I was reading through some Cisco blogs to catch-up on what’s going on in their world for a current project of mine, when I saw this blog entry for today called “Beat this uptime” from Omar Sultan at Cisco.
They have some servers with five, seven and even nine years uptime. Great, except the utilization is so low as to not warrant the electricity they’ve used. Rather than an uptime boast, these systems seem like a great opportunity for a green datacenter consolidation and save the electricity!
Hmm, I love the smell of virtualization in the morning. I emailed Tim Sipples, it will be interesting what the mainframe blog makes of this, I know that when I was new technology architect for System z they had some customers with uptime in the 3-4 year space, running millions of transactions per day and at 85%+ utilization. I never checked for Power Systems but I suspect there are many similar out there.
[Update:] Actually if you post with stats on your system uptime directly to the Cisco blog, you can win a fleece. I missed that when I first read it!
Live partition mobility on power6 sounds like a great way to combat this on power. move large lpars that don’t do anything at night off current frames, and consolidate them as smaller lpars on a single frame. Then, when morning comes, power all the large frames up again, migrate the lpars back, and bump up the hardware. Uptime stays high, and you can turn off the unused frames at night. Same principle could apply to a TSM server, or batch processing servers that do most of their processing at night. IIRC, this was one of the major points the sales & marketing guys were pushing when power6 came out.
Yes of course, but this wouldn’t qualify for the Cisco uptime challenge since the same server hadn’t been up all that time, even if the OS or application had. Come to think of it, I don’t know what AIX, IBM i or Linux on Power would report as OS’s when they get migrated… I’ll ask.
Mark:
You make a good point about this being a case study for server consolidation and virtualization. Although, in this case the two servers I noted are in different locations and the last system is actually a Ethernet switch.
The good news, as Swimpunk points out, is that there are a number of tools like server virtualization, VM portability, etc to start reclaiming those wasted watts and CPU cycles.
Omar Sultan
Cisco
On a Power 6 System with AIX 6.1 the uptime will show how long since it was originally booted even after it has been Migrated. This could drive some nice stories of Greener Data Centers with years of application uptime while driving very high utilization on a Power System.