So what is Matrix? Basically, it is the Insight systems management software already on the blades, merged with the orchestration software that came from Opsware (formerly known as Opsware Workflow) that has been given that graphical templating environment to make it easier to provision, patch, and manage servers and their software. HP has created all of the templates for the Opsware tool necessary to manage its various server blades and chassis, its Virtual Connect virtualised I/O for storage and networking, and the usual suspects in terms of operating systems (Windows, Linux, and HP-UX). Thome says that HP is working on templates that will allow Matrix to control Microsoft, Oracle, and other databases, and then popular application software that runs atop all this. Customers can use the Matrix tools to create their own templates for automating homegrown code.
As we previously reported, Cisco has embedded its management code inside the switch atop its chassis, which is known as the UCS 6100 fabric interconnect. HP is putting the Matrix software onto an x64 blade in its BladeSystem chassis. (Not on a service processor inside the chassis, which may or may not make a difference to IT managers.) One of the key features of the Matrix software stack, called Virtual Connect Enterprise Manager v1.30, can manage LAN and SAN connectivity for hundreds of BladeSystem enclosures and can assign failover assignments for up to 200 server-to-network virtual connect domains.
The Matrix hardware includes the c7000 chassis, as well as the current ProLiant G6 generation of blade servers, using Intel's quad-core "Nehalem EP" Xeon 5500 or Advanced Micro Devices' quad-core "Shanghai" Opterons. Presumably the next generation of quad-core "Tukwila" Itanium processors and the current generation of dual-core Itanium 9100 processors are also part of the Matrix hardware mix.
According to HP, a Matrix environment can be extended to manage up to 1,000 systems, which is far more scalable than what Cisco has achieved with California. Using the top-end 40-port version of the UCS 6100 fabric, a California blade system tops out at 320 servers.
HP was also keen to point out that it has its own storage, unlike Cisco. The Matrix system will include a new virtualized direct-attach storage array for the BladeSystem, which is comprised of a SmartArray 700m SAS RAID controller and a six-disk enclosure that plugs into the c7000 chassis. The controller on this blade and its disks connect to the chassis, which in turn maps storage to blades. Each server thinks it owns the disks assigned to it, as if they were plugged into SAS ports on the blade, but they are actually virtual SAS links that can be tweaked.
HP is also announcing an external StorageWorks Modular Disk System 600 array, which has a 5U enclosure that sits near the BladeSystem c7000 chassis and can hold up to 70 3.5-inch SAS or SATA drives. This external unit links to the chassis through an SAS controller, just like the internal one does. A single enclosure, equipped with the proper number of SAS modules, can have up to six MDS600 arrays linked to it, and the storage is mapped to the blades in the same way as with the internal disk enclosure for the c7000. The MDS600 array will be available at the end of April, with a starting price of $10,000. HP's StorageWorks EVA 4400 arrays can be plugged into the Matrix system, too, and these are in fact the default storage devices for Matrix right now.
HP was also keen to point out that the virtual SAN appliance software, which it got through its $360m acquisition of LeftHand Networks last October, will be deployable on a future release of the Matrix blade system, too. This P4000 SAN appliance software will be deployed on HP's SB40c storage blades. HP also says that starting on May 1, all of the former LeftHand products will be available across HP's server line and with HP part numbers.
Matrix vs. California...or just go with Liquid
Unified computing - the complete solution (which seems to elude the two big vendors playing in the space today) is a very disruptive force that will fundamentally change how data centers are designed/operated, as well as who the technology leaders will be moving forward. But, the true destructive benefits of the approach are being held back by the big players. Why?
Well, my opinion is that if you're a vendor worried about what to do with the boat loads of '90's era infrastructure that you've manufactured and sold to customers as recently as last month, you've got a problem. Do you want to explain why you sold something that was something less than the optimal solution available to your customer? Of course not. So, your best course of action is to slow the roll-out of unified computing in order to wring the last bit of revenue out of the old stuff that you can while constructing an on-ramp for your customers with the speed of a road crew.
Data center architects who are interested in building and investing into forward looking infrastructure have learned time and again that during water shed moments such as these, the best solutions come from independent vendors who are unencumbered with the past. Cisco used to be one of those companies...20 years ago. Today, look to independent company's such as Liquid Computing (www.liquidcomputing.com), the only vendor that's actually shipping product and has real customers in production.
A fully featured system can see, manage and control compute+network+storage. Anything less is just an on-ramp.