The True Cost of a Core
- 26 Aug 10
Servers are becoming more powerful as manufacturers are finding new ways to get more cores into a CPU. Today it’s not uncommon to see hexa and octa-core processors shipping at the same price points the dual- and hexa-cores shipped yesterday. Where manufacturers once got their performance improvements through raw CPU speed, they are now getting their getting the majority of performance improvement through more cores in their processor chips.
Unfortunately the economics of additional cores for performance aren’t the same as improvements through improved clock cycles because software manufactures have largely tied their technology licensing to the number of cores on a system, and their pricing isn’t decreasing as the number of cores on these new servers increase.
For example, say you buy a basic server with two hexa-core processors, so you’re looking at 12 cores on the box. Now let’s suppose the list price for Oracle Database is $47,500 per core. So your list price to run an Oracle database on your new server will be $285,000. And that’s not counting tuning packs, diagnostic packs, management packs, or even maintenance -- which is calculated as a percentage of the base price. It turns out the cheapest part of this equation may be the hardware!
So if you’re planning on running software from the big vendors, conduct a solid sizing exercise and be sure to buy just the number of cores that you need. Leave empty sockets for growth, but you might want to choose models that let you scale with fewer cores to avoid breaking the bank. Avoid sharing servers with more than one software package that is licensed per core (i.e. Informatica and Oracle DB), or you could end up paying double for server capacity that you’ll never be able to fully realize. And when you DO add cores, be sure to also purchase the additional licenses to stay in compliance. I’ve heard that software vendors’ compliance teams occasionally check up on you, and running with a few extra cores could break more than your annual budget.
Last Updated on Wednesday, 22 September 2010 06:24.
Cheap Cloud Based Backups and Disaster Recovery for your Enterprise Apps
- 15 Aug 10
Remember the old days (circa 2006) when doing disaster recovery involved shipping tapes to an off-site storage facility and provisioning emergency hardware at a disaster recovery facility? Well, fast forward to 2010 when virtually unlimited hardware and disk storage are cheap and plentiful. Making real-time backups to the cloud (with the right software) is not only doable but almost a no-brainer. But once your backup is in the cloud, the next logical step would be to launch a virtual instance of the server that you backed up. But that’s like, Star Trek futuristic stuff, right?
Well no more. CloudRepica has a Software as a Service offering that does just that. Their enterprise-grade replication software is capable of creating real-time disk images of any server. Then they combine that with the unlimited disk resources and phenomenal reliability of the cloud to create a real-time disk backup and recovery service. This offering is packaged and delivered in a Software as a Service model where you pay only a monthly fee. There are no upfront costs, no licenses, no hardware, no media, no consulting labor and no facilities costs.
So here’s how it works: CloudReplica installs their replication software on one of your Microsoft servers. This isn’t any run-of-the-mill software either. It’s an industrial-strength, lightweight, flight-level replication software that’s been in use by large enterprises for 16/17 years. Technically, it uses a non-blocking filter/driver which listens for changes at the byte-level, which means only the portions of the files that change are replicated and there are never any locks because of the replication. The changed bytes are compressed, encrypted and transmitted to one of their cloud hosting providers, such as Amazon or Teramark, where the identical disk image is maintained. Features within their replication software insure that you can always bring your system back to a good known restore point.
Last Updated on Wednesday, 22 September 2010 06:14.
What I expect from Oracle Open World 2010
- 02 Sep 10
I’m going to Oracle OpenWorld again this year, and I just finished building my schedule. Wow. This year there are so many sessions I want to attend in the same time slots that I won’t be able to see a fraction of what I want to. Guess I’ll have to skip the session on how to author blue-ray disks using Java in favor of a product roadmap session I need to attend. You see, I’ve started a new project and I have a whole laundry list of stuff I need to come up to speed on. I’m sure it’ll be an informative but exhausting week – it always is.
Gazing into my crystal ball, I’m expecting to hear more about the Fusion applications that were introduced at the end of the 2009 OOW. I think Oracle isn’t re-inventing all of the functionality in their mature ERP/CRM product lines like PeopleSoft, JDE, EBS, and Siebel. But all the same I’m expecting to see some products that are ready for launch and looking snazzy with the deep integration with BI and other apps that Oracle has invested so heavily in.
Speaking of BI, I’m looking forward to seeing the new release of OBIEE. BI apps just look cool, and their functionality makes things like PeopleSoft Matching functionality seem boring in comparison. I'm hoping to see support for a ton of data sources and the ability to publish interactive reports to latest generation mobile devices. Unfortunately I think I missed the BI boat at some point in my career, so bring on the 3-way match!
Last Updated on Wednesday, 22 September 2010 06:18.
