Brainiac Corner with Amrith Kumar, Founder and CTO at ParElastic

Posted by VividCortex on Nov 5, 2013 6:10:00 AM

At the Brainiac Corner, we meet with some of the sharpest minds in the system, database, devops, and IT world. If you’d like to share your thoughts on pirates, ninjas, the future of system administration, or any other relevant topic, please don’t hesitate to contact us.

This week we are lucky to hear from Amrith Kumar, a Founder and CTO at ParElastic, a venture-backed, pre-built MySQL sharding solution. We’ve gotten to know Amrith through the conference circuit and were even more excited to share his thoughts with our readers.

Brainiac_Corner_with_Amrith_Kumar

How did you get from stork to brainiac (i.e. what do you do today and how did you get there)?

I’ve always been fascinated by computers and things like this. The first one I used was called an SDK-85. It had 256 bytes of RAM, a 2KB ROM a couple of analog ports and maybe a digital port. It ran at 3Mhz. It was obviously a great improvement to move up to a TS-1000 (the American version of the ZX-81) which ran at 8Mhz and had a QWERTY keyboard. This also meant we got a television and a tape-recorder ☺

I never studied computer science because it seemed like too much work to get into one of those courses; so instead I studied math which was relatively easy. Then, after getting an MBA, I started working for CitiBank in India where I found that crashing the Stratus Computer we worked on was more fun than writing software that computed your mortgage interest or balanced your checkbook.

Which meant that in ‘95 I was hired by Stratus (I firmly believe they did just so they eliminate their support nightmares). Two years later they decided to move me to headquarters in Marlboro, MA where I got to work on two of their proprietary operating system kernels, and briefly the windows kernel.

In ‘02 I joined a company called Netezza that was building a cool distributed parallel database for data warehousing and I was really taken by the whole MPP/parallel database/shared-nothing concept.

In ‘09 (and this continues today) I was shocked to see how much hype was being generated by some crackpot solutions of highly dubious value in the NoSQL space, and I decided to build a more traditional solution. Which brings me to ParElastic, we make traditional relational databases both elastic and horizontally scalable.

What is in your group’s technology stack?

Our product is today written entirely in Java for portability and speed. We build on the shoulders of standard relational databases like MySQL and operate in a variety of environments. We are deployed on a variety of public clouds, and in some large enterprise data centers.

Standard relational databases are at the core of our product offering and the Database Virtualization Engine ™ that we’ve developed serves to virtualize a group of independent database instances and make them appear to be a single virtual instance.

Who would win in a fight between ninjas and pirates? Why?

The ninjas will win many of the battles, but they will lose the war. The ninja is stealthy, plans well, dresses sharply, and generally is fighting to defend something. In contrast, the pirate is loud, has a rough idea of what he wants and tries to get it by brute force, is dressed like a slob (mostly) and is fighting with nothing to lose but his (or her) own life. Yes, the ninja wins the battles, and fewer ninja are killed in skirmishes. But in the long view the pirates win, and they win big. But they take more losses. More importantly, the pirates have more fun.

Ninja’s are the big company types, pirates are the entrepreneurs and start-up types ☺ which would you be cheering?

Which is a more accurate state of the world, #monitoringsucks or #monitoringlove?

Today I see a world more in the #monitoringsucks camp because monitoring is mostly analogous to a historian accurately recording that which happened. But there are a few people moving this closer to #monitoringlove by making monitoring predictive, by making monitoring drive self-diagnosing and self-healing systems, and by making monitoring something more than a post-mortem.

Yes, I know my MySQL server is down, thank you so much for that insightful piece of reporting. That is #monitoringsucks.

But if my system could look further and identify that the server is down but the instance on which it was supposed to run is still working, and that the volume on which the “data” should be stored is mounted but unresponsive, and that there are error messages indicating that the problem relates to an unresponsive volume, that’s beginning to look more like #monitoringlove.

What if I got a text message indicating what the problem was, and gave me three options to fix it and all I had to do was to reply 1, 2 or 3 to the text message and it would be done for me? That would be #monitoringnirvana!

In six words or less, what are the biggest challenges your organization faces?

Making databases “virtual” is hard stuff!

What’s the best piece of advice you’ve ever received?

These things collectively represent the best advice that I’ve received over the past several years.
- Life is too short to not be having fun.
- Iterate quickly, and with definite purpose.
- Time is the most precious commodity for an entrepreneur.
- True learning comes from mistakes, hopefully others mistakes but maybe your own; If you must make them, try very hard and to only make new ones.

What principles guide your expertise in your given domain?

The world around us is changing very rapidly, some say at an increasingly rapid rate! There is noise all around, finding the signal within it is hard. Just reading about it is not sufficient to remain current or even relevant any longer; it takes continuous practice. We all know the truth in that statement when it comes to sportspeople or doctors but it is just as relevant in the area of high technology. In any technology profession today, to remain relevant and retain a current skillset, one must continually update their knowledge by reading, experimenting, and most importantly making mistakes.

True learning only comes from making mistakes. If there’s one thing I know it is that when you stop making mistakes you are by definition stagnant. Try and learn from other people’s mistakes where you can but if you must, don’t make the same mistake over and over again.

I was lucky to have worked for some great managers who gave me the latitude to make many mistakes (and boy did some of them regret it ☺ ), and today I continue (on a regular basis) to make mistakes and hopefully continue learn from them!

What is your vision for system administration in 5 years?

The system administration and devops roles today are still too heavily dependent on human beings doing repetitive tasks and attempting to diagnose and correct failures manually. This is an un-scalable situation as systems are becoming ever more complex.

Furthermore, human beings are terrible at repeatedly doing the same action, error rates are surprisingly high. Computers on the other hand are very good at reliably doing the same action over and over again without generating an error (unless the human chooses to come in the way that is).

We should strive to build systems that can diagnose most failures and either correct them automatically, or at worst present the human “administrator” with a set of possible options with the associated “costs” and allow the human to make the choice but not actually the act of implementing the action!

IBM’s demonstrations of Watson have wowed all of us but at the end of the day all it does is have the ability to very quickly search enormously large volumes of data and provide a set of options with associated probabilities. Yes, it can be programmed to act on the option with highest probability if this probability meets some threshold and the resulting apparatus has been shown to beat a chess grandmaster and other Jeopardy contestants.

That’s where system administration needs to head, in a hurry!

Recent Posts

Posts by Topic

see all