Analysing a so-called "Comparison" about Virtualisation at IBM Developerworks
Whenever you want to dismiss the claims of a competitor or want to set your own or prefered technology in a better light, you should do some research on your topic. Otherwise you may end up with a document that´s outright ridiculous.
I found a really strange piece of “comparison”. It´s called “A comparison of virtualization features of HP-UX, Solaris, and AIX”. It´s written by Mr. Ken Milberg. And i wasn´t able to stop my shaking the head in disbelive. This text reinforces my personal impression, that this author is just a hired gun to publish claims even IBM doesn´t want to make. But let´s dissect his newest blurb. You shouldn´t read it … it´s just a really abysmal document. I´ve sacrified my time to do it for you, so don´t waste your own ;)
I will start with just four sentences of Mr. Milbergs document:
- Scalability -- Only eight CPUs and 64GB RAM on one machine
- Server-line -- Only low-end Sparc servers are supported
- Limited micro-partitioning -- Four partitions on T1, Eight on T2
- No Dynamic allocation between partitions
The truth is:
- Scalability: 4 sockets, 32 cores, 64 pipelines, 256 threads, 512 GB of memory
- Server-Line: I would call the Sun SPARC Enterprise T5440 not really low-end SPAC services
- Micro-Partitioning: Up to 128 LDOMs on a T2+ system, Up to 64 on a T2 and 32 on a T1.
- Of course you can resize the LDOM without rebooting the system. At the end this is the way, you initially configure the system: At the beginning all CPUs and memory resources belong to the control domain, you take them away from this domain to give them to the guests.
Four claims, four times utter bullshit. I could stop the dissection now, but the article starts to get even more funny. So let´s go ahead:
Sun also offers hardware partitioning, which allows their high-end servers to be divided into four-process partitions. These are referred to as Sun DSD's. In many ways this technology is similar to IBM logical partitioning, which was introduced in 2001, with no real virtualization capabilities.
DSD was the name of the technology in the Sun Fire Enterprise line. You were able to split those systems at the granularity of system board, thus the granularity was 4 sockets per domain minimum. That´s correct.
But the Sun Fire Enterprise almost reached the end of its lifetime and now we sell the M-Class systems. There is something called quad-XSB mode. The documentation for Dynamic Reconfiguration on the M4000/M5000/M8000/M9000 states:
SPARC Enterprise M4000/M5000/M8000/M9000 servers have a unique partitioning feature that can divide one physical system board (PSB) into one logical board (undivided status) or four logical boards. A PSB that is logically divided into one board (undivided status) is called a Uni-XSB, whereas a PSB that is logically divided into four boards is called a Quad-XSB. Each composition of physical unit of the divided PSB is called ane XtendedSystemBoard(XSB). These XSBs can be combined freely to create domains
In a M4000 you can create 2 partitions, in a M9000 you can create up to 24. Despite the statements of Mr. Milberg you are able dynamically move resources from one domain to another. This is a really old trick. I´ve done this on one of my E10K in 2000 and the system with the capabilty of creating 16 DSD. And this system was introduced in March 1997. So Mr. Milberg´s comment “… IBM logical partitioning, which was introduced in 2001 …” is good for some amusement. At the end of the Sun part of the document he even starts to celebrates the advantages of WPARs without mentioning the disadvantages. To close my article: The whole article is an insult to the real meaning of the sentence “This article explores all of these topics in detail.” I´m not really sure how such an article was able to pass the editorial quality control at IBM. PS: Okay, when you really want to read this botch job … here is the URL: A comparison of virtualization features of HP-UX, Solaris, and AIX