutions and on operating costs associated with massive farms of spinning media.  Delivering results to a search query, placing the right advertisement in front of the right consumer at the right time, or building the perfect page (News Feed?) are I/O intensive applications operating on gargantuan data stores.  SSD has the potential to collapse the memory hierarchy, at least for many applications at most firms.

What happens when an application that depends on some batch processes today (alerts, people who like this like that, matching on dating and jobs sites, behavioral targeting) can be turned into real-time processes?   I haven’t seen examples of these things yet, but I certainly expect to see next generation consumer web applications written with the assumption of dramatically better I/O characteristics thanks to SSD.

While the world is suffering from a nasty recession, innovation marches on.  Solid-state drives will help us regain balance between compute and storage, which will lead to lower overall costs and usher in a new class of applications.  And this will lead to some very interesting startup opportunities.

*I would like to thank John Colgrove for reading a draft of this note and providing his feedback.  John is an EIR at Sutter Hill Ventures, previously served as a Symantec Fellow and VP of Technology Strategy for the company’s Data Center Management group, and was a founding engineer of VERITAS Software which was acquired by Symantec.

People respond to incentives.  

These four words are at the heart of the social sciences and the key to hacking human nature.  Mankind has developed positive incentives and negative incentives.  Positive incentives include fame, praise, respect, degrees, badges, medals, [good] grades, points, and various monetary rewards (which can be traded for most other goods).  Negative incentives include prison, the dunce hat, [bad] grades, gossip, ridicule, physical intimidation / harm, and monetary loss (lawsuits, depreciation of assets in disrepair, etc.).  It’s important to remember that “incentive” does not necessarily mean “money.”

In the early days of the internet a small community of academics relied on community norms to govern the system.  The systems governing the exchange of email were limited.  What incentive would one academic have to spam another?  The same goes for usenet, which was a very open message board with limited governance.  The first generation of web search engines made assumptions that publishers of content on the web would accurately tag their own sites so that search crawlers could help index the web.  Who knows better than the content creator?  In small communities like the early internet, the cost of being a bad actor were high — transparency and the fact that it was a small community were powerful incentives.  

But then email became pervasive outside of academia and SMTP was (and still is) ill equipped to deal with a world with bad actors — even a small percentage of spammers can ruin the experience for everyone else.  Usenet was quickly overrun by spam as the internet went commercial.  And the idea of having publishers tag their own content for better search and indexing failed once incentives to cheat grew and the cost of doing so decreased.

While most people think about incentives on the web being monetary things like About.com paying their guides or virtual currency in name-your-favorite Facebook game, understanding incentives has been at the core of many of the web’s greatest inventions.  And I suspect that will continue to be the case.  

1.  Google’s PageRank.

Google Search “won” by hacking human nature to fight the spam problem.  The insight behind PageRank is that you can figure out what a site is about and its relative value by what other people say about the page rather than the publisher herself.  Publishers speak with their links (and the content around those links).  Not only do you need high quality sites linking to you, but you also get punished for linking to bad content.  Brilliant.

Crappy content fades away and bad actors live under the threat of being removed from Google’s index.  A potential benefit of having a monopolist in the search market is that the threat of being removed from Google’s index could literally lead to bankruptcy for many potential spammers (not the case in a fragmented world of search providers).  That’s a painful pill to swallow and one that arguably leads to a better search experience for all consumers.  

2.  Facebook’s domain-based authentication.

Early social networks had many of the same issues [and still do] as usenet.  Consequently, users [particularly women] were often afraid to share authentic details about themselves.  

Until 2006, the only way to get into Facebook was with a university domain (e.g., name@harvard.edu).  Domain-based user authentication not only limited fraud, but allowed Facebook to automate the creation of groups by University (and later corporation, etc.).  Furthermore, the cost of being a bad actor surely carried greater risk than having some lightly authenticated name@hotmail.com account.  

What other mechanisms of authentication can improve authenticity?  How about sending physical letters with a unique code to every home on the planet to create a location-authenticated social network?

3.  Less is more with IM buddy lists.

Of all of my online collections of friends, the smallest by far is my IM buddy list.  There is a huge cost of interruption.  With a small list of people, I can leave Adium running in the background whenever my computer is on.  While I’m fairly open to receiving and making invites on other social networks, I cultivate my IM buddy list like no other.  LinkedIn smartly suggests that you only invite and receive invites from people you know.  But instant messenger applications needn’t make such suggestions, as users quickly learn the costs of too many or unknown friends.  I get more email spam in a day than IM spam in a year.

Good entrepreneurs understand how to manage risk, often implicitly rather than explicitly.  Risk can be a great friend to the entrepreneur.  Risk leads to fear, often leaving arbitrage opportunities all over the place.  Startups are the stealth vehicles of the business world with respect to risk.

YouTube took massive legal risk that others like Google wouldn’t, and arguably couldn’t, take.  Until they bought YouTube, of course.  Many very talented people experience early success in their careers.  As they find more success, the “risk” of a startup grows in lockstep with their success.  So startups can also be seen as an arbitrage on tenure, in a sense.  Mark Zuckerberg is a very talented guy.  If he had joined McKinsey & Company or Goldman, Sachs like many of his peers probably did, he would probably would have done very well.  And he would likely be working as an Associate at one of those places or in an MBA program at Harvard or Stanford right now instead of serving as C.E.O. of one of the world’s most successful web companies.

I recently had dinner with a successful entrepreneur.  When I asked him about his motivations for his new project, he said that he wants to be more successful this time than last time.  He told me that he’s worried about failing this time, which wasn’t the case the first time.  It’s surprisingly how often I hear that.  His comment reminded me of something Steve Jobs said in his 2005 Stanford Commencement speech,

I didn’t see it then, but it turned out that getting fired from Apple was the best thing that could have ever happened to me. The heaviness of being successful was replaced by the lightness of being a beginner again, less sure about everything. It freed me to enter one of the most creative periods of my life.

Risk can lead to fear.  Fear is useful as a survival mechanism in a life or death situation, but it’s not the entrepreneur’s friend.  So how can a startup manage risk intelligently?

1.  Market risk.

The single best strategy to mitigate startup risk is to make some massive market your sandbox.  Warren Buffett makes sure he gets a valuation that provides a margin of error (rumored to be at least 25% — a huge feat in the public markets).  The best “source” of margin of error for the technology entrepreneur and investor is a gargantuan market.

In the consumer market it’s often hard to define the market — this is especially true with advertising supported businesses.  While an entrepreneur working on a new ad-supported sports site may study the size of the “sports” industry, is that really relevant?  A lead generated from a sports site can be a perfect substitute for a lead from a social network, for example.

In the enterprise market, entrepreneurs often focus on some new set of features that some new emerging technology will enable.  This approach is exciting.  It’s fun.  And it’s often bad business.  Don’t invent the problem, invent the solution.

2.  Product risk.

If the market is the “problem,” then the product is the “solution.”  An approach that has been used frequently in the consumer market with great success is doing more with less.  The cognitive overhead of an incumbent’s feature rich product can often be replaced by a very focused, fast, and effective single-minded product from an upstart.  Good examples of this include Google’s single-minded homepage, Facebook’s focus on social networking [initially] just for college students, and Twitter’s very narrow focus on <= 140 character status updates.

The same approach would be a welcome addition to the enterprise market.  Enterprise products are typically a cumulative mess of feature requests from customers over many, many years.  Market requirements documents stuff hundreds of features into every release, often with individual features being added to please one big customer.  Why not take a fresh look at huge existing enterprise products?  What are the few features that a big chunk of the market really needs?  How can we focus on just those features and delight customers?  In addition to removing all of the crud that just gets in the way for most customers (and the development team, customer support, sales, etc.), focus will allow for innovation in the right places.  That, taken with a fresh approach based on new technologies that have emerged since the incumbent product started 5, 10, 15, or even 20 years ago makes for a very soft target.

Finally, the one feature that is at the top of everyone’s list is price.  If you can offer more for less, that’s interesting to everyone.  Open Source software has done best not in creating new markets, but rather by offering a free alternative in huge existing markets (Linux, JBOSS, MySQL, etc.).  Software as a service can lower installation, maintenance, and training as Salesforce has demonstrated.

If you can nail the few features that matter and offer your product at a price under your competitor’s cost basis, that’s a strategy.

3.  Execution risk.

This is the risk that venture capitalists love.  A great team will almost certainly create something of value in a big market.  You can control execution risk more than any other risk.

I thought about breaking out technical risk from business risk (go-to-market, legal, and the like), but technical risk isn’t what it once was.  Older venture capitalists and executives speak about the good old days when you really didn’t know if a team could build what they set out to accomplish.  In the world of IT today, that’s extremely rare.  In most cases, a great team can deliver the goods as long as the right product is being built for the right market.

Legal risk varies from business to business, but is more often a source of competitive advantage than anything else.  Just make sure that you understand the scope of the risk before you decide to arbitrage the legal system, as many music startups have learned the hard way.

4.  Financing risk.

In today’s environment this is a risk with which everyone is well aware.  But financing risk is always a risk for an early stage startup — just more so during times like these.  Financing risk can be mitigated by limiting market, product, and execution risk.  In particular, a great C.E.O. can often convince investors to have a little faith.  In subsequent rounds a great VC is not just a source of further capital, but can also encourage others to join the effort.

Having said that, investors have learned just how hard it is to get user traction with consumer products, so even a great C.E.O. may have trouble raising money without a product in the market with good usage traction.  And it gets harder and harder as your product is in the market longer and longer without traction.

I am sure that there are plenty of other risks out there to think about that I have failed to mention.  Viewed as an asset rather than a liability, risk is a powerful weapon.  In particular, have the courage to never let fear cloud your judgement.  Risk is just another asset with its benefits and its costs.