wrong tool

You are finite. Zathras is finite. This is wrong tool.

All of this has happened before and all of this will happen again

by 7 Comments

One of my favorite shows on television was battle star galactica, and it’s theme that the past and future were tied forever in a cycle. As someone who has lived through three busts, the feeling of being a cylon trapped in a cycle of life and death is … familiar. And in each case, the bubble was caused by a drop in interest rates and in each case the end of the bubble came with an increase in interest rates.

Right now traffic is as bad as it was in 1999, in spite of the tremendous infrastructure improvements that were put into place since then. And there is this feeling that this guy wouldn’t be out of place here:

pets_com
Without any shadow of any doubt the total amount of investment in the bay area has gone crazy. The question is when will it end and was it correlated with an interest rate drop…

And then this shows up (my own annotations):

2015-01-12_1419

What the chart shows is how between the start of QE in 2008 and the end of QE the total aggregate investment in Seed rounds grew from under 50ish billion to over 500 billion.

What the chart also shows is how with the end of QE in Q3 of 2014, the total amount of money in seed rounds has collapsed from over 500 billion to less than 200 billion. Twitter commentary from the VC community suggests that there is basis in fact.

Turns out we had a bubble begin with a drop in interest rates (QE Start) and and end with an increase in interest rates (QE End).

Speaking from my own narrow field of view, in early 2014 people were running around getting funded, in late 2014 I heard a lot fewer stories and recruiting pitches to go join the new new thing…

What does this mean? Assume the trend holds…

First of all it means that there is going to be, once again, much better traffic for those who survive the wave after wave of layoff. Many companies will fail and the never ending and increasing investment dollars that allowed for endless seed rounds will disappear creating a large overhang of employees who will, unfortunately, move on to other pastures. Secondly it means if you are in one of those seed companies, you better be figuring out how to get to sustainability faster … and finally it means that the quality of investments will pick up….

Oh and housing prices will drop again…

For folks who have not been here in a similar bust the experience will be traumatic as the go-go atmosphere evaporates. For those of us with more grey hairs, we are reminded of the saying: This time it will not be different….

 

 

 

The problem with image backup services like iCloud

by Leave a Comment

At NetApp in 2008 while we were struggling to define feature set of ONTAP 8, a product manager asked me if it was okay to ship backup in 8.0 and restore in 8.1. He quickly realized that this wasn’t going to work…

Unfortunately Apple has made a similar mistake with iCloud.

The problem with iCloud is that it is an image backup service. An image backup copies every disk block and restores every disk block. An alternative to an image backup is a file backup that restores every file. An image backup has the nice property that it is faster to restore than a file backup. The problem with an image backup is that it is a an all or nothing restore. If a part of the image has a corrupt file, then the backup is useless.

For example, if you have a firmware image in the iCloud backup that’s corrupt, then restoring the backup from iCloud restores the phone into a corrupted state.

And that is a problem. Turns out if you have let’s say 5 years of photos in those backups, there is no way to restore the pictures because the image contains the corrupt firmware.

This isn’t impossible to fix. Many backup apps that do image backup are also able to do file restores. Unfortunately Apple doesn’t offer that service.

Fortunately third party restore apps do!

The real lesson is that if you are using an iPhone, please use a service like dropbox or google drive or one drive. In this manner if the backup is corrupt you don’t lose your data or spend money on someone tool.

 

Because some 1’s are 0’s and some 0’s are 1’s

by 1 Comment

When I was an aspiring software engineer at my first job at SGI in 1996, I asked my friend and mentor why a specific kernel debugger wasn’t working.

He gave me an answer, but even then my unwillingness to be satisfied with a curt answer was – well – annoying. He looked at me as I tried to get a more interesting answer than just – well the configuration didn’t work and realized that he needed to get on with his life than answer one more annoying question from this freshly minted new grad.

His response to my: well okay but really what was going what was the fundamental issue?

His answer was legendary:

Because some of the 1’s that should have been 1’s were 0’s and some of the 0’s that should have been 0’s were 1’s

I love the answer. It was a brush off, but at the same time was deeply insightful. At the end of the day the reason software doesn’t work can be reduced to a string of 1’s and 0’s that aren’t correct.

And many years later my son repeated this to me when I asked his mom why the database wasn’t working:

Apparently some sayings endure.

It wasn’t me, it was my neighbor

by 1 Comment

Keith Adams of Facebook pointed me to this paper:

Flipping Bits in Memory Without Accessing Them:
An Experimental Study of DRAM Disturbance Errors

The fun part is that we have a new and exciting way to induce memory corruption. Reading memory.

The interesting part is the proposal of a probabilistic algorithm to address the issue.

This continues my enduring belief that reliance on reliable hardware to make software work is increasingly a fools errand. As we patch more and more of the cracks, eventually we will have to stare at the chasms and rethink software design.

In the meantime, the next time I get a memory corruption I am pointing my boss to this paper.

 

Looking back on “Here Comes Another Bubble”

by 1 Comment

In 2007 the Richter Scales captured the zeitgeist of the Silicon Valley bubble atmosphere with this hysterical video:

The irony about this song was that the 2007 bubble was killed by Lehman’s. A lesson in mass hysteria.

And as I was re-watching it last night I wondered – so what happened with those scrolling startups? And was Peter Thiel right or wrong?

bubble-3

When it comes to Facebook, he was probably right. In fact, the valuation of Facebook is staggering and an amazing achievement for the folks at that company!

That got me wondering about the rest of the companies … The hype around MySpace wasn’t probably justified:

myspace

And then if you look at the scrolling startups, I tried to circle the ones I knew were still in business. Admittedly I didn’t look through every single one — there are a lot but was struck by how few of them are around:

bubble-2 bubble

Starting a company is hard. Creating investor value is hard.

And sometimes it’s worth remembering that for ever Facebook there is a scrolling morass of companies that go nowhere.

And like the song says, those folks are probably back at work looking to start another new thing…

 

 

Regulation or Encryption for Big Data

by Leave a Comment

A few weeks ago I called for a Hippocratic Oath for data scientists.

Now the New York Times is calling for regulation.

We can fight regulation all we want and we will.

But…

Customers will fight us. And they will fight us by moving to vendors that offer privacy and encryption. Where data is wholly owned by the customer.

And the customer will win. Because if we don’t act like their data is a sacred trust they will learn to distrust anything that they cannot control.

And they will control it through encryption.

And then we will have the medical industry where breakthroughs are there if only we could look at data …

What’s your employer’s security policy for your ID?

by 1 Comment

After reading about the attack on Sony Entertainment, I am thinking that instead of worrying about hackers stealing from your bank we should be worried about hackers stealing from major corporations.

And that isn’t really surprising.

If identity is the seed of theft, then as banks and retailers get marginally better the soft squishy underbelly of HR departments becomes an inviting and easier target.

To date, a company cares about their business secrets being stolen, followed by their customer data being stolen and ending somewhere below not-at-all their employee’s private data being stolen. Employees don’t quit their job if their employer loses their data… At least not yet.

Identities are simply too valuable an asset for the bad guys to not want to steal, and unlike retailers where you can choose to not give them something that’s worth stealing your employer must have everything…

Me-thinks there is an opportunity for an innovation where companies can use tokens from some really secure system instead of the real raw data that they are just simply incapable of securing.

This Sony attack isn’t the end it’s the beginning.

 

Hardware not Software is Eating the World: Refinements…

by Leave a Comment

After I wrote my comments about hardware, a bunch of my software friends reacted …

And it got me thinking. My argument is a little bit more subtle than I thought.

Without a shadow of a doubt – A (capital A – no typo) disruptive product is the combination of software and hardware. That almost seams silly to have to write … After all is hardware with no software anything but inert rare earths?

Let me add some refinement.

1. I was really talking about as an innovator where do you go looking to go create the next disruptive product

My point was not that software is irrelevant, my point was that disruptive products emerge where new hardware is not where old hardware was. And that the new hardware creates new kinds of software systems that are very different than the old ones, and the combination of the new hardware and the new software is what displaces old software.

If you are looking to CREATE new disruptive technologies, the most important thing to be looking at is hardware not software. The hardware will create the opportunity to create new products that may have been previously imagined but impractical to build or sell.

2. It’s about utilization

When I say utilization I mean how much of the true capabilities of the hardware are being wasted versus being fully utilized. For example, the smart phone in 2007 was underutilized compared to it’s true potential leaving 100’s of billions of dollars of unrealized value in innovation.

Innovation really happens when the existing software systems are unable to take advantage of the new hardware capabilities creating an opening for a new software system to leverage the hardware differently. Typically what happens is that the old software made certain assumptions about the hardware that changed and a different software system making different assumptions is able to create new experiences that are profoundly different than the old ones.

What has happened, typically, is that the hardware vendors ability to create new hardware has exceeded the pace of innovation in software of the incumbent players creating an opening for new vendors.

3. Physics begets hardware begets software that is not math.

A lot of the new hardware that we are seeing come into existence is the result of new deeper understandings of the materials we are using.  Ultimately that science into materials is what is creating the opportunity for disruption. Hardware is the – in this vast oversimplified narrative of the world – first order derivative not software.

And yes I know some believe math begets physics, but I also know that software is not math.

In short

Clearly any novel and new disruptive product is a combination of hardware and software … My personal bias’ when looking FOR disruptive products is that they must be hiding in places where new hardware is being underutilized and the product must have software that is making trade-offs that are inherently incorrect for the old hardware. And if the new product succeeds it will do what seems impossible – kill old software.

Because hardware comes first, software later in this tale – and that the triumph of the new system is to kill the old software then my original point of view remains – new hardware is the place to go looking for when you want to create disruptions and therefore hardware not software is eating the world because it is allowing the world to be radically changed.

And I know Mr. Andreesen meant something else.