At the end of 2013, Salil Deshpande, Managing Director with Bain Capital Ventures and member of the Aria Systems board, popularly made a few predictions in the cloud and virtualization industries for 2014. Among Salil’s predictions were the notions that moving apps to the cloud includes moving your ancillary problems to the cloud and that apps will need to manage their data in memory rather writing them to a disk.
So how did Salil fare with his predictions? He recently took a look back with some interesting results. His full article, which first appeared on VM Blog, is below.
Niels Bohr, one of the greatest thinkers ever and the founder of quantum mechanics, had a sense of humor, once remarking, “prediction is very difficult, especially about the future.”
I don’t feel that I can predict the future any better than my peers; however, it forces me to think about the future and solicit feedback. And, it puts just enough pressure on me to have fun, extrapolate and be wrong, which I enjoy almost as much as being right.
Let’s see how I did. These were my five predictions for 2014. (1) memory is the new disk, (2) I got 99 problems and I brought them all up to the cloud (3) event-oriented non-blocking reactive applications (4) venture firms will raise larger funds in 2014 than they have since 2007 (5) for now Bitcoin is a commodity not a currency.
How did I do?
Memory is the new disk.
A minority of people thought I was saying that Flash SSDs are the new disk, which isn’t saying much, even though flash is the most disruptive thing in storage in many decades and its impact is yet to be fully realized.
No, what I meant is that DRAM is the new disk. And I made two investments (for Bain Capital Ventures) along this theme. One is Redis Labs, the company that’s commercializing the open-source Redis NoSQL database. The other is Hazelcast, a JCache-compatible distributed in-memory computing framework that lets you create complex data structures that are cluster-distributed and cache-coherent.
In-memory computing started out as a special-purpose use case, as caches of disk databases. But DRAM has become cheap enough that it’s now reasonable to throw together a cluster of commodity servers with terabytes of DRAM. This means it’s become a more broadly applicable use case within the last year or two. For example, per datacenter, Twitter holds 105 terabytes in Redis, out of which it serves 39 million queries per second, and it has many datacenters.
In the one year we’ve been invested in Redis Labs, it has gathered 3,600-plus paying customers (many with multi-terabyte databases) for its cloud service, with its on-premise version just released. The January 2015 Wired Magazine article Open Source Keeps Chipping Away at Oracle’s Empire noted that Redis is the fastest growing database, currently #3 in popularity among all NoSQL databases (after MongoDB and Cassandra) and very close to becoming #2. Similarly, one of the biggest ecommerce sites in the world uses Hazelcast on a large number of commodity servers with 256 gigabytes of RAM at multi-terabyte scale with response rates in microseconds via JCache.
I got 99 problems and brought them all to the cloud.
Analysts have been predicting for years that recurring revenue models would become a fundamental means of merchandising in the 21st century.
That time has finally come. I invested in Aria Systems, which is a beneficiary. Prior to cloud apps, only industries like telecom and insurance needed to deal with complex recurring-revenue management. But in the new world everyone needs to. And managing all the permutations and combinations of different pricing tiers, charging models, billing frequencies, product bundles, metering and provisioning is incredibly complex.
Enterprise resource planning systems like SAP ran manufacturing-economy companies that made widgets. In the digital economy, systems like Aria are the new ERP. Seventeen of the Fortune 500 companies launched on the Aria Active Monetization platform in 2014.
Event-oriented non-blocking reactive applications will be ever more important.
In late 2014, Amazon announced AWS Lambda, a distributed compute service that runs code in response to events and manages the compute resources automatically and elastically for you, so you don’t have to think about machines, disks, memory and queue servers.
I made two investments in this theme. One is Typesafe, the company that’s commercializing Scala, Akka and Play, the language and frameworks that promote event-oriented reactive programming. The other is Iron.io, which provides an AWS Lambda-like queuing and execution service on all clouds (not just Amazon) and also on-premise.
Venture firms will raise larger funds in 2014 than they have since 2007.
In fact, the venture industry set a post-dotcom bubble record. Venture Capital Journal reported that U.S. venture firms raised $33 billion for new funds in 2014, the most since 2001. That compares to $29.5 billion raised in 2007, $25 billion in 2008 and $17.6 billion in 2013. The 10 largest funds took in more than 40 percent of the $33 billion. My firm, Bain Capital Ventures, raised $935 million in 2014. VCs were also incredibly active in 2014, investing more than $40 billion, the most since the dotcom days.
Bitcoin is a commodity not a currency (for now).
To be a currency, Bitcoin needs price stability, which it does not have. And it’s still awkward for ordinary people to transact using Bitcoin. But as a store of value and an alternative to fiat currencies conjured up by sovereigns in which one must put one’s faith, Bitcoin continues to strengthen in suitability and it remains a topic of future prediction.
About the Author
Salil Deshpande is a Managing Director at Bain Capital Ventures. Based in Palo Alto, he focuses on infrastructure software and open source. His “greatest hits” include Lending Club (peer to peer lending), SpringSource (application servers), Buddy Media (marketing platform), DataStax (NoSQL databases), MuleSoft (cloud and enterprise integration), and DynaTrace (application performance management); and his current portfolio includes Typesafe (the Scala language; Akka and Play frameworks for Java), Redis Labs (theRedis NoSQL database), Hazelcast (in-memory data-grids for Java), Aria Systems (ERP for recurring revenue businesses), Iron.io(queues and workers in the cloud), Concurrent (Cascading and Driven), and ZeroTurnaround (faster Java redeployments).
Note: This article first appeared on VMblog.com.