Archive for October, 2011
Its that time of the year
When its time to start booking hotel rooms and planning your agenda for the UKOUG conference in Birmingham (UK) for the 4th to the 7th December. Yep, all the cool sessions from Oracle Open World, minus an awful lot of the pure marketing. So just to wet your appetite here are my highlights in advance .
- Oak Table Sunday. A whole day of wit, wisdom, argument and education from World Renowned database experts. Probably beer as well.
- International Speakers who rarely come to Europe. Connor McDonald, Greg Rahn, Maria Colgan, Iggy Fernandez, Cary Millsap, Marc Fielding
- UK speakers who’ve been there, done that and wear the scars. (too many to mention)
- Key Notes worth attending – Ray Wang, Andrew Mendelsohn, that man Cary Millsap again.
- The Balti Triangle
- The Oak Table Challenge
I’ll be speaking too, an entry level “things really, really not to get wrong” session, and a kindly scheduled advanced session on statistics, data lifecycle, obscure optimizer issue and er getting decent execution plans at 16:15 on the last afternoon that last one might end up being a pub discussion if I’ve anticipated numbers correctly (0) .
Anyway, especially if you’re employer is a UKOUG member, you should be there. If you want a flavour of OakTable discussions – arrive on Sunday and book.
Possibly Related Posts:
First Impressions of EM12C
One of the major announcements at Oracle Open World last week was the launch of Oracle Enterprise Manager 12c, though I’m going to refer to the product as em for the rest of this blog. EM is a product that I both love and which completely infuriates me from time to time. Its worth understanding my reasons for this attitude before we look at the new release. First up then why I love it.
- Database Performance Pages
The database performance pages, also available in database control of course, by and large focus on exactly the right things, namely response time and throughput. Moreover they provide a clear picture of database performance that is straightforward for humans to understand. A good picture almost always gives clarity and understanding more quickly and to a wider audience than text, as Florence Nightingale so eloquently understood about 150 years ago. The problem with many data visualisations is that they hide the core message, for example here. The em performance pages nearly always show the right thing clearly. - Central Repository. So many infrastructure management tasks benefit from a central repository of observations. I’d hazard a guess though that most database infrastructure management is still done via scripts. This approach means that items like capacity planning, comparison of time periods and so on are not readily available or rely on the knowledge of the individual administrator both of the product and the environment that they are administering.
- Management templates. In my experience many environments end up with either different versions of common scripts monitoring their infrastructure or else different subsets of scripts monitoring the infrastructure.
In short then I love EM because it enables me to roll out clear, consistent, sensible management across a client, and moreover to use nice graphics to communicate deep technical issues clearly to non-technical staff. What’s not to like? Well actually this list of pet peeves.
- Navigation. The navigation in EM10 and 11 is, frankly, appalling. Multiple lists of links, the same list in different orders on different pages, the same page having 2 identically labelled links going to different locations. Then there’s the use of the back button or rather the lack of a reliable back operation – combined with the MOS experience someone, somewhere needs to tell Oracle that breadcrumbs are not the only navigation aid.
- UI. I love the database performance pages as I said, yet when you navigate to the performance pages for non-oracle targets you get a completely different experience with different graphics, often focussing on different things. Oh and using a different technology and with a different look and feel. UIs really do need to be consistent.
- Support. Historically a new release of the database, or a new patch version of SOA has resulted in your Enterprise Monitoring solution being uncertified against your enterprise technology stack – and in Oracle support refusing to take calls.
- Security Configuration. A lot of the Oracle inspired articles out there seem to imagine that the infrastructure of your typical enterprise has no firewalls, common passwords, ubiquitous sudo access and so on. So for example you are expected to be able to log on remotely to database servers from the management server or servers as SYS, the firewalls are expected to allow network traffic through on all listener ports from the OMS and back to the OMS on various upload ports etc. Deployment assumes that sudo is available, passwordless ssh is permitted and so on.
This release brings a number of significant architecture and UI changes. I thought it would be useful to evaluate the new release, and especially compare it to the lists above. My usual approach when looking at a new Oracle product is to fire up a new CentOS vm via Virtual Box . Then in conjunction with the documentation available at the OTN docs site. For EM10 this is a perfectly reasonable approach. With em11 I also was able to get away with it. For em12 the minimum specs haven’t changed much from 11, but you really do need them. That means you will want
- A database server with at least 2g ram available for the db.
- An application server with at least 4gb ram available.
The poor old laptop that I have here wasn’t going to cut it. I did manage to successfully install the product on a single server with 2gb ram, but the install itself took 5 hours.
For the purposes of this exercise therefore I setup an Amazon AWS VPC environment as follows.
- em repository machine – type m1.large which means 7.5g ram and high i/o capacity.
- em app server – type m1.large
- db target server – also type m1.large though I could have got away with a small server here.
overall I left this setup running for 3 days and with the CPU usage and data uploads this ended up costing me ~ $70 including taxes. Its likely that most test instances of em12 will have to end up on production hardware/vms. Its also worth noting that the installation still took 1.5 hours. I’ll not cover the installation here because
- its relatively straightforward.
- MOS has a note on it (MOS login required)
- Martin Bach covered it here (OEL 5.7)
- Sve Gyurov covers it here (OEL 6.1)
The main points to note are that the installer is much, much simpler to use, Weblogic 10.3.5 is not separately required but that a certified database is. Unlike in previous releases the installer will correct various of the common pre-requisite failures for you and allow you to correct others after the event. When the product is installed however you discover the major architecture change in EM 12. EM12 effectively now provides a monitoring and alerting framework and monitoring plugins that are separately developed and maintained provide the actual target management functionality. Even MOS integration is a plugin rather than a core feature now.
How do we do against my pain points then?
Navigation.
EM12 provides a menu driven application navigation style rather than the sea of links management style. Moreover, unlike the sparse adoption of this in em11, this navigation is pervasive. I at least won’t be left scanning the columns of links at the bottom of the page thinking “I know the all metrics link is here someplace”.
In addition em12 introduces the concept of job related home pages, so a DBA can setup a Databases home page, an IT manager can have a management overview and so on. This functionality was available via groups in earlier versions but required manual setup and careful thought.
UI:
I think the best way to illustrate this is with a couple of screenshots. The first is the new database home page.
I hope you’ll agree its clear, straightforward and well laid out. The lists of links are banished to be replaced by a neat menu structure shown in action below.
overall then the ease of use experience as compared to prior versions is a fantastic improvement.
Support:
Here I can only give a qualified thumbs up since we will not know for a while how the new plugin architecture works. In principle requiring product development teams to update the plugin with each new patchset/release should result in a much better ownership experience, especially for thise customers who have been told in the past to upgrade their enterprise monitoring system to stay certified when they apply a patchset or point release in order to bug fix the technology product they really care about. In practice it wouldn’t surprise me to see plugins trail product updates by some considerable distance.
Security:
Here things are much as they were in em11. I still think that in many organisations the required configuration will mean much rewriting of security policies in order to allow the management host to connect through corporate firewalls, run privileged o/s commands and so on. Bear in mind also that to make the best use of the MOS integration many organisations are likely to end up exposing the host that has these privileges to the internet.
In conclusion I am very much impressed with em12 , the architecture and UI decisions look to be broadly the correct ones and the areas of weakness appear to be being addressed.