January, 2011


31
Jan 11

GSB Talk on Mozilla and Scaling

Last week I got to attend a class at Stanford Business School taught by one of my favorites, Huggy Rao. The course is on “scaling” — an over-used word, but one that Huggy’s been really digging into lately — resulting in some great insights. This particular class covered a case study authored by Huggy with Bob Sutton on the rise of Mozilla and Firefox, so it was fun to participate in.

Huggy asked me to do a quick 10 minute introduction to the class. I chose to talk about the differences between then and now — how much has changed in the 5.5 years since Firefox’s initial 1.0, and what the new challenges of scaling are. So, naturally, my first comment to the students was that most of the case was irrelevant to today’s world. That Mozilla was amazing and unique and special — for lots of reasons that include (1) breaking the MS/IE monopoly distribution and usage of the browser, (2) doing it in a way that enabled lots of innovation and competition that we’re seeing now, and (3) finding our own way through the journey — not behaving like anyone else in the market ever really has. So that’s cool. In that battle, though, access to users was probably the biggest challenge — it looked impossible when Mozilla started, and it’s remarkable — incredible, really — that we ultimately have gotten the reach we have.

But fast forward to today’s world, where we have more than 600M users on Facebook, more than 400M users of Firefox, and networks like LinkedIn and Twitter with global reach of a hundred million or more. Combine that with the rise of the Apple App Store and mobile devices — with something approaching 200M user accounts that all have credit cards associated with them. (And if there’s any doubt, these numbers are truly huge. I put in some cultural references in my talk — about 100M people will watch the SuperBowl. And only about 20M watch the nightly news in America; 30M listen to NPR. We think of these institutions as huge, but they’re nowhere near Internet scale at this point. The new networks have left them behind, quite handily.)

So now a huge part of the world is accessible, a huge part of the world is ready and able to download an app or click on a shared link. Which means that access is no longer the chief initial obstacle to scaling. That means you can see companies like Zynga or Groupon rise from nothing to massive practically overnight. Clearly, the initial challenge is about rising above the noise of an increasingly crowded field of ways for people to spend their time and money, but it’s very, very possible to get to tens or hundreds of millions of users quickly. Which means that now you’ve got companies that are dealing with huge, complex, global user bases at an extremely early point in their history. My view is that scaling successfully — which means sustaining that scale over time — will be dependent on figuring out how to make the teams and processes in rocket ship organizations operate effectively.

I know not all the analogies in the slides are apples-to-apples, but what’s clear is that we’re living in an era of hyper-distribution, where things can change very, very quickly. I’m really glad that smart people like Huggy and Bob are thinking about how to help us all learn how to manage these in the future.

Fun conversation, thanks to Huggy for the invitation! My few slides are below — they’re very incomplete and mostly served to provoke some interesting discussion. (PS — the deck is sort of a tweener deck graphically between my Mozilla-style slides and what I’ll use here at Greylock — haven’t been here long enough to monkey with the Greylock slides yet. :-))


30
Jan 11

Congrats to SayNow

It’s been a busy January for the companies I know the best — first TripIt was acquired by Concur and last week SayNow announced that it’s been purchased by Google. I was a minor advisor to SayNow, and have been involved since the beginning — since before the beginning, really, and I could not be happier for Nikhyl & Ujjwal — these two entrepreneurs deserve all the success and recognition they’ve gotten and will get, and really deserve a lot more.

They started SayNow with a bunch of intuitions that voice communications were an undervalued and underused part of the web — it’s obvious in hindsight, but they were quite early to that conclusion, nearly 5 years ago now. The road from insight to today was anything but obvious, though. Nik & Ujjwal spent lots of time experimenting, building, listening — really trying to figure out how to make the most useful voice service they could.

What impressed me most about their journey is that they were very often working so hard without the adoration of the TechCrunches and GigaOms of the world — they were working on relatively unsexy problems that really mattered, but maybe weren’t the most tweetable. They pivoted many times, they worked incredibly hard, and they just wouldn’t give up.

I’ve known Nikhyl pretty close to forever — probably since my second year at Stanford — and he’s one of my very closest friends and confidants — we’ve talked about so many challenges and successes over the years that it’s hard to imagine my career (or life) without him. And I’ve known Ujjwal since SayNow’s founding, and always been blown away by his commitment, hard work, intelligence, friendliness, and ability to just get things done.

For the two of them and the whole SayNow team, I’m incredibly, incredibly happy for them, and impressed by them, and I can’t wait to see what they do at Google.


17
Jan 11

Drobo S

I’ve been meaning to pick up a Drobo for our house for quite some time — thanks to Aneel I finally got one last week — and, as I share below, I really, really like it. They’ve done a great job, and it should be a great storage and backup solution for us for a long time to come.

[quick disclosure: Greylock Partners, where I work, is an investor in Drobo.]

First off: Drobo is essentially a big box of hard disks that acts like a single disk. It uses a non-RAID technology called BeyondRAID to spread out data in a way that if 1 disk fails, you can still get to all your data. (There’s also a setting so that you can make it robust to 2 simultaneous drive failures, but that’s not the default, or what I’ve set mine on.)

It also has an extremely nice characteristic that you can always hot swap any drive for a replacement or a bigger drive if you run out of space. (Just another step towards robots being in control — when the Drobo needs more space, it blinks a yellow light at you, effectively saying, “Human, I require more disk — please run down to Fry’s and deposit the new disk beside my blinky light. That is all.” I, for one, welcome our new robot overlords.)

Anyway, back to the Drobo. I chose the Drobo S, which is a 5 bay unit that’s direct attached to a machine via eSATA, Firewire 800 or USB3. I dithered back and forth on whether to choose that one or the Drobo FS, which runs everything over Gigabit Ethernet as a networked file server. I chose the direct attached because we have an iMac that’s pretty much always on, and the vast majority of the data we want to store on the Drobo is from the iMac — and I figured this would be a simpler configuration. I’m still on the fence — I think I would have been just as happy with the FS.

The out of box experience with the Drobo is exceptionally good. The packaging is more like a consumer device, with easy-to-follow 1-2-3 steps on top, clearly marked, and the Drobo itself wrapped in a protective cloth. Essentially, here’s the process:

  1. insert the disks into the Drobo (in my case, 5 1TB Quantums)
  2. install the Drobo software onto the host computer
  3. plug it in
  4. run the Drobo dashboard to specify how you want the partitioning scheme to work, etc

Altogether, took about 10 minutes, and the Drobo was happily running, showing to my iMac as a 16 TB volume (even though I’ve got 5 TB of physical space and more like 3.7 TB of logical space once you factor in the redundancy).

That was really it — it couldn’t have been simpler. No complicated decisions, no significant forward looking capacity planning, no nothing. When the yellow blinky light comes on, I’ll feed it more disk.

To my mind, this is really the only solution there is that’s simple, performant, stable, expandable, redundant, and not incredibly industrial and expensive. It’s quiet (much quieter than I expected for 5 drives running) and doesn’t get hot, and so far has been pretty good about spinning down after disuse. (It could be more aggressive there, but might be pilot error on my part — need to look into it.)

Anyway, I like it a bunch.

Using it for backup with Time Machine

It’s worth talking about my use case here, since there is a minor wrinkle. The reason that I’m so excited & interested in home storage is that our family’s data needs are growing, quickly. We’ve got about 40k digital photos that are getting increasingly large per picture. We’re taking more and more HD video with our Canon 7D. Most everything we watch for SPL that isn’t streamed is ripped from other sources and stored. In fact, in the year or so since we got our iMac with what I thought was a pretty reasonably sized 2 TB disk, we’ve gone from about 600 GB of stuff to 1.3 TB.

It’s harder than you might think to regularly and reliably back up 1.3 TB in your house.

We had a 2 TB external drive that was doing it for a while, but that’s not really big enough once your main data set is over a terabyte, especially with Apple’s Time Machine, since it’s pretty aggressive about making deltas every hour, day, week, etc. So we needed something bigger than that, but the 3 TB drives aren’t too available yet, and I wasn’t really all that keen on replacing that one again in a year or so as our data grew.

Which is why the Drobo is perfect. Just add disks as I want backup to scale.

There is a small problem though, in the interaction between the design of the Drobo and the design of Time Machine. The Drobo just wants to be the biggest disk it can be — up to (at least) 16 TB, so it tells the iMac to just keep giving it data, letting it (and you, human with a car and a credit card) take care of the physical details.

Time Machine is designed to find an external disk (that you select), start putting files on it and then putting new (or changed) files on it again and again and again until it fills up the disk. So that’s a virtue if you’ve got a dedicated disk that’s backing your data up, to a point.

But since the Drobo advertises to the iMac that it’s a 16 TB disk, and Time Machine hasn’t really learned limits yet, Time Machine will happily fill up your physical space, which will cause your yellow Drobo light to come on, which will cause you to install a new disk (and repeat, and repeat), all the way up to the 16 TB that’s the maximum.

So that’s not great. What you really want to be able to do is to tell Time Machine to just use a certain amount of space and then start rewriting over older copies. There’s no way to do that at present in the Time Machine settings (which I feel is silly — with big disks like Drobo, it’s an easy thing to add), so you have to instead make Time Machine think it’s got a smaller disk than it does.

2 ways to do that: (1) partition the Drobo into a logical Time Machine volume, or (2) use sparse bundles. I didn’t want to do the first one, because in the future I might want to expand the Time Machine volume, which would mean repartitioning the Drobo, and losing my data in the process.

But sparse bundles (or, more precisely, sparse images), which I also use to encrypt my work and personal directories on my laptop, do the trick perfectly. With a sparse image, you can set up something that looks like a disk to your Mac, with an optional maximum size that you set, but that only takes up the space that it’s actually using.

So I set up a 3.5 TB sparse image that Time Machine backs up to — as it gets to full, it’ll do the right thing, keep the total backup space used under 3.5 TB, no problem. And I can resize the sparse image at any time in the future from the Terminal.

And that’s working perfectly at this point.

I used an Automator script called Time Tamer to do things pretty automatically for me. And there’s a good discussion of some of this stuff on a couple of blogs — this one was most helpful to me.

Bottom line: the Drobo S is going to be perfect for us, and I couldn’t be happier to have it. I know it’s decidedly nerdy to get excited about a storage solution, but there you have it. ­čÖé


17
Jan 11

Cleopatra, by Stacy Schiff

This biography of one of the most controversial women in history is on a lot of folks’ top 10 lists for 2010, deservedly so. Cleopatra’s life has been chronicled by many — very often writers with political or cultural agendas — so it’s always been tough to get a fair picture of her. Further, her story is almost always told in the context of the history of the end of the Roman Republic — the decades of civil war that resulted in Octavian “restoring the Republic” by becoming Emperor Augustus.

This book is neat because it really focuses on Cleopatra, Egypt, and her Ptolemaic/Hellenistic context, with (principally) Caesar and Marc Antony as the 2 Roman figures who she would have extremely significant relationships with.

I found the beginning of the book to be terrific because it gave me a ton of context of Egyptian & Ptolemaic history that I didn’t really know. The middle was a little bit of a slog, between Caesar’s death and the real hostilities between Antony & Octavian. The end was incredible — a real page turner, even though I knew most of what would ultimately happen.

I’m not sure you’ll love this book if you don’t also really enjoy Roman history, and as I mentioned the middle was a little slow – but overall I really enjoyed it, and it’s significantly changed the way that I view that 20-30 years of world history generally and Roman history specifically.


13
Jan 11

TripIt Congratulations

This morning TripIt announced that it’s signed an agreement to be acquired by Concur, a publicly traded travel & expense company. It’s a fantastic outcome for all involved, and I’m very proud to have been involved as a member of the board of directors over the last several years.

For me, it’s a great example of doing things right. It’s an awesome team that made an awesome product that lots of people can’t live without — simple as that.

I started using the product way before I got involved in the company. I can still remember waking up in Tokyo at 4a suffering from jet lag, going through my news feeds and noticing a financing round in 2008 that Mark and Bryce from OATV had participated in (they did the original round of financing in 2007 — great foresight!). I sent them a short note, saying something like: “Awesome investment, it’s a terrific product. Couldn’t live without it.”

That led to Mark introducing me to Gregg Brockway and Scott Hintz, the founders, and over time we developed a relationship that turned into an outside board seat for me. It has been a wonderful relationship. If you’ve ever met Gregg and Scott, you know what I mean: they’re extremely hard working, smart, humble, thoughtful founders — really everything you could ask for in a founding team. They’re always ready and willing to engage on any aspect of the product or the business; they’ve always put building a great business at the top of the list of concerns. I love working with both Gregg and Scott, full stop.

And they’ve put together a team that is like that as well. The developers, marketers, designers and other managers — everyone’s been committed to building something great. (Andy Denmark, in particular, deserves an award for being patient with me as I nitpicked various things over the years — to his eternal credit, his response was always very positive.) And it includes the other members of the board as well — Mark Jacobsen from OATV and Mike Kwatinetz from Azure — they’ve been amazing and fun to work with, and I can’t imagine the past several months working with anyone else.

So I wanted to say thank to you the whole team — it’s been deeply gratifying to be involved, and I know you’ll continue to do amazing things with a bigger platform like Concur. To my mind, it’s a testament to doing things the right way, and to building products people love.

Onward and congratulations!