Super-fast arrays: We already have them
e.g.
char x[100]; /* Fast. */
Stealth-cloaked startup claims to be developing super-fast arrays. How fast? Well...
Vexata is a stealthy storage array startup with a product technology that it has been telling people is "25 times faster" than arrays from leading rival vendor and VC favourite Pure Storage. That's the claim made in pre-stealth emergence material, anyway. We have no product against which to measure performance data. The VCs …
-
Tuesday 4th April 2017 14:43 GMT Chris Gray 1
bypass?
[Note: I used to work in the storage industry, but have been out of touch for 10 years.]
Typically a storage system sits between the consumer machines and the raw disk drives. So, I/O's have to go through buffers in the storage system when going in either direction. That doesn't allow latencies as low as a raw SSD device. So, they must be doing something different. Some thoughts:
Perhaps they have some custom chips that are able to reroute and rewrite I/O requests so that host requests go directly to the appropriate SSD, without having to read the data into storage system memory first. It should also be possible to use those same chips to observe the data going by and buffer it for caching purposes, if that is appropriate.
More custom silicon could perhaps allow them to rewrite, on-the-fly, FibreChannel requests into SATA/whatever requests to the actual disks.
If this is not what they are doing, I'm curious to find out. And if its not, and anyone uses the above to make money, please send some of that money my way - I could use it!
-
Wednesday 5th April 2017 15:24 GMT Gigabob
Re: bypass?
I think you are very close to the mark - note the investors, Intel etc - they are using FPGA's to build new I/O stack for the data center that explosively improve on search times across the cloud storage environment. Stands to reason a new interface and attachment software is required to enable this.
-
-
Tuesday 4th April 2017 15:27 GMT Androgynous Cow Herd
Performance has limited value
The press release could have been written for Violin or Datrium. There are a few shops that can monetize ultra low latency. Most places don't care about the "fastest" or the "quickest" if they give up the stuff that matters more over the lifespan...scalability, data protection, availabilty and integrity, and these days, predictive analytics.
It's neat that you can go fast. But if that is all your customer cares about and all you can deliver, you will lose them quickly. Shops that only value performance have no loyalty to any vendor, and performance isn't "Sticky"
-
Wednesday 5th April 2017 05:22 GMT Alpy
I stopped at Fibre Channel!
Honestly? Fibre Channel? Why would you design a super quick storage array and glue legacy technology such as fibre channel to it? Ethernet and Infiniband technology is currently pushing storage around at 100Gbps and 200Gbps while using RDMA (RoCE) to obtain very low latency and very high throughput. Fibre Channel is dead in modern infrastructures. Let it die!
-
Tuesday 11th April 2017 01:19 GMT Steve Chalmers
Speculation
How about this as sheer speculation: this is a native NVMe over Fibre Channel array, where an FPGA or ASIC (in the style of Apeiron or Kazan) is hardware-distributing incoming I/Os to a collection of NVMe drives on a collection of PCIe buses.
The network business made the transition over a decade ago from software routing packets to last year's leading edge layer 3 switch ASIC routing about 3 billion (3 x 10^9) packets per second at a latency of about half a microsecond. Hardware of this type is very hard to design correctly and get all the corner cases right, and expensive to engineer, but quite doable.
The hard part is the software, both behind the scenes on data integrity and error recovery, and visible services. In this new world as in networking, this software will run as the "control plane", setting up the hardware tables which allow individual I/Os to execute entirely at hardware speed in the "data plane".
Oh, and if Vexata has any sense, they will not fall into the trap Violin did and design only for legacy access, they'll also offer a forward looking programming model (my favorite is using the hooks in a supercomputer connect like InfiniBand to execute storage reads and writes directly from user space) which eliminates I/O path length at the server. Or maybe they've already figured out something better...
-
Friday 28th April 2017 10:35 GMT hoola
Fibre Channel
Before stating the FC is "legacy" you need to understand about latency. IB is very fast, over short distances. It is also a bugger to cable. Ethernet in all its variants, FC over e iSCSI ultimately suffer from the same problems, latency. Ethernet is cheap to deploy until you need low latency, and distance at that point you start having to put fibre in. If I am deploying fibre and segregating as recommended (that is not the same as a vLAN) then you may as well use FC as it is the correct tool for the job.
Biting the hand that feeds IT
