Oracle Cloud : 17 Months Later

I had a couple of comments on a previous email from an Oracle salesperson. In response to one of them I linked to my Oracle Cloud : First Impressions post. His reply to that was,

@ Tim. You are referring to a review from August 28, 2015? Facts from 17 months ago-Seriously? #OracleCloud has changed *significantly* in 17 months. Suggest you relook/re-evaluate your research.

So here is the update of the stuff I said 17-ish months ago and the state of things as I see them now. Remember, this was a direct challenge from an Oracle salesperson.

Oracle Cloud : Look and Feel

At the time I was pretty positive about this. At first there were some inconsistencies in the navigation menus. They later made some changes to make things more consistent and I reflected that in my article in an update.

My biggest problem at the time was how slow the interface for Oracle Public Cloud was. Well, 17 months later it is still slow. Sometimes taking 30+ seconds to respond when you click a tab. I’ve heard people say, “It’s your internet connection!”, before. Nope! Same at work (a University with a scarily fast internet connection), home and at Oracle HQ. 🙂 In comparison the Azure or AWS the Oracle Public Cloud interface feels slow and unresponsive.

Oracle Cloud : Ease of Use

I was also pretty positive about this in my initial review. Same here really.

My major gripe was with the firewall interface. If you access this via a service like DBaaS, it does limit the view of the rules to that service now, which makes it more usable. Before you could only edit them from the Compute > Network section, so you saw all the rules for all your services. That was a nightmare.

You still can’t edit the rules, with the exception of enable/disable. I don’t like this. It feels like someone has put a wrapper over basic firewall commands, rather than thinking how a user might like to use this interface. Don’t get me wrong, the firewall interface is better than it was, but it still feels like a botch-job to me.

A quote from someone who shall remain nameless,

My router at home has a better firewall interface than Oracle Public Cloud.

Remember, I’m not talking about the security of the firewall. I’m talking about the interface here!

DBaaS Specifically

No multiplexed redo logs. That was the case 17 months ago. That’s the case today. I don’t consider an instance without multiplexed redo logs fit for anything but playing. It is an Oracle recommendation to multiplex redo logs. See here.

Oracle recommends that you multiplex your redo log files. The loss of the log file data can be catastrophic if recovery is required. Note that when you multiplex the redo log, the database must increase the amount of I/O that it performs. Depending on your configuration, this may impact overall database performance.

So before you use DBaaS for anything other than messing about, multiplex those logs.

Management Fragmentation. That was the case 17 months ago. That’s the case today. Some of the service management is done in the main cloud interface. Some is, or can be, done in the DBaaS Monitor application, which has improved, but I’m still not really a fan. For other tasks you jump across to [DBConsole – 11g | DB Express -12c]. This is still a fragmented approach to management and doesn’t feel like a consistent DBaaS offering to me.

Firewall. I’ve mentioned this before. There have been improvements in this area, but I still don’t like it. DBaaS still builds a bunch of disabled public rules which you would be insane to enable. I can only imagine that they are there for inspiration. As mentioned previously, you can’t edit them. So my comment from 17 months ago stands. Do it properly or don’t bother.

Access to Management Interfaces. My original point was, because of the fragmentation of the management interface explained above, some of the management functionality is not available until you define firewall rules to access it. This is because of a design choice. It felt wrong then. It feels wrong now.

Access to OS. This is very much *my opinion* and I’m sure many others will not agree. I feel like a Platform as a Service (PaaS) offering, which Database as a Service (DBaaS) is a part, should separate you from certain bits of the infrastructure. I originally thought access to the OS and therefore the Oracle installation was a good thing. As I started using the system it felt too much like Infrastructure as a Service (IaaS). I would prefer a fully managed service, which Oracle are going to offer at some point in the future. Like I said, just my opinion of what PaaS and IaaS are… So my opinion remains unchanged since the original post. Please don’t comment about Schema Service or Exadata Express as exampled of fully managed services. I don’t need to get into another fight!

Oracle Database Cloud Service – Virtual Image. I thought it was a waste of time when I tried it. I still think it’s a waste of time. If I’ve got to do everything myself, it’s not DBaaS. Get rid of this option and make me use IaaS.

Conclusion

That article from 17-ish months ago is still pretty much on the button today.

Oracle Public Cloud has certainly changed a lot over that time, but most of the points I raised all that time ago have not been addressed, or at least not to my satisfaction.

These issue were raised in that original blog, directly using a private feedback space set up for Oracle ACEs, and in some cases directly with PMs. I believe I’ve done everything I can as a self-proclaimed fanboy to get this stuff sorted.

I’m sure, based on my previous post about the cloud licensing policy change, some people will see this as me angling for a fight, but I’m really not. I had no intention of writing this post until I was directly challenged to do so by this sales guy. If he had taken the time to check his facts before issuing this challenge, I guess he would have kept his gob shut!

Cheers

Tim…

Author: Tim...

DBA, Developer, Author, Trainer.

9 thoughts on “Oracle Cloud : 17 Months Later”

  1. It would have been so nice if the Oracle sales rep had taken the time to deconstruct your 17 month old review and replied to each observation with ‘True but this has subsequently been modified and performance/whatever is now 3 times quicker’ instead of a blunt ‘Suggest you relook/re-evaluate your research’.

  2. Hi Tim,

    Really good staff as usual and I share a lot of opinios with you on these points.

    I dont know if you have already written about the performance of:
    Oracle Storage cloud service & Oracle compute cloud service – Block Storage

    when running on IaaS,

    Did you?

    If so, could please point me the articles?
    I found no docs about the experience of this when running Databases on IaaS.
    Oracle also does not give any metrics to rely on

    Best Regards!!!

  3. Hey Tim,
    oh well … it is also dead slow for me (EMEA2) and these are not the only issues with Oracle Cloud.

    Some examples from a few days ago:
    * https://twitter.com/OracleSK/status/824943337497034752
    * https://twitter.com/OracleSK/status/824951068949876736
    * https://twitter.com/FranckPachot/status/826505335833817088

    I also replied to an Oracle sales person who wanted to know how satisfied i am with the Oracle Cloud. I wrote him a detailed feedback with screenshots of issues, etc. but never got any response / feedback! This is how it works nowadays!

  4. Andy: That would require thought… 🙂

    Fernando: No. Not written about it. I’m doing functional testing, not performance. I’m using trial accounts. 🙂

    Stefan: In my experience, it can take up to 24 hours to notice you’ve released some resources. I’m not sure if you would be charged for these… 🙁

    Stew: I think I’m on about -5000 points after the last couple of days. 🙂

    Cheers

    Tim…

  5. Great re-review Tim. I wonder if Oracle want to discuss any of this with you further. I’m still surprised at the changes in the pricing policy. Will be interesting to see if Non-Authorised Cloud environments will be able to take advantage of this change?

  6. I’d be curious on hearing from anyone that has been “saved” by multiplexing their redo logs on any form of storage that was mirrored at some other level (OS, hardware etc).

  7. Connor: As you know, since the bad old days of using non-RAID disks, multiplexing has never been about disk resilience. It’s about protecting against file corruption and user error, both of which I’ve experienced in my time. I’ve actually experienced a file system corruption issue in my current role, which affected some data files and one copy of the redo logs, so yeah. Been there and done that. Doesn’t happen often though, I grant you. Can’t remember if I’ve seen a user error issue with regard to redo logs, but have seen it from a data file perspective. 🙂

    I agree, you can argue the relative risks of this, and certainly it will vary depending on the storage layer and experience of the users etc. Having said that, and read a number of pros & cons discussions over the years, I’m still of the opinion that if you are trying to protect the whole chain to minimise the chances of data loss in a failure situation, it makes sense to multiplex the redo logs. Only if that becomes a specific issue from a performance perspective should you consider an alternative. Other opinions will vary of course…

    It does say something really important about Oracle Cloud though. There is an explicit mention in the docs that you should multiplex, then the DBaaS service doesn’t take that advice. Then on the other side we are told that Oracle does Oracle in the cloud better than anyone else, but doesn’t follow their own “recommendation”? I could mention other “recommendations” that DBaaS doesn’t follow, but that would just be me being bitchy. 🙂

    Also, DBaaS doesn’t stop file system access, so a service isn’t protected by Oracle’s “superior management”. A customer can still get on the system and do something stupid if they have less experience. I would suggest protecting them from themselves to some degree would be sensible.

    Cheers

    Tim…

  8. Some of my gripes were
    1) Default file system for the Database (non-RAC) was not ASM. ASM makes it more difficult to delete redo log members
    2) Adding space to the DBAAS offering required a reboot. This was 5 months ago and I haven’t checked this since then
    3) Extreme performance with an ext4 filesystem would seem paradoxical.

Comments are closed.