The biggest thrill of working in the tech industry was inventing with my co-workers and delivering products that helped our customers solve their business needs. Here is a list of those products from beginning to end.
- High Availability Disk Array (HADA). The team that I worked on came up with an architectural design before I formally joined the team. I was straight out of college at the University of New Hampshire. My assignment was to implement RAID5 algorithms (introduced by Berkeley in 1986), write the algorithms that performed hot repair of disk drives, and handle all of the data integrity issues that might result from disk failures, controller failures, and power failures. The failure conditions were so complex we had to come up with solutions like parity shedding. The project was known within Data General as Project Corona. The only server that the HADA storage system connected to was Data General's AViiON platform (which had a VME interface). This restriction led to the next invention.
- CLARiiON. By replacing the VME interface with the SCSI protocol, and shrinking the HADA form factor down to something much more compact, the resulting RAID5 system could be attached to many more server platforms, e.g., Microsoft, Sun, IBM, SGI, HP, etc. We overcame the RAID5 write penalty by creating the industry's first mirrored write cache. These two first inventions, and the software that underpinned them (FLARE) lasted for nearly 30 years until the Unity platform was replaced by Dell's PowerStore product. My passion for building data path software that delivered fast, trustworthy data to applications became a lifelong pursuit. To learn more about how our customers used the storage systems, however, I moved over to a new group and this new team began inventing storage management products.
- Navicli. Originally known as CLARCLI, this tool complemented the CLARiiONs ASCII-based Gridman interface. I wrote a command line interface (CLI) that thousands of our customers used to manage and configure the CLARiiON. We found that this tool really helped them with scripting and automation. As the CLARiiON family of storage systems continued to grow to 7-slot, 10-slot, 20-slot, 30-slot, and Fiber Channel products, we realized that our customers really wanted a graphical user interface (GUI) that could manage multiple varieties of storage across their entire enterprise.
- Navisphere. The Navisphere architecture was created for several foundational reasons (all documented here). Client-server architectures were all the rage and the Navisphere GUI and CLI both used a common library (internally known as RAID++, the first time that many of us learned C++ and object-oriented programming) that allowed our customers to connect to, monitor, and configure multiple storage systems from a single remote interface. Some of our architecture and design approaches were documented here.
- Storage Area Networks. As the CLARiiON / Navisphere technologies experienced rapid adoption, and the storage systems began to scale in size, the Fiber Channel protocol opened up a way to have multiple servers connect to the storage system and carve up the system into "virtual CLARiiONs" over the FC network. This resulted in the rise of Storage Area Networks, and it also attracted the attention of our main competitor EMC. They acquired Data General, and a whole new spectrum of innovation opportunities became available to me. Within 6 months I had transferred to EMC's Hopkinton MA facility. This jumping around from project to project helped me choose the name of my eventual blog (Information Playground).
- EMC Storage Scope. In 1999 so many customers were buying EMC's Symmetrix Storage system that they began to ask for by-the-drip, storage-as-a-service billing (because the systems were a huge capital expense). This, of course, was the beginnings of the need for cloud-based billing models. My first project at EMC was to work with my team to design a storage management framework that could inspect the usage of internal storage system resources by different customers so that they could be billed for their OPEX. This became the product known as Storage Scope.
- PowerPath Migration Enabler. One of the reasons that EMC was the #1 stock on the New York Stock Exchange was their SRDF software that provided a mirror copy of the Symmetrix system at a remote site. Every time certain customers needed to buy a Symmetrix storage system, they bought two of them, plus an SRDF license to boot! My team began to experiment with building device drivers at the server level that could access both systems and improve performance. This experimentation led to the creation of the PPME product that could do a live migration of all data from one system to another.
- Centera Seek. I spent three years in the Centera business unit at EMC, learning about this new technology called Content Addressable Storage (CAS). Customers would use a proprietary Centera API to store fixed content to the system. They would also associate metadata with the content, getting back a unique content address in return. There was no good way to search the metadata and content so we designed a Centera Seek appliance to do so. The idea also allowed us to do chargeback and reporting on the usage of the Centera itself.
- CAS Router. My experience with PowerPath software made me wonder if we could intercept requests coming to a Centera storage system and insert value-added functionality before passing it on to the Centera itself. I built a proof-of-concept internally to prove this point, and it eventually became a product that allowed a huge Oil & Gas company expand beyond one Centera to another by intercepting the routing and forwarding it somewhere among multiple systems. I wrote an article about how that idea developed.
- VNXe. About twenty years into my career EMC decided it had too many storage systems (CLARiiON, Centera, Celerra) and decided to build an architecture which componentized the best parts of each platform and put them into one platform. The company wanted all of it to run on LINUX. My role was to architect the data path and control path and work with the teams to build and deliver it, which took about 3-4 years. The story behind building VNXe was anchored by an amazing innovation by one of my colleagues called CSX.
- Unisphere. The building of the VNXe system gave the company an opportunity to build a brand new control path and storage management architecture called Unisphere. The goal of the system was to not only manage the VNXe but eventually other products at EMC (e.g., the Symmetrix product line, and eventually the Unity storage system). This involved integrating a UI technology at the time that was growing in popularity: Flex.
- Unity. When the VNXe grew up, it became EMC's flagship midrange storage line: Unity. There is a great Wikipedia article about Unity that traces the lineage that I've mentioned above.
- GINA. The twelve products mentioned above took 25 years of my career to build! I decided to step out of software engineering at that time and become EMC's Global Director of Innovation? What was the first thing that I did? Design an internal software tool with a global team of innovators working in Ireland, Russia, China, Cairo, India, Israel, and the US. I wanted to call the tool INGA (the name of one of the innovators) but we decided to call it GINA (Global Innovation Network Analytics). The tool centralized company-wide innovation data and ran analytics on top of it. This launched an industry-first innovation analytics tool inside the company that helped us measure knowledge of our global innovation network. Later on in my career we built another incarnation of GINA and created a production-quality version (Project MultiVac) running within Dell's IT department.
- Data Confidence Fabric (DCF). The final invention that I worked on in my career was called a Data Confidence Fabric. What I really liked about this invention was (a) it got me back to my roots of delivering trusted data to applications, and (b) we were able to open-source the DCF software to the LINUX Foundation (a challenge that I never faced before). The open-source codebase was known as Project Alvarium, and the evolution of that idea is documented here.
Comments
You can follow this conversation by subscribing to the comment feed for this post.