
news flash, memory is faster than disk
somehow I fail to find this all that impressive.
Texas Memory Systems has absolutely creamed the SPC-1 storage benchmark with a system that comfortably exceeds the current record-holding IBM system at a cost per transaction of 95 per cent less. TMS submitted a RamSan-630, a 3U box holding 10TB of single level cell flash that delivers – according to TMS – one million IOPS ( …
For reading, maybe...
But once a flash memory device runs low on free blocks, sustained write rates can suffer greatly, and in many cases SSDs end up being slower on writes than a good mechanical hard disk drive:
-- AnandTech: The SSD Anthology: Understanding SSDs and New Drives from OCZ
-- -- http://www.anandtech.com/print/2738
Look at the reports, Go ahead and I'll wait:
http://www.storageperformance.org/benchmark_results_files/SPC-1/TexMemSys/A00105_TMS_RamSan-630/a00105_TMS_RamSan-630_SPC1_executive-summary.pdf
http://www.storageperformance.org/benchmark_results_files/SPC-1/IBM/A00087_IBM_DS8700_SVC-5.1-6node/a00087_IBM_DS8700_SVC5.1-6node_executive-summary-r1.pdf
For the tested ASU storage capacity. Which is cheaper per GB?
...than IBM. But the IBM system is a classic example of a benchmarking product built for a big headline number but not something that anyone would purchase in the real world. Their costs for capacity and performance are both stupidly high.
I can see a place for the TMS unit where ultra-high performance over a relatively small amount of storage is required, but given it is still coming in at up to 5x the cost of capacity of other SPC-1 results it isn't exactly what you would call general-purpose storage. Saying that it is cheaper than IBM for both capacity and performance is no more of an accolade than "not in a position to stiff customers who have no idea what they are buying but know it needs to be blue".
One reason the TMS system is cheaper is that there has always been a niche to get a good $/IOP score with a small quantity of flash. Few vendors occupy this niche, because there isn't a huge market for it.
The SVC config tested is over 97TB and has a "minimum" of 25% headroom to scale the performance as it only has 6 out of a possible 8 nodes, it's not an unrealistic config for Enterprise. Will the RamSan scale ten times and see a linear increase in performance or will it top out and simply add cost and capacity, but not performance, hence ruining it's $/IOP..
Horses for courses.