DMX, DMX2 and the Early days of Symmetrix
It’s getting pretty late here in Atlanta. Travelling to this part of the US for two days for business visit to win some business. Had a long good day, couple of meetings at the office, flew in Atlanta, had dinner, some drinks and now this blog……. hope I can finish it……
I am really not sure if the title of this blog is exactly what this blog is about, but I think it is quite closer….
So from the Symmetrix (SCSI based Disk Arrays), EMC moved aggressively to release the all Fiber based disk array, now called Symmetrix DMX (Direct Matrix Architecture). While the DMX was a major move or an upgrade from SCSI – 40Mbps Disk I/O to a new speed of 2048Mbps Disk I/O.
Early in the days the major bottle neck use to be around the Disks speeds. With the I/O’s per second reaching 2 Gbps on the DMX1 and DMX2, the disk drives now work at the same speed as the fiber directors I/O’s per second. The DMX was a notch above with atleast 40 times better performance than its predecessor Symmetrix.
The backend of the DMX Symmetrix came from the acquisition of Data General. The DMX800 was one of those hybrid boxes that got EMC Symmetrix engineers really confused about how to use the DG Clariion backend on a Symmetrix Enterprise Level System and they created the DMX800 (oh believe it or not, I have had customers telling me they have ghost like issues on that box).
A Clariion architecture to be implemented with a Symmetrix Enginuity Code is a DMX800 (don’t get me wrong, I am not talking bad about the Clariion, we are comparing the enterprise level machine code running on a really strong mid tier machine). EMC Symmetrix Engineers learned a lot from it and then came the DMX1000 and DMX2000. With the early success, EMC went to work on a much larger config of the DMX3000. As much as the DMX3000 is more or less the same technology as the DMX1000 and DMX2000, the memory requirements, power requirements, device numbers, address spaces, memory, algorithms, data structures, productive laptop (Service Processor) required for handling the memory and address space on Symmwin were much different. The Enginuity Code was now able to handle drives (volume addresses) much more than the older Symmetrix 5.5 and the new DMX1000 and DMX2000.
So during that time came the Enginuity code 5669 on the DMX800 and the DMX1000. Ask the Symmetrix engineers what happened? Every day the drives would drop, loops would drop causing backend issues, it was a support and engineering nightmare in real sense. These were the good old days of EMC making tons of money, but again as the call volumes went up and they starting selling more boxes internationally, it got EMC to start creating product support labs in Sydney, Australia and Bangalore, India on top of Hopkinton, MA and Cork, Ireland.
Almost after a year of the DMX, EMC now was at a point of releasing the DMX2 which in essence would be 2 times more faster than then its predecessor, the DMX. The DMX processors per director were of speeds of 500 Mhz Power PCs. With the DMX2 now each director was configured with atleast 1Ghz Power PC chips.
This was more a data-in-place upgrade like they offered on the Clariions back in the day. This would be a change of all directors throughout the box. The only drawback was this was an offline change and customers get scared away from the offline events. With the DMX2, now it would support the 73GB – 15K disk drives, Single memory board going up to 32GB and the largest memory configs being supported upto 256GB.
Also with the DMX1000, DMX2000 and DMX3000 and the 5670 code came a native support for RAID5 (3RAID5 and 7RAID5) on top of Mirroring (M1/M2), RAID S (3+1 and 7+1), BCV and DRV devices. Now with the support of RAID5, EMC all of a sudden started competing in the market between the mid tier and the enterprise applications. Enterprise customers requiring data storage for mid tier applications and non critical applications started using RAID5 along with some customers using it for critical business continuance backups.
Raid5 was a big boon for system administrators and storage administrators, which in essence they were all used to across all their other storage platforms and systems. Now with almost a similar protection level, you could lose a member device of a Raid5 and still be fully functional, like losing a mirror. There are still drawbacks of losing another member in the same Raid group or losing a mirror whose partner is already down. The concepts of BCV’s would come in to play in these situations.
Anyways, the DMX is where EMC made a big impact into the storage market. EMC used to be a 6 Billion company in 2004 to being about 13 Billion in 2008. A lot of it has to do with the software acquisitions but their move to get into the fiber based storage market has put them to being the storage industry solutions provider company over the past decade.
The internal transformation of EMC under its existing management from being a Hardware Storage Company to being a Hardware/Software Storage Company to being a Storage Solutions company has kept EMC as a leader in this industry…… span>