Today was the first of a 2-day HP ProLiant Gen8 Tech Day in Houston. The ten bloggers who were invited are all from the US except for me. Some of them I have met from other HP events like HP Discover in 2011 and an event in Madrid in the same year. Since I moved to Inovativ which is a 100% System Center & Hyper-V consultancy firm, hardware hasn’t been so much of an issue for me. You deal with whatever server, network or storage hardware the customer has. When I worked for a system integrator, HP was the one and only vendor we worked with, so I can say I know the product range quite well. It was surely a fantastic opportunity to come to Houston and be brought up to speed on the new HP ProLiant Gen8 server family.
We had a couple of really interesting sessions today, both from marketing and technical folks at HP. One good introductory presentation was by Gary Thorne (VP Strategy Industry Standard Servers and Software). Gary’s presentation was on Transforming the server experience with customer-inspired innovation. Most of the changes you will see in the Gen8 stem from over 100.000 customer interactions. Whenever there was a server down situation, HP would ask the customer to detail the problems so HP could learn from them. After all, the aim is to get the most amount of uptime with the least amount of effort, speed deployment of systems and deliver the desired performance. This is not something you get after just one round of interviewing customers. It is more like a multi-year project and it takes a large budget for R&D. HP spent $300M on R&D which is at least twice the amount spent on the previous generation.
HP has developed something they call ProActive Insight Architecture which takes much of the hands out of the operation. This required the server to have a lot more intelligence. Right now HP calls the Gen8 the most intelligent server platform on the planet. A sea of sensors collect thousands of system parameters needed to take appropriate action when problems arise. There is no agent involved which means the server starts logging right from the moment is turned on.
Deployment is also much faster and easier because all the firmware and software is already local to the server. It takes only a few minutes to deploy a server and with the right tools many systems can be deployed simultaneously. Another manual and very time consuming task is firmware/driver/software update management. The new HP System Update Manager (SUM) now does everything automatically and in the fraction of the time. It can be done for 1 system, for a full blade enclosure but also for an entire datacenter of around 5,000 servers. HPSUM handles different generations of servers and deals with HP Onboard Administrators, HP Virtual Connect and is able to coordinate reboot behavior. In clusters VM’s are moved off before a server is updated.
One particular problem HP improved on is storage connected directly to the server. CPU and memory speeds have been increasing quite substantially over the previous years. Storage has significantly lagged this trend. For this reason HP developed a new generation of the Smart Array controller optimized for solid state disks. This accounts for raw performance boosts of 100x in terms of IOPS. The controller caches data in solid state (DRAM) and finds a way to put the hottest data in the fastest place. It is able to run data in and out of cache at very high speeds. HP developed algorithms that see through a virtual machine on hypervisor platforms. Storage subsystems normally work best with sequential IO streams, but with multiple VM’s on one box this IO tends to become randomized streams. The new Smart Array controller attempts to de-randomize on a per VM basis and coordinates the reads and writes in an attempt to avoid latency.
The new Insight Online website helps customers to track their server statistics as well as the server’s warranty information. It maintains a complete snapshot of the server and makes the support effort a lot easier. This service does not come at additional cost. At the same time all this collected information about the server’s health, can be the starting point for Pro-Active Support and to catch problems before they happen. It uses compliance analysis of the system components and makes suggestions for improvement.
Over 150 design improvements in HP ProLiant Gen8 were inspired by customers. One nice example is the Do-Not-Remove LED on the disk drives which are switched on if a logical drive has run into a problem situation. Let’s say a RAID-5 has a drive failure. In that case you want to make sure you do not pull one of the remaining four disks. These four drives will have the Do-Not-Remove light turned on. This greatly reduces the chances that an engineer pulls the wrong disk drive and destroys the entire RAID set.