The Center for Large-scale Data Systems Research has developed a big data benchmarking community and has held big data workshops. I recently got an announcement that they are looking at moving the big data benchmarking effort to SPEC and specifically SPEC Research Working Group on Big Data Benchmarking.
Having an industry standard benchmark is always a good thing, if the benchmarks are not gamed. Sadly, many industry benchmarks are gamed. For example, the SPC-1 benchmark has responses numbers with 100s of thousands to over 1 million IOPS. What are real users with real applications doing and needing for IOPS numbers? Do a significant number of real buyers, with real applications, use all-flash arrays for their storage needs?
I think for the most part the likely answer to these questions is no, yet many organizations seem to buy hardware based on SPEC benchmarks rather than understand their workloads and buy hardware based on their requirements and of course budget Do not get me wrong – I am in favor of benchmarks and most, if not all, of the SPEC benchmarks are well suited to measure what they are designed to measure. The issue I have is that the hardware configurations are not designed around real world representations of the real problems, nor does anyone (that might be an exaggeration, but just about anyone) goes out and buys the SPEC benchmark configuration that is at the top of the heap.
In the benchmark world today you have to run SPEC benchmarks and publish, given the environment we live in. I am not sure that, given how SPEC benchmarks are used, it makes good sense. But on the other hand what other agreed upon benchmarks do we have that measure aspects of the hardware and software that we need to have measured? Clearly, SPEC storage benchmarks are a double-edged sword, where they are well designed and well conceived, yet at the same time the underlying hardware used is often not configured for the real world.
Photo courtesy of Shutterstock.
Labels: data storage, big data, benchmarking, big data storage, big data analysis
posted by: Henry Newman
I saw that the Fibre Channel Industry Association announced that 128 Gbit Fibre Channel will be available broadly in 2016. Just to be clear, that means – quoting from the press release – “Gen 6 doubles the 16GFC data throughput of 3,200 megabytes-per-second (MBps) to 32GFC, which enables 6,400 MBps full-duplex speeds. Gen 6 also provides an option to quadruple 32 GFC to 128GFC throughput, thereby achieving full-duplex speeds of 25,600 MBps, based upon the seamless inter- and backward-compatibility of Fibre Channel technology.”
Think about this folks: We are going to have an 8x improvement in FC performance in two years. This is unprecedented for any technology that I can remember ever, but I have only been around for 30+ years.
There are a few things that bother me about this announcement besides the performance increase claims. First, what is FC going to plug into? PCIe-3 16 lane is 96 Gbit full duplex. So that does not work and I have not heard about anyone thinking or talking about a 32-lane PCIe 3.0 slot anyway.
PCIe-4.0 specifications are supposed to be released in late 2015 according to the PCI-SIG. You would think that the FCIA would talk about interfaces but there is no mention of the words PCIe or interface anywhere to be found in the press release. It is going to be fairly difficult if not impossible for a specification from the PCI-SIG to be released late in 2015, a CPU vendor and motherboard vendor to develop a slot and interface, another vendor to develop an interface chipset for the slot, and all of this to be widely available in 2016.
The PCI-SIG has not been known for being on time, much less early, given the complexity of the work they’re doing. Look what happened with PCIe 3.0 and how late it was and how long after release it was before there were products, even though Intel and their partners released Sandy Bridge systems about 4 months after the release of the specification. There weren't widely available interface cards for a good 8-10 months. It would be nice if the FCIA was even close to correct, and maybe they know something I do not know – but most likely they do not and they’re dreaming. Just my opinion, of course.
Labels: storage networking, enterprise storage, Fiber channel
posted by: Henry Newman
Micron recently presented their Winter Analysts meeting and offered a review of the industry and some hints on the direction on where things are going. A few highlight from the slides include that the supply growth for both DRAM memory and NAND. The reasons for this can be found on slides 41 and 49 where future products are discussed, including 3D memory cubes and a mystery “new generation” of memory.
I read an article that Paul Shread wrote a few months ago on Nantero, which designs carbon nanotube memory. There is a huge need (slide 54) for low power non-volatile memory for both mobile devices and for larger memories near the processor, which are needed for things like data analysis and table space.
Another interesting slide is 56, which looks at the growth and bandwidth needs of the graphics memory market, which really surprised me. Slides 60 through 66 focus on the NAND storage market and Micron’s product offerings. Micron has a wide product line that includes everything from PC storage to enterprise SAS storage, but looking at the web site I cannot get many details as I am not a reseller. I did get one brochure for the P410m and it did not seem to support the ANSI T10 data protection model.
So what does this all mean for our future in the data storage industry? I think Micron and likely other companies are going to making some major changes from 2015 to the end of the decade in the area of non-volatile memory as the market demands changes for mobile devices that need both low power usage and non-volatile memory. The number of mobile devices around the world is in the billions and it’s growing faster than any type of device, therefore the first vendor to market with something that changes the game will win. Read the Micron PDF and see if you see what I see.
Labels: SAS, Micron, Carbon Nanotube, mobile device market
posted by: Henry Newman
I am sure all of you have seen the latest on Violin Memory (the company has five buyout offers after a disappointing IPO). So what are things going to look like in, say, mid-2015 and 2016?
In opening up my crystal ball, I think what will happen is that the NAND suppliers themselves are going to take a lesson learned from others in the industry and move up market. This means that both Samsung, Intel, Toshiba and others that make NAND are going to start making storage devices that can be sold in the market.
NAND is a commodity part that had low margins, even storage devices such as a disk drive have low margins but are likely better than NAND, but storage appliances have way better margins. I think that this is happening in other markets and it will start to happen in the SSD market.
The question is: can these new vendors pull off this huge change? Selling NAND or even selling NAND-based storage is far different than selling to SMB, which is far easier to sell to than the enterprise. There are issues of customer knowledge and expectations on product. The fact of the matter is that today people are looking for integrated products that are appliances that can address multiple types of requirements from big data, to file storage to even object storage. There has to be a lot of software written. The only way to accomplish that in a short period of time is through acquisition.
That means that smaller software companies and hardware companies that have NAS, support big data analysis, support objects and/or have software that helps all of this are going to be on the auction block very soon. If I am right, by mid- to late 2014 we are going to see purchases in this area by the NAND suppliers. Time will tell if I am correct.
Labels: SSD, flash memory, IPO, Violin Memory
posted by: Henry Newman