That Oracle would trumpet at OOW 2016 a VM-based offering, the dense IO shape, is shocking, and disappointing, when Linux Container (LXC) technology offers order-of-magnitude better elasticity, density and performance relative to hypervisor-based VM technology.
Last year, I asked a certain founder of a company known the world over for its sturdy and magnificent Linux Desktop software at SCALE 14x conference if he felt container technology made the Xen-hypervisor-fork foundation technology of AWS obsolete, and his answer was no, and as I recall he cited full-isolation as the main reason for his answer. Still, that same company has staked a great deal on LXC technology, and has led, and continues to lead this field.
Those of us who champion container technology know that LXC smokes hypervisors on all metrics and features except complete isolation any day and twice on Sunday. I routinely build 6-node Oracle 12c multitenant ASM Flex clusters in LXC containers that perform at bare metal speed, turning in bare-metal SLOB benchmarks, which on a VM on the same hardware would choke, and probably be crashed by intensive SLOB tests .
So one such as I asks why did Oracle opt to go to market with what I would call a “me-too public cloud VM offering” rather than boldly use its reputed engineering prowess to solve the isolation problem and be a pure-play “container-cloud” giant, a space under-represented by the “public cloud big-4” of AWS, Azure, Google, and IBM. Note, to be fair, Google has Kubernetes, Amazon has their EC2 Container Service, and Nardella over at Microsoft is strengthening it’s ties to Canonical Ltd. and has the Azure Container Service. But all these offerings tend to be oriented around Docker and underlying VM layers (i.e. Amazon Containers run ontop of EC2 VM layer).
At OOW 16 yesterday, I popped over to the Oracle virtualization and container kiosks and found what I would describe as a lame offering that the Oracle reps described as effectively the “only” Oracle container offering which they called “Bare Metal As A Service”. That’s BMaaS I guess, right? Smirk. They told me that that is the extent of Oracle’s container cloud offering…you can get BMaaS from Oracle Cloud and build whatever you want on it, including for example, LXC containers. You’d think Oracle could have built some GUI provisioning and management tools for LXC containers technology. However, to be fair, Oracle now fully supports Oracle 12c database in LXC containers provided you run a suitably recent version of Oracle Linux Unbreakable Enterprise Kernel (UEK). Oracle for those who might not know, has its own fork of the RedHat-family of kernels called Oracle Linux. They offer this support for Oracle database 12c in LXC containers for both Oracle Linux 6 and Oracle Linux 7. And to be sure, since it’s a true bare-metal platform for LXC, it deserves mention here.
Oracle, like BlackBerry, missed the boat by about 10 years on Public Cloud just as BlackBerry denied reality about Apple iPhone. Oracle still had, and has, a chance to lead containerized clouds, but apparently lacks the will to spend on R&D to “make it so #1” and solve the challenging full-isolation issue holding container adoption back.
Therefore, I predict that like BlackBerry, the Oracle Public Cloud has a murky future at best. Currently the Oracle Public Cloud does not even make it into the 2016 Gartner Magic Quadrant. BlackBerry went multi-platform in recent years trying to catch up to Apple, but the new approach at BlackBerry also cannibalizes BlackBerry’s “Service Access Fees” which it loses every time someone junks their old BlackBerry handset. SAF for those who don’t know is per-phone, per-month fees BlackBerry used to collect from BlackBerry users. SAF is quickly eroding at BlackBerry and it remains to be seen if software and services products at BlackBerry can close the breach on vanishing SAF revenue.
Similarly, Oracle is getting intense pressure on all sides about its exorbitant support fees for Oracle database, and to compete in the cloud yet preserve support fees Oracle has to somehow match AWS incremental cloud cost while not cannabalizing it’s own SAF-like support fees.
This leads to a big clue about why Oracle eschewed superior LXC containers technology for its public cloud. Containers run at 10x the density of VM’s and hence Oracle likely did the math on what a container cloud would do to Oracle support revenue (which remember is processor/core based) and said to itself “let’s not go there”.