The SPC-1 benchmark is cobblers, thunders Oracle veep The DataCore SPC-1-topping benchmark has attracted attention, with some saying that it is artificial (read cache-centric) and unrealistic as the benchmark is not applicable to today's workloads. Oracle SVP Chuck Hollis told The Register: "The way [DataCore] can get such amazing IOPS on a SPC-1 is that they're using an enormous …
I guess being a SVP of Oracle removes your sense of irony. After all, it's not like Oracle stack their appliances (such as Exadata) full of flash, operating as a cache is it!! They're not exactly shy with the DRAM either......
I'm waiting for Oracle to create their own benchmark.
The ExaSPC-1 or the SuperSPC-1.
"We actually only used 1.25TB (per server node) for the DRAM (2.5TB total for both nodes) to get 5.1 million IOPS, while Huawei used 4.0TB [in total] to get 3 million IOPS."
Somehow Datacore failed to mention their parallel server had a usable capacity of 12Tbyte, while the V18800 came with 69Tbyte usable capacity, 4,8TB usable per TB cache on datacore vs 17TB per TB on Huawei´s system.
jftr...
...is that while Datacore's cost-per-TByte (ASU capacity) was more-or-less equivalent to Huawei cost-per-Tbyte, each of the DataCore Terabytes delivered an IO-density of 431 IOPS/GByte, while the Huawei Terabytes delivered only 44 IOPS/Gbyte.
For my money, if I can get an IO density that is 10x greater with DataCore, and my cost-per-TByte is comparable, I'd have to be stupid to buy slow terabytes when I can buy 10x faster TBytes for about the same price.
And that's before I consider that DataCore IOPS (response times) are 3x faster...
It's a no-brainer. SDS rules, HDS drools...
-Disclosure NetApp Employee Opinions are mine, not my employers-
Mr. Hollis has never been much of a fan of SPC-1, or at least the uses to which it has been put by various vendors including that of my employer. I'd be more impressed if Oracle put up an SPC result, because most of the Oracle ZFS array performance comes from Oracle throwing LOTS of cheap CPU and RAM into their arrays, not the most efficient use of resources, but if they're cheap enough, most people don't care. Pot meet kettle ... you both appear to be covered in soot.
In any case, benchmarks like SPC-1 can be remarkably useful tools, even when wielded by a Vendor .. though it's usually worth noting that the benchmark usually proves a couple of interesting engineering points, which marketing usually spins into as much gold as they can (that's their job).
That data core can stitch together 72 CPU cores to get 5M sub 1ms IOPS without much if anything in the way of data services, is an impressive piece of engineering, I know of other more efficient IOPS/CPU core numbers out there, but that still stands as a great result.
It doesn't invalidate the benchmark, which is still remarkable in its disclosure rules and verifiability, but it does mean that people need to look a little more closely at that full disclosure report to figure out how useful / applicable that result is to their needs. Being #1 and having bragging rights for a few months is cool, but having the best overall fit for a customers needs is still probably more important.
Does anyone still listen to anything Chuck Hollis has to say? Good to see hasn't learned a thing and is still happy to help his competition by drawing attention to their achievements.
I thought people (and especially El Reg) would have let him slide into his cesspool of irrelevance when he became the perrenial punch line in the Nutanix vs. VMware idiot-fest.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
