Monday, March 2, 2020

An Open Letter to Oracle for SPARC & Solaris





[Solaris Logo, courtesy Sun Microsystems]

An Open Letter to Oracle for SPARC & Solaris

Abstract:

The Market had always wanted revolutionary improvement from systems and operating systems, in order to produce cost of ownership improvements and a competitive against other software players. Many promises were made from Sun Microsystems, some were carried out by Oracle after acquisition, but there are still many needed features that must be completed.








SPARC M8 Laundry List

There is really not much that is desired, with Oracle producing an amazing Hardware Architecture with Software Stack to run on top. With a processor running at 5GHz, 32 cores per socket, 8 threads per core, and 4 instructions per clock cycle, hardware crypto-decrypt units, hardware decompression engines, Java streams accelerators, and Oracle RDBMS [numbers & query] accelerators - what else could be needed?


SPARC Trademark

The SPARC Trademark appears to be getting watered down, lately. More press releases are needed for the architecture, to exercise the rights of the holder, for legal purposes. The organization has seemingly gone mostly dormant, for the past number of years, with new Oracle partners in the SPARC environment not contributing press releases through the international organization. This needs to end. People offering emulation need to be participants in the trade group and contribute their dues, which are very small.









[Flash Card, courtesy Sun Microsystems]

Hardware

With competing organizations, like IBM, now owning Red Hat Linux, Oracle is left with the most scalable hardware in the market, with access to only a few select larger customers. The hardware reach must go downward, which IBM now has with their RedHat Linux on single board computers.


From a hardware perspective, scaling small for Makers and scaling big for Industry is needed:
- Intel CPU has proven itself as a terribly risky proposition to RISC, with never ending streams of security vulnerabilities - a commitment to secure SPARC is needed more than ever!
- Optane / QuantX / 3D XPoint NVRAM from Intel / Micron must be accommodated in a useful way, perhaps a native chassis interfaces controlled by firmware for persistent ARC, L2ARC replacement, or SLOG intent log replacement.
- a shrink of die, if for nothing else than re-adding future integration
- - network integration to a SPARC die for makers less expensive storage nodes
- - cross-bar switch for vertical scalability in single chassis for SMP systems
- - InfiniBand endpoint, for glue-less MPP integration or engineered systems
- - Memory, for inexpensive Compute Nodes in an engineered system
- a hardware compression engine has always been missing, to make 100% compressed datasets as native as 100% encrypted datasets
- 1 TB addressability per socket is a good start, but there needs to be a drive towards 2TB or 4TB of memory per socket, to work with ever increasing datasets.
- an Open Source 1x core 8x thread implementation, single board SPARC platform for Makers
- an Open Source OpenFirmware implementation to be used on a single board SPARC platform, for Makers
- an active push into the Universities & High Schools with single boards running Open Firmware and Solaris




If Oracle wishes to stay relevant, they must be more than a hardware knock-off company and start feeding the market with desire for their products from the lowest levels.






[RedHat and IBM logos, courtesy IBM]

Software

With competing organization, like IBM, now owning RedHat Linux, Oracle is left with the most scalable OS in the market (Solaris), with Engineered Systems based upon a knock-off version of IBM RedHat Linux. The software reach must go downward & broader with Solaris.
- Open Firmware needs to be re-released & re-spec'ed by the IEEE, so makers to have a future
- promises for rebootless-patching via K-Splice on SPARC must be delivered, as K-Splice has less of a future with similar rebootless patching now available on Linux (from other competitors like IBM, who own the OS that Oracle Linux is a knock-off of)
- promises for ZFS to expand out to a true horizontally scalable clustered filesystem need to be fulfilled, so perhaps small Solaris single boards can be combined for horizontal scalability
- A small version of Solaris, to run on small SPARC single boards, for makers
- A decision to support Java on SPARC, for makers to code in.
- A small Oracle RDBMS, to run on a SPARC single board, for makers, to make Oracle RDBMS relevant in today's computing environment
- Cooperation or Commissioning of Open Source Solaris alternatives, like Illumos, to run in the Maker environment, or ZFS Clustering, to maintain relevance to emerging human resource base.


When people come out of school today, they do not think of Oracle as a provider of anything. Software is a way to change this, and not by merely someone who copies IBM's operating system.




SPARC Solaris Cloud

Oracle released Solaris 11, as the First Cloud OS, yet only customers are building clouds with it.
- It is long overdue for Solaris 11 ZFS to provide block storage to Oracle cloud
- It is long overdue for Oracle RDBMS and Apps to run in Oracle Cloud under Solaris 11
- It is long overdue for LDom's to be available in a SPARC Cloud
- It is long overdue for Cold and Live Migrations of LDoms and Zones to an Oracle Cloud
- It is long overdue for Storage Migration of LDoms and Zones to an Oracle Cloud


This is not rocket science, as customers do a lot of it today, but it is time for Oracle to monetize on the Cloud with SPARC & Solaris, instead of refusing to provide visible cloud solutions to customers.








Your Own Destiny

It is written: Without Vision, The People Perish. A maker's dream might be a shrink of the M8 processor, with 1x core comprised of 8x threads, and the rest of the die filled with an ethernet & raw memory. A secondary version holding a crossbar on-die & headers on the board, so boards can be stacked on one another, similar form-factor to a Raspberry Pi, to create a scalable SMP platform. Allow OpenFirmware and Solaris to reside on it, while adding Oracle RDBMS on it, for educational purposes, and small business as a mini Oracle Database Machine (charge per additional board, aggregate ethernets for increased throughput.) End the dependence upon Intel & IBM & licensing other processors for the future - drive your own destiny. Facilitate third-parties to drive the hobby & educational & MPP markets (i.e. Illumos.org & unixpackages.com) while concentrating on the




Conclusion

The industry had watched, with great anticipation, the union of Oracle and Sun Microsystems. The creation of the fastest processors on the market, from SPARC family & a fabulous Solaris 11 OS has been the apex of computing, but resting  since 2017 is insufficient. A minor amount of investment or even just partnership with existing Open Source organizations is all that is needed to take Oracle to the next level - sustainability of Oracle SPARC, Oracle Solaris, and Oracle RDBMS for the next generation. With so many very good RDBMS's in the Open Source market on top of the Linux that Oracle copies from IBM RedHat, Oracle must decide to be a leader again, rather than a shell of a copy of competitors. Without advancing SPARC & Solaris, anyone can do what Oracle does.





No comments:

Post a Comment