The servers were in a child’s bedroom. In 2005, Anand Babu Periasamy and Hitesh Chellani started Gluster inside Chellani’s home, with noisy machines occupying his daughter’s room. Silicon Valley has a fondness for the garage as a founding address. A bedroom is less cinematic, and the landlord presumably has stronger opinions about bedtime. Yet this was where the two engineers began a business that Red Hat would acquire six years later for approximately $136 million.
Periasamy, usually called AB, has since co-founded MinIO and become its co-CEO alongside Garima Kapoor. The businesses differ, and the computing world around them has changed. A recognizable preference remains: put the intelligence in software, make it accessible, and be suspicious of the apparatus that accumulates around it. Follow his career and storage becomes a surprisingly personal subject. It concerns who gets to build, who must ask permission, and how much complication a useful idea can survive.
A computer, then a way in
Periasamy’s fascination with computers began in ninth grade. He has described seeing one for the first time as love at first sight. His mother encouraged him to become an engineer; he credits her support with getting him through his studies. He graduated in computer science from Annamalai University in Tamil Nadu in 1999. Before there were acquisition announcements or investor meetings, there was the much more immediate attraction of making a machine do something.
At NextGen in Delhi that year, internet access was restricted to a few employees. Periasamy figured out how to connect when the service came on in the evening, listening for the modem and investigating the available ports. Colleagues started asking him for help, and he ran a proxy on his computer. The anecdote has an appropriately practical scale. The obstacle was access. He wanted to get online. He worked out a route.
The attraction of free software gave that instinct a larger purpose. Code could be studied, changed and shared. A programmer could follow an idea beyond the boundaries of a job description. His later career would repeatedly turn on that freedom, even when pursuing it required the decidedly unromantic work of finding customers and meeting payroll.
Thunder before the bedroom
At California Digital Corporation, where he served as chief technology officer, Periasamy worked on the Thunder supercomputer for Lawrence Livermore National Laboratory. Installed in 2004, Thunder ranked second on the June TOP500 list. Its specifications give the achievement a useful physical outline: 4,096 processor cores, a Linux operating system, and measured performance of 19.94 trillion floating-point operations per second.
The problem was how to make many components work together at a level that any one component could never reach. That question would recur in storage. A collection of ordinary machines becomes useful at a different scale when the software can coordinate it. Thunder was a concrete demonstration of this kind of engineering, before Gluster made it the basis of a business.
When California Digital closed its hardware and free-software business, Periasamy faced a less elegant coordination problem: finding work for himself and his team. He wanted to continue working on free software. He and Chellani eventually started their own company. Founding a business was a way to keep the work going, with all the obligations that followed.
A room full of hardware, a bet on software
Gluster began with $200,000 from angel investor Anil Godhwani. The early meetings took place at a Starbucks in Fremont, California; the working address was Chellani’s home. By the time the company sought its first institutional financing, it had reached a precarious arithmetic: $25,000 in the bank and 13 people to pay. Its $4 million first round arrived amid the 2008 financial crisis.
Gluster’s proposition was to combine commodity storage and computing resources into a shared pool through software. Customers could increase capacity by adding machines. The software provided the coordination. For organizations confronting growing quantities of video, documents and other unstructured data, that offered a different set of choices from buying a single expensive storage system and then buying a larger one.
Red Hat announced its agreement to acquire Gluster in October 2011. At the time, Gluster’s customers included Pandora, Box.net and Samsung. Periasamy had been its co-founder and CTO, responsible for the file system’s architecture and strategy. He subsequently joined Red Hat’s Office of the CTO. The transaction put his work inside a company built around selling and supporting open-source software.
“Now, we are the storage of Red Hat.”AB Periasamy · October 2011
That line, delivered at the acquisition announcement, has the neatness of a programmer’s joke: the same terms, rearranged, produce a new result. It also gives the deal a human scale. A venture that began in a borrowed bedroom had found a home where its original technical preference made commercial sense.
The right quantity of something
MinIO was founded in 2014 by Periasamy, Kapoor and Harshavardhana. Its name joined minimalism with I/O, the shorthand for input and output. In a 2015 interview, Periasamy described an ambition to make cloud storage manageable for an ordinary JavaScript developer. The imagined user mattered: someone building an application, who needed storage to work without becoming a separate career.
The team had tried other approaches before settling on Go. An early prototype used a C++ core with JavaScript extensibility; the languages considered also included Haskell, Rust and Java. Periasamy liked Go’s productivity and its opinionated design. The decision was consistent with the product’s intended audience: reduce the amount of machinery required to get useful work done.
Years later, he described minimalism as “the right quantity of something.” Too little leaves the job incomplete; excess adds baggage. His feature requests face a formidable admission policy: explain why the addition will help everyone. Removing a feature has a considerably easier time getting through the door. Many software products appear to fear an empty checkbox. His stated preference allows one to remain empty.
In a 2020 presentation, he illustrated the idea with a 44-megabyte static binary. The historical size was part of the argument: a storage server could be compact enough to run at the edge while serving a much larger deployment elsewhere. He discussed Raspberry Pi clusters and 5G towers, bringing a subject usually associated with data centers down to devices an audience could picture.
His explanation of MinIO’s internals stayed close to those design choices. Nodes had equal capabilities. There was no hierarchical metadata server directing the rest. Objects were distributed across drives and nodes, with erasure coding helping the system tolerate failures. Simplicity in the interface still required careful engineering underneath. Removing a component made the remaining components more accountable.
The cloud needs somewhere to put things
MinIO’s object storage speaks the Amazon S3 API, the interface through which applications store and retrieve objects. In conversations about multiple clouds, Periasamy has emphasized the ability to run compatible storage in different places: a public cloud, a company’s own infrastructure, or an edge deployment. The application has a familiar way to ask for its data even when the location changes.
That is a particularly revealing version of his access argument. The programmer is spared some of the work of learning a different storage vocabulary for every environment. The organization gains choices about where the system lives. Compatibility is therefore part of the product’s appeal, alongside its performance. A useful interface can travel further than the service that first made it familiar.
In January 2022, MinIO announced $103 million in Series B financing at a $1 billion valuation. Intel Capital led the round, with SoftBank Vision Fund 2 and existing investors participating. The company reported 855 contributors and more than 9,000 pull requests since 2015. Those were dated measures of a community making and revising software, as well as evidence for a business seeking to expand.
Periasamy’s connections extend beyond MinIO. Public interview biographies describe investor or advisory relationships with companies including Postman, Starburst and Yugabyte, and board roles at H2O.ai and Manetu. His involvement with the Free Software Foundation of India links the commercial career to a longer commitment. The same person can discuss a financing round and the freedom to modify a program; both subjects have remained in his orbit.
Now the waiting happens at the GPU
By 2024, his discussions were increasingly about AI. He described customers facing pressure on budgets while others expanded their infrastructure because of AI demand. That combination made the storage question more urgent: how do you feed demanding applications without making every increase in data an increase in operational difficulty? His conversations continued to return to replication, cloud-native architecture and keeping systems understandable.
At Cloud Field Day in June 2025, Periasamy presented AIStor’s S3 Express compatibility and integration with NVIDIA GPUDirect Storage. The discussion concerned the work required to move data into AI workloads efficiently. A GPU can perform extraordinary calculations, but useful performance depends on the route by which the necessary data reaches it.
A June 2026 MinIO presentation turned to MemKV, a distributed memory layer for AI context. As an AI application’s accumulated context grows, its cached computation can exceed GPU memory. Evicting that cache and calculating it again consumes time. The proposed memory layer addresses that repeated work. It is a new technical setting for a familiar question in Periasamy’s career: what can the system avoid doing twice?
There is room in this technically serious life for a less predictable pairing. Periasamy has identified himself as a fan of both Star Trek and SpongeBob SquarePants. The combination supplies a welcome corrective to the ceremonial gravity of enterprise infrastructure. His earlier description of software as an art form offers another view of the engineer: a person interested in how something is made, as well as what it can sell.
A bedroom full of servers, a distributed file system, a compact object store, a memory layer beside an AI model. The settings keep changing. Periasamy’s public explanations keep examining the work hidden inside them. Somewhere between the machines and the person trying to use them is a complication that can be removed. He has made a career of looking for it.