New Sun T2 based servers

by Mads
October 21, 2007 at 23:40 | categories: hardware, sun

Recently, Sun announced their latest CoolThreads servers based on the UltraSparc T2 processor.
If you were familiar with the UltraSparc T1, then the new T2 is fairly similar, but without most of the limitations from the T1. Some of the most interesting news are:

  • 64 threads (up from 32, still 8 cores)
  • 8 fully pipelined floating point units (up from 1)
  • 8 x crypto accelerator
  • Dual 10Gbit Ethernet and PCI-E integrated onto chip (and the crypto accelerator to feed it full speed)

The 3 machines are:
T5120 - compares to the T1000 and fixes two of my largest complaints about the T1000 by having a redundant power supply and making room for 4 2.5" sas drives. All in a 1U package.
T5220 - compares to the T2000. Not much else to say other than there now being room for 8 2.5" sas drives. I rarely think I'd pick a T5220 over a T5120 unless I needed the extra internal drives (very unlikely) or wanted the 1.4GHz model.
T6320 blade - a blade version of the 5120 which is quite interesting because Suns blade enclosures let you mix and match between UltraSparc T1, UltraSparc T2, AMD Opteron and Intel Xeon based blades. Unfortunately, the 6320 blades appear to be unavailable for the time being.

At the ASF we have two T200s - see Out with the old and in with the new. Eos and Aurora are quietly working along through a decent workload and we're quite happy with them (allright, so Eos is very busy and Arora is mostly there as a backup, but redundancy is a good thing). The load usually only goes high when bots run wild or similar forms of abuse hits, but most times we manage pretty well.
There is however another area where an UltraSparc T2 based server could do a whole lot of good and that is on our Subversion repository. Keeping it afloat is currently a bit of a task as there seems to be popping up more and more stupid Continuous Integration tools that keep hitting us like there's no tomorrow. There seems to be no end to the silliness such as trying to trawl the whole harmony tree every 30 seconds looking for updates (one bad example had 2 ips belonging to a large micro......... company hitting us with between 400.000 and 1 million requests/day). It doesn't often affect other users of subversion, but it would be very nice to be able to keep up with the load a little better and not have to firewall early on when we get hit. A T2 based server would give us a whole lot more headroom than the current dual processor box and thinking about moving everything to SSL is no longer utopia.
I'm sure it could keep up with the load (as long as we can find a suitable storage array to duct tape it to) as the performance figures from bmseer are looking pretty darn amazing.

Looking at the US T2 based servers from a work perspective, I'm also expecting to see a few of them in the near future (if I get a say). There are plenty of candidate systems on older and much less efficient platforms.


Sun SPOT and Project Looking Glass

by Mads
October 21, 2007 at 21:22 | categories: hardware, sun, solaris

Usually graphical noise such as Project Looking Glass really doesn't interest me one bit. And the first time I looked at it, I did find it rather pointless as it tends to make things rather confusing for something as simple as showing a couple of terms and a browser.

Sun SPOT or Sun Programmable Object Technology. A really cool programmable toy that I've often been wanting to buy (but can't because they don't ship outside .us). The downside of course being that programming them you have to use java, but I suppose it would be worth the pain to get access to the range of sensors such as accelerometers that Sun SPOTs have.

So, how does the two relate? (via).
The videos below show a demo of just how to put the two together:

If I wanted a Sun SPOT before, I really want one now!


Out with the old and in with the new.

by Mads
August 05, 2007 at 23:20 | categories: hardware, sun, solaris, asf, httpd

For quite some time, the infrastructure team at the ASF has been running our websites, mail-archives and wiki on a Sun Fire T2000 Server kindly donated by Sun. Along with the T2000 there's also a Dell SATA raid donated by ask.
Naturally, the machine is running Solaris 10 and that along with dtrace has already allowed us to find and correct pretty serious performance issue. Our load was hitting 500 and beyond and was close to knocking the machine over. Some digging around with DTRACE showd us an insane number of syscalls and almost all of them being reads.
More digging around with the following one-liner by Brendan Gregg:

# Read bytes by process,
dtrace -n 'sysinfo:::readch { @bytes[execname] = sum(arg0); }'

It gave a very clear picture that almost all reads were of 1k size and that allowed Joe Schaefer to create a patch for apr to Use buffered I/O with SDBM..
The current look of things is a lot better:

  httpd                                             
           value  ------------- Distribution ------------- count    
              -1 |                                         0        
               0 |                                         987      
               1 |                                         0        
               2 |                                         6        
               4 |                                         296      
               8 |                                         30       
              16 |                                         147      
              32 |                                         130      
              64 |                                         47       
             128 |                                         140      
             256 |                                         460      
             512 |                                         118      
            1024 |                                         19       
            2048 |                                         72       
            4096 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 336511   
            8192 |                                         11       
           16384 |                                         3        
           32768 |                                         0        
           65536 |                                         8        
          131072 |                                         0        

With the change, our load has dropped from over 500 to somewhere between 5 and 10.

For a long time we've also been wanting to add some redundancy by placing a similar setup at our European site. The board approved our request to go shopping and after lots of hassle trying to buy the machine from Sun (being a small customer at Sun is rarely much fun and I think we were even more unlucky than usual). Eventually we got there and Sander along with Colm got the machines racked.

The pictures are by Colm
before

after

The upper picture show the old Itanic and a broken X-serve. Below is the "after" picture, showing Aurora which is now the European mirror of Eos. The machine above Aurora is a Sun Fire X2200 M2 Server that will serve as a mail frontend.

and so ends the tale of how the rising Sun replaced the sinking Itanic :)

Sunny Day

by Mads
April 13, 2006 at 20:51 | categories: hardware, sun, solaris

Tuesday last week I went to a Sun thing titled "Powering the Participation Age". Most of the day was spent chatting to Sun reps and listening to a couple of semi technical presentations. One of the presentations covered the T2000 servers and was quoting Colm MacCárthaigh's and his T2000 testing. Of course I couldn't keep quiet and had to mention that the numbers were probably far from what Colm would get in later testing and his latest posts adds about 1/3 by upgrading to Solaris Express, but the really amazing thing is that he gets better performance running Linux on the T2000!

Scott McNealy also made an appearance and did a very good job at explaining what Sun has been doing recently and why. For the first time, I began to understand where they're going and I'm not as worried as I used to be. Only time will tell wether they will succeed. As usual, Scott managed to make fun of the company that I won't be working for much longer, maybe I'm biased by having made the decision, but he wasn't far off the mark - if anything, things are probably worse than he joked about :)


Mysql on the T2000

by Mads
April 07, 2006 at 22:01 | categories: hardware, sun

T2000 Ian Holsman has a nice writeup of the benchmarks he has been doing with the T2000 in "How does MySQL perform on a sunfire?".
I've still got plenty of ideas for sqeezing a bit more out of the box, but I don't really have much time to test it myself before I don't have access to the T2000 Try & buy at work.


Next Page ยป