Come with me in my time machine for a moment. Back to the ancient days of the 90s: quartet singers, Victorian architecture, the Babbage Difference Engine…wait, not the 1890s. The 1990s: grunge music, McMansions, and the Mosaic browser. Remember in those days we also had things called “open systems”, which were servers instead of mainframes that ran UNIX.
They were called “Open Systems” because they ran Unix instead of IBM MVS, DEC VMS or DG RDOS, and they allowed applications from other companies to be developed to run on them. Each server vendor, though, had their own flavor of Unix: Solaris, HP-UX, AIX, Tru64, IRIX, etc.
We all Know what Happened: Linux Happened
Unix had always been freely available, but as noted each Unix vendor had their own proprietary version. When Linus Torvalds created Linux, he and made it open source he fundamentally changed the market. Starting in the mid-90s Linux began to come to prominence and with it the entire open source movement including copyleft theories and open source licenses like the GNU Public License (GPL) and the Apache License. By the late ’90s and early 2000s, start-ups began to appear like Cobalt Networks and VALinux that used inexpensive ‘white box’ servers instead of custom built server hardware. At the same time, the Internet was happening, and web companies needed racks upon racks of inexpensive servers that could be deployed and managed quickly and inexpensively to match the tight revenue models of those early days. The rest, as they say, is history.
And then there’s Storage…
As it tends to have done throughout computing history, storage tends to follow its compute big brother in both business trends and technology trends. So while open systems and Linux were rocking the compute world in the ‘90s, storage was rolling quite contentedly with proprietary Network Attached Storage (NAS) and Storage Area Network (SAN) solutions for both mainframes and servers.
Any semblance of openness in storage only recently began upon the introduction of software-defined storage (SDS) and scale-out storage architectures. Both architectures take full advantage of Linux and COTS hardware – applying it to the storage world in slightly different and sometimes conflicting ways (which we will cover in a different post.)
Enter Open Source Storage: Hello Ceph
But it has only been in the last few years that an open source storage project, namely Ceph, has been architected in a way that makes it a serious contender to support top tier enterprise applications. Inktank, funded by DreamHost and acquired by Red Hat in 2014, was the original company behind Ceph and included its original creator Sage Weil. Red Hat now deserves a special place in the Ceph community, but the community has grown beyond Red Hat to include a newly formed advisory board with top industry members: Intel, Cisco, Canonical, SUSE, SanDisk, Fujitsu and CERN. Many other companies (including Concurrent) are also participating actively in the ecosystem.
Customer traction for Ceph-based storage systems is broad and increasing in velocity. There are too many customers to name but the list includes Bloomberg, Flickr, CERN, and Wal-Mart. Ceph has also become the most popular popular choice for storage deployment with OpenStack.
So, the storage industry is clearly having its Linux moment. Whether the shiny, new players in the scale-out storage market realize it or not, they are about to be over-run by the same kind of wave that swept over the proprietary minicomputer and workstation market in the 1990s. The traditional, established storage players like EMC, NetApp, HP, et al, have large embedded bases which will give them an opportunity to react. The new storage market entrants that lack a deep user base, will find it hard very soon to differentiate their offerings as the Ceph community catches up and eventually passes them feature for feature in capabilities and embedded base of deployments.
So is Ceph the Linux of storage? What do you think?