c0t0d0s0.org

One of the favorite misconception about our UltraSPARC T1/2 systems is the "Each core has 1.2 GHz, so divide it by four threads, each core has 300 MHz"-meme. This meme was introduces by our beloved market companions like HP or IBM. At first, this sounds quite intutive, but do you really think that such a processor would be the fundation of the fastest single socket SAP system? Denis Sheahan explains in Lesons learned from T1 the misconception behind this FUD:

This line of argument doesn't hold because most commercial code chases pointers and is constantly loading data structures. On average a commercial application stalls every 100 instructions for a variety of reasons such as TLB miss, I cache miss, Level 2 cache miss etc. When a thread stalls it is usually delayed for many cycles, an Icache miss for instance is 23 cycles. So even though a thread is running at 1.2GHz it usually spends 70% of its time stalled. This is why major processor manufacturers create ever deeper out-of-order pipelines in an effort to avoid this stall.
All this stalling is perfect for CMT. The hardware automatically switches out a thread when it stalls and shares its cycles amongst the other 3 threads on the pipeline masking the stall. With this technique we can utilize the pipeline 75% - 80 of the time provided there are enough threads to absorb the stall

Even this explanation leaves out some cases: The cryptographic units work largely parallel to the pipelines. Thus the computational power of one core is even higher than you would assume from the pure frequency. To say it a little bit simplified: For cryptographic workloads you can see this processor as an 64+8 threads and 8+8 cores system, albeit the 8 cryptographic cores are specialized ones ...

There is a slight but extremely important change in the licensing modell of Oracle. The Multicore FAQ states:

Oracle Database Standard Edition can only be licensed on servers that have a maximum capacity of 4 sockets. When using RAC to cluster this limitation is mirrored in the cluster. It may be licensed on a single cluster of servers supporting up to a maximum of four sockets.

What does this means: Our workhorse server V490 is now usable with Standard Edition. So you can save a vast amount of money. Instead of 3 Licenses (each core is counted as 0.75) Enterprise Edition you need only three licenses Standard.

Or you can build an incredible powerful OLTP RAC-Cluster with 4 T2000 servers. Although they have 32 threads there is only one single socket per Server. 128 parallel threads = 4 UltraSPARC T1 = 4 Sockets = within the limitation Standard Edition. Standard Edition = RAC included.

And now the kicker: Niagara 2 equals 64 threads per socket. Means 256 threads per 4-node cluster. But: Still only 4 sockets. And you surely know it: Still Standard Edition.

The OpenVZ project announced the port of OpenVZ to the UltraSPARC T1. Choice is good, but we have Container inside of Solaris right now, LDOMs in the next weeks and XEN in production solaris within this year. I don´t really see the point for an additional virtualisation system on this processor besides of the "been there-done that" factor.

Diese Präsentation muss ich mir unbedingt besorgen. Der Vortrag hat wirklich umfassend beschrieben, wie man mit der SSL-Beschleunigung der UltraSPARC T1 CPUs umgeht. Ich werd einfach mal fragen, ob ich den Inhalt ins Deutsche uebersetzen und in mein Wiki setzen darf. So ... nun gehts weiter in einen Vortrag ueber ZFS

After we open-sourced the design of the UltraSPARC T1, i´ve predicted here in my blog and at customer presentation, that we will see stripped-down version of the the T1 sonner or later for the embedded market. And the "sooner or later" is right now. The processor is called S1 and can be obtained unter GPL at Simply Risc. It´s a slightly extended single-core version of the T1.

Since the integration of the floating point coprocessor in the x86 processors nobody really thought about floating point. Now they are ubiquitous and a high percentage of users think about it as nescessary. Joseph Darcy talks about what every computer programmer should know about floating point arithmetic. After this you understand, why floating point is quite unusual in many commercial software. Floating point arithmetic is such an unaccurate stuff, unusable when both sides of your quarterly report has to match. This fact enables the development of a processor (the UltraSPARC T1) with only one FPU without beeing slow in many tasks.

Normally i don´t quote our own press releases, but this one is really interesting: Sun Fire T1000 Server With CoolThreads(TM) Technology Outperforms All Competing 1U Systems on the Market

The third operating system is available for the T1: FreeBSD is selfhosting now. It´s stable enough to run through a make buildworld. It´s facinating how fast the ports are done to this architecture.
I think that the delopers see the T1 as an very attracting target. And additionally the ports are done against sun4v, not against the T1. sun4v was developed to build an abstraction between processor hardware and operating system, so the work on kernel ports should be done for the foreseeable future.

Here is another T2000 review. This one has mysql as the focus:

Man our T2000 rocks out for MySQL serving. It requires a bit of tuning to get going, but it our scales our existing Opteron clusters 4-20 to 1.

itweek resports about about Jonathans statements regarding support of Ubuntu and UltraSPARC T1. Well, i would still use Solaris (alone for dtrace and zones), but this really opens another market for the T2000.

news.com reports about Colms Benchmarks on the T2000. Most interesting comment from Colm cited in the article:

The Niagara servers only consumed about 220 watts while the Itanium sucked up 400 watts. "We were kind of surprised about the power consumption" on the T2000, he said. "I thought maybe my power meter wasn't working."

When you don´t believe you us or a vendor-supplied tool, when we talk about the need of energy efficent servers, talk with your HVAC guy.

When talking with the HVAC guy before he left he said something interesting. "Don't buy Dell." He's not a server guy, he's not in IT, he's an HVAC guy. "They run hot as hell," he noted.

Colm made an additional benchmark. The T2000 seems to be faster on Linux than on Solaris, but I assume that he hadn´t switched on the write cache of the discs. This can make quite a difference. On Linux the write caches are activated by default.

AMD, EPA and Sun are working together to define a metric to compare the energy efficency for servers. Perhaps this will end the brute-force-tactics (mooooooore frequency, moooooore) in server design and benchmarketing with unrealistic configuration (more than thousand disks for a good TPC-C-Benchmark will ruin your energy efficiency benchmark ).
Perhaps Sun rides with the UltraSPARC T1 on the first wave of an idea, whose time has come.

IT Jungle reports about Niagara and Niagara 2 and makes an interesting point:

Moving workloads does not necessarily mean companies want to take their operating system with them, and a lot of shops that have dumped Solaris for Linux might be fairly easy to convince to dump Linux for Solaris--if the price, thermals, and support are right.

A company which made the homework at the last migration should have no problem to migrate back to Solaris. So : You have no argument for the further usage of Linux