HP goes sweet sixteen with HP-UX 11i update
Serviceguard but no leather boots
The established Unix operating systems change at a glacial pace these days, but they do get tweaked from time to time. In the case of Hewlett-Packard's HP-UX 11i v3 operating system for the company's current Itanium and legacy PA-RISC servers, HP-UX gets an update every six months or so, with Update 4 debuting tomorrow.
The changes are not huge in Update 4, but they will nonetheless prove useful to HP-UX shops. Most of the useful changes come in an add-on for HP-UX called Serviceguard, which provides high availability clustering for up to 16 server nodes. But regardless of where the tweaks have been made, HP has focused on improving uptime and cutting down on the work admins have to do.
HP-UX 11i v3 Update 4 also includes an online download option, which doesn't seem so radical if you are used to using Linux, but old school operating system providers have been hung up on media distribution for some strange reason. Anyway, customers can now get an HP-UX license in a few hours from HP's replicated servers instead of waiting a few days for it to arrive by mail.
To further make the life of administrators easier, HP has extended the online patching and updating features of its Unix operating system with a new feature called dynamic root disk, which allows for an admin to spawn a new partition on an HP-UX box, plunk a copy of the operating system there, and point updates at this secondary partition.
Once the patches are applied, admins can do testing to make sure it is working, leaving end users banging away on the old partition until the techies are satisfied that the new partition is solid. When they want to move customers over to the new partition, they just reboot the box and start the machine on the new partition, and if something has gone awry, they can reboot and be back on the old partition. HP says that this dynamic root disk can cut the time downtime associated with an HP-UX upgrade in half.
Update 4 includes a disk scrubbing feature that allows information to be truly scrubbed off disk drives, with the data being randomised, rather than simply having pointers to the data removed, which is what operating systems normally do when you delete files.
According to Brian Cox, director of Business Critical Servers software planning and marketing at HP, companies like to repurpose their Unix boxes as they add new machines, but there are all kinds of compliance issues regarding data scrubbing. With it costing as much as $150 per drive to have a disk drive scrubbed, this can add up to money in a heavily configured system. HP-UX now has its own data scrub feature, which is free to anyone with an HP-UX 11i v3 support contract, and because customers tend to have a set of disks for the HP-UX operating system and another set for data, the data disks can be scrubbed while the OS disks are left alone, which means admins do not need to reinstall the operating system - also a pain in the neck.
On the clustering front, the Serviceguard A11.19 clustering extensions to HP-UX 11i v3, which is part of a set of HP-UX systems software that HP calls the Virtual Server Environment, is getting a new feature called fast failover. This feature was custom coded for two-node clusters for selected HP-UX customers who needed the absolute smallest recovery time if a server failed - think banks and stock exchanges - and is now being mainstreamed. The fast failover feature can now span 16-node clusters, not just two-node setups, and can reduce the failover time from 30 seconds with the prior Serviceguard release to 4 seconds.
Serviceguard has also been tweaked so the planned downtime associated with cluster maintenance and configuration - such as adding or deleting cluster resources, changing parameters of application software while the application is running - no longer results in downtime. According to HP, the U.S. and Japanese governments have mandated that their agencies support the IPv6 protocol, and Serviceguard A11.19 supports IPv4 and IPv6 networks, or networks that mix environments as customers transition to IPv6.
Finally, HP is announcing a feature for some clustering extensions dedicated to the Serviceguard-SAP combo called, appropriately enough, Serviceguard Extension for SAP. This new feature provides failover clustering for a key feature inside SAP's supply chain management modules called LiveCache.
The supply chain module is a big part of the SAP ERP application stack, particularly for manufacturers who use SAP software to manage their shop floors. The supply chain database was a performance bottleneck for the SCM modules as shop floor workers and the ERP suite banged away on it, and so SAP took an instance of its open source MaxDB database (which is one of many possible backends for the MySQL database management system) and embedded it in memory on systems running SAP's ERP software to house that SCM database. This is what LiveCache is, and it significantly improves the performance of the SCM software. But sometimes this database gets corrupted or the machine supporting it has a failure of some sort, and when it crashes, the system is bollixed.
So now HP is using Serviceguard to keep a fully replicated spare copy of the MaxDB database on another HP-UX box, and the SAP suite just points to that backup copy if something goes wrong with the primary. Now, instead of taking hours to rebuild that SCM database, recovery happens in about 2 minutes. LiveCache clustering for SAP ERP works with ServiceGuard Extension for SAP versions 4.51 and 5.0 and with LiveCache 7.5 or later from SAP, and works with HP-UX 11i v2 or v3.
HP-UX 11i v3 Update 4 and Serviceguard A11.19 are both available on April 17. ®