Original URL: https://www.theregister.co.uk/2014/01/21/diablo_technologies_ceo_talks_of_san_bottlenecks/

Diablo Technologies’ boss: 'As a CEO I have made a lot of mistakes'

Willem ter Harmsel chats with Ricardo Badalone about SAN bottlenecks

By Willem ter Harmsel

Posted in Storage, 21st January 2014 23:02 GMT

With IBM adopting Diablo Technologies' Memory Channel Storage (MCS) for its gen 6 X-servers, we thought it timely to have a look at the background to this through a Willem ter Harmsel interview with Ricardo Badalone, Diablo Technologies' CEO.

Read on to find out about Badalone's thinking on SAN bottlenecks and a move to distributed storage architectures.

I had a great Skype call with Riccardo Badalone of Diablo Technologies … I heard about Diablo Technologies recently and found myself very interested in the background of this company.

Riccardo started the company a decade ago in 2003; he co-founded it with Franco Forlini and Michael Parziale. Diablo is based in Ottowa, Canada.

WtH: On a personal note, what has been your personal drive to start Diablo?

Ricardo Badalone: I am hugely competitive, I want to win badly. Not at any cost, but I do feel very competitive. Thinking and speaking about our competitors actually changes my body language.

What really drives me is the idea of our technology being used by global companies that are number one in their market. Allowing them to use it and profit from it. Financial rewards are important too, but is not my main drive.

Ricardo Badalone

Ricardo Badalone, CEO of Diablo Technologies.

WtH: The success seems recent, has it been a rough road?

Ricardo Badalone: It has been a rough road and as a CEO I have made a lot of mistakes. What we did do correctly is that we transitioned away from markets where we were easily replaced and moved aggressively into markets where we have a unique technology.

I have grown from a world where I worked at a low level of things. The concept of what a disruptive technology is, has grown over time. I have learned that as you move up the stack, the impact you have on the market and financial rewards increase exponentially. That evolution has been the biggest lesson, I would never move back into a purely silicon based technology again.

We went through two economic two downturns where it was hard to stay profitable. There have been many times where Diablo nearly had to close the doors, where the stamina of our team and our investors kept us afloat.

All these hard times have given us an experienced, battle-scarred board of directors and management team. You find that through the bad times you develop a culture where you are not allowed to quit. Through experience, the fabric of your company tightens and you realise there is a very close bond, it becomes a family.

WtH: How do you see the future of storage?

Ricardo Badalone: As a general development, I see that people want to increase the amount of work done per server. Storage does not create work.

In the short/middle term, server-side flash will be on the rise. Creating a flashpool at the server level does three important things; First you can reduce the memory footprint required per virtual machine/run more VMs. Second, all of the VMware metadata related to the management of the guests are now stored local to the machines – and does not have to go onto the SAN. Third, all of the IO that a guest is trying to read from the SAN, is now serviced from a cache that has an 80 per cent to 90 per cent hit rate. These three effects combined allow customers to scale up the amount of work done by the servers, while leveraging their existing investments in their SANs.

In the long term, we will move away from the SAN and toward distributed storage architectures. Typically SANs have been the bottleneck for the reason that people have wanted stateless servers. If a system goes down, you want to be able to seamlessly move your machine over to another machine and not lose any data.

VMware, IBM and others have been building distributed SANs that will eventually displace existing physical SANs. We will move to a distributed storage architecture that Google has been doing for over a decade. Distributed architecture takes advantage of scale and allows for work to take place on the machine where the data resides. You stop moving around data and start moving around work. Added to that, distributed architecture allows for linear scaling. Companies like Nutanix and Simplivity and VMware with vSAN are already doing this, which I think is brilliant.

WtH: How do you feel about the future of X-86?

Ricardo Badalone: I believe many applications would really benefit from purely in-memory computing. The driving force behind is that many customers would greatly benefit from real-time analytics. Increased throughput with the same latency will enable customers to process more data real-time.

Through new memory technologies evolving from what Diablo is currently offering with MCS, applications will be able to map IO structures directly into different memory regions. Operating systems like Windows, Linux and especially VMware will be able to identify different types of memory devices as a different classes of memory. They will be able to directly write to a special persistent memory layer, as if writing to RAM with Byte-addressable transactions. MCS is the start of that new development.

WtH: What are your views on storage companies moving toward 100 per cent focus on software?

Ricardo Badalone: I believe that me-too hardware producers need to be careful. For storage companies with strong software skills, a move away from hardware to focus fully on software is in some cases brilliant. Even Diablo might in the future consider to move away from hardware and move more into software if we see that our hardware becomes commoditised.

On the other hand, a move toward 100 per cent software focus is very bold, because companies like VMware, EMC and Microsoft are aggressively drawing functionalities into their hypervisors and software stacks. If your goal of starting your company is to be bought by EMC or VMware, that is no threat – but it is very hard to build and sustain an autonomous software company in the storage space. ®