ZFS and Database fragmentation

Disk Fragmentation

Disk Fragmentation – O&O technologies.
Hope they don’t mind the free advertising

Yet another customer was asking me for advice on implementing the ZFS file system on EMC storage systems. Recently I did some hands-on testing with ZFS as Oracle database file store so that I could get an opinion on the matter.

One of the frequent discussions comes up is on the fragmentation issue. ZFS uses a copy-on-write allocation mechanism which basically means, every time you write to a block on disk (whether this is a newly allocated block, or, very important, overwriting a previously allocated one) ZFS will buffer the data and write it out on a completely new location on disk. In other words, it will never overwrite data in place. Now a lot of discussions can be found in the blogosphere and on forums debating whether this is really the case, how serious this is, what the impact is on performance and what ZFS has done to either prevent, or, alternatively, to mitigate the issue (i.e. by using caching, smart disk allocation algorithms, etc).

In this post I attempt to prove how database files on ZFS file systems get fragmented on disk quickly. I will not make any comments on how this affects performance (I’ll save that for a future post). I also deliberately ignore ZFS caching and other optimizing features – the only thing I want to show right now is how much fragmentation is caused on physical disk by using ZFS for Oracle data files. Note that this is a deep technical and lengthy article so you might want to skip all the details and jump right to the conclusion at the bottom :-)

Read more of this post

Managing REDO log performance


I have written before about managing database performance issues, and the topic is hot and alive as ever. Even with today’s fast processors, huge memory sizes and enormous bandwidth to storage and networks.

warning: Rated TG (Technical Guidance required) for sales guys and managers ;-)

A few recent conversations with customers showed other examples of miscommunication between IT teams, resulting in problems not being solved efficiently and quickly.
In this case, the problem was around Oracle REDO log sync times and some customers had a whole bunch of questions to me on what EMC’s best practices are, how they enhance or replace Oracle’s best practices, and in general how they should configure REDO logs in the first place to get best performance. The whole challenge is complicated by the fact that more and more organizations are using EMC’s FAST-VP for automated tiering and performance balancing of their applications and some of the questions were around how FAST-VP improves (or messes up) REDO log performance.

Read more of this post

Oracle and Data Integrity: Data in, Garbage Out?

Stop Corruption

A trivial question:

What is the basic function of a storage system?
I would say, the trivial function of a storage system is to store digital data and getting it back when you need it.

To be specific:
get the data back exactly the way you stored it.

You would probably say “Duh, of course!”

A storage system (as simple as a hard disk or as sophisticated as an EMC VMAX) is supposed to store data and give it back unmodified. But recent research shows that simple disk drives are not as reliable as you might think. Enough material is available that explains why and how often disk drives fail to return the correct information, often without any error as if the corrupted data is perfectly valid. See below for more references to this issue.

Read more of this post

Application processing at lightning performance – The hourglass view of access times

HourglassEven in these modern times, when lots of things are changing in the ICT world, some lessons from the past still hold true.

Previously, I discussed the I/O stack in a typical database environment. As virtualization has complicated things a bit, the fundamental principles of performance tuning stay the same.

Recently I was browsing through old presentations of colleagues and found another interesting view on response times in an application stack. Again, I polished it up a bit and modified it to reflect a few innovations and personal insights.

The idea is as follows. We as humans have problems getting a feel of how fast modern microprocessors work. We talk in milliseconds, microseconds, nanoseconds. So – in the comparison we assume a 1 Gigahertz processor and then scale up one nanosecond to match one second – because this fits better in human’s view of the world. Then we compare various sorts of storage on the “indexed” timescale and see how they relate to each other.

Read more of this post

Through the wormhole with Stretched Clusters

Last year, EMC announced a new virtualization product called VPLEX. VPLEX allows logical storage volumes to be accessible from multiple locations. It boldly goes beyond existing storage virtualisation solutions (including those from EMC) in that it is not just a storage virtualisation cluster – but merely a storage federation platform, allowing one virtualized storage volume to be dynamically accessible from multiple locations, as if they were connected through a wormhole, and being built from one or more physical storage volumes.

Wormhole in space
Read more of this post

Stretched Clusters – Alien storage

In my previous posts I described how Oracle ASM can be used to build stretched clusters. I also pointed to some limitations of that scenario. But I am by far not the first one in doing so – and some of EMC’s competitors attempted to build products, features and solutions to overcome some of the limitations in host mirroring.

A while ago, some guys I met from an EMC partner, confronted me with the question why EMC, the market leader in external storage and premium Oracle technology partner, had not offered a solution for these limitations. They pointed to a number of products from competitors that – allegedly – solved the problem already. Also they pointed to the architectural simplicity of these solutions.

Alien Storage

At that time I had no good answer (which does not happen to me very often). I was not aware of how these products worked and I asked some questions on that. In that period I was also confronted by our enterprise customers who started demanding an EMC solution for stretched clustering – so I started digging. Could it be that EMC was over-passed by some of these alien storage start-up companies in continuous available storage solutions? It seemed to be the case.
Read more of this post

How does Oracle keep data consistent on filesystems?

Sometimes there is questioning on how EMC can provide consistent snapshots on storage systems, given the fact that file systems like ext3, ufs and the like may keep write data in file system (server) cache.
Read more of this post