· Technology
 · · Servers

Meat-Grinder! · It’s days like these that make it fun work­ing for Sun. The new server’s of­fi­cial name is the T5440; they call it a “mid-range” box, but to me it looks like a mon­ster; count the num­bers for cores, thread­s, RAM, and so on. It’s as­tound­ing what you can fit in­to a 4U box these days ...
It’s Slow · The Pen­guin­istas like to brag about how GNU/Lin­ux runs just fine on low-rent hard­ware, by con­trast with com­peti­tors like Vista that need the lat­est gleam­ing iron to be use­ful. And they have a point; but on­ly up to a point. I can tes­ti­fy from per­son­al ex­pe­ri­ence that an el­der­ly 333-MHz Dell with a re­cent De­bian to­tal­ly sucks wind when you run WordPress. And the re­al point is, it ain’t op­er­at­ing sys­tems that bog your com­put­er down, it’s app­s.
VirtualBox and Solaris · Wow, we must have some pret­ty sharp M&A peo­ple. I’m no vir­tu­al­iza­tion gu­ru, but I thought that at least I knew about the se­ri­ous play­er­s. But now we’re buy­ing in­notek, who make Vir­tu­alBox; I’d nev­er heard of ’em but the re­views look good. I tried it out and learned a cou­ple things ...
The T2 Servers · Th­ese T5x20 servers we’re an­nounc­ing to­day are a big deal. My bet is that they end up mak­ing Sun a lot of mon­ey; but on the way, they’re go­ing to bring the whole serv­er busi­ness (not just Sun’s piece of it) face to face with some re­al dis­rup­tion ...
Testing the T5120 · This was go­ing to be a Wide Fin­der Pro­ject progress re­port, but I end­ed up writ­ing so much about the serv­er that I’d bet­ter ded­i­cate an­oth­er frag­ment to the com­par­isons of all those im­ple­men­ta­tion­s; es­pe­cial­ly since there are still lots more im­ple­men­ta­tions to test. So this a hands-on re­port on a cou­ple of more-or-less pro­duc­tion T5120’s, the T2-based serv­er that’s be­ing an­nounced to­day. Head­li­nes: The chip is im­pres­sive but weird; as­tound­ing message-passing bench­mark num­ber­s; fight­ing the US DoD ...
[1 comment]  
Important: Constellation · [Missed this one last week in Europe] We made a big splash ear­li­er this sum­mer at the High-Performance Com­put­ing show in Dres­den; a con­cert­ed at­tempt to grab a big­ger piece of the HPC mar­ket. I thought the cov­er­age was pret­ty in­tel­li­gen­t: the big deal about Con­stel­la­tion isn’t the blades or the disks, it’s the big switch. Any­one can stuff a bunch of racks with blades with hot chips on ’em; the hard part is get­ting them to work to­geth­er, and what­ev­er the ap­proach, it’s easy to get bot­tle­necked on the mes­sag­ing. The big new 3,456-way In­fini­band switch is a brute-force as­sault on the prob­lem, com­bin­ing sil­i­con, bus-ware, con­nec­tor, and ca­bling wiz­ardry: Josh Si­mons has close-ups and un­veil­ing shots, but Jonathan has the best pic­tures. Why is this in­ter­est­ing? Josh has the num­bers: 19% of the world­wide serv­er mar­ket.
Blade Marketing · We an­nounced a bunch of blade stuff this morn­ing and I hon­est­ly couldn’t think of any­thing use­ful to write about it, since it’s a decade or two since I was a sysad­min and dis­cus­sion of I/O den­si­ty and the like tends to go over my head. Ap­par­ent­ly you can have Xeon and Opteron and SPARC and Win­dows and So­laris and Lin­ux all in the same chas­sis, which sounds kind of ter­ri­fy­ing to me. I did get to hear Andy Bech­tol­sheim give the spiel on the I/O de­sign on­ce, I un­der­stood maybe 20% but it sound­ed awe­some. Any­how, there is an in­ter­est­ing an­gle: it turns out we gave Scoble over at PodTech a pre-release ex­clu­sive for his video cov­er­age. Which is not ex­act­ly high-gloss mar­ket­ing, most­ly just John Fowler in a monochrome meet­ing room with some journos, pulling dull sil­ver blades and I/O mod­ules out of dull sil­ver chas­sis box­es and talk­ing about the tech, tak­ing a few ques­tion­s. I think it’s im­mense­ly more ef­fec­tive than most product-launch bumph. As for the prod­uct­s, I don’t re­al­ly have an in­formed opin­ion; but if you un­der­stand blades, you prob­a­bly will if you watch the video.
Big, Big Iron · Yes­ter­day April 17th we did a great big honk­ing an­nounce­ment about server­s, er make that an an­nounce­ment about great big honk­ing server­s. Frankly, I know pret­ty well zilch about this class of ma­chine; I can re­late to Web pumps like the T1000/T2000, and to an x86 meat-grinder like the X4100 (which an ir­ri­tat­ed lit­tle bird work­ing for one of our cus­tomers just told me that you can’t buy at the mo­ment be­cause they’re sold out­); but these supertanker-scale things we cooked up with Fu­jit­su are from an­oth­er plan­et. I went pok­ing around blogs.­sun.­com look­ing for some­thing hands-on. Bin­go (but, er, anony­mous bin­go, which ir­ri­tates me; I’ve ex­pressed my opin­ion on this in­ter­nal­ly). First, “EXOTERIC” was down in the trench­es with Fu­jit­su and the iron, see New Sun-Fujitsu SPARC En­ter­prise Sys­tems. Se­cond, “BM Seer” fires a vol­ley of bench­mark­s: bang bang bang bang. I may not grok the fin­er points of main­frame tech­nol­o­gy but you can fol­low point­ers from the an­nounce­ment to the M9000 and look at the heavy-metal pic­tures. Ouf.
Rock Secret Contest · Check out Jonathan’s piece to­day on the Rock. Some­times he makes me ner­vous... but hey, let's have some fun with it. The first three peo­ple who can de­ci­pher his art­ful clues and guess (by com­ment­ing here) what the Big Cool Nifty Rock Se­cret is will win a round of blog ap­plause from me and al­so a neat Sun sweat­shirt or hood­ie (my own, un­opened con­fer­ence loot sit­ting at the back of a clos­et, this is not a Sun con­test etc etc) ...
Intel Inside · For some cheap chuck­les, check the Web sites and con­sid­er the sub­tle dif­fer­ences be­tween Intel’s spin and ours. I had no idea this was go­ing to hap­pen this week, but a few months ago I was in a room where Andy Bech­tol­sheim al­lowed that, yes, In­tel was play­ing some damn de­cent catch-up, so the writ­ing was pret­ty well on the wal­l. Andy’s ex­pla­na­tion of the fin­er points of the trade-offs re­quires the best part part of two white­boards to lay out and an alert well-caffeinated mind to fol­low, so I won’t even try. But... ain’t com­pe­ti­tion won­der­ful? For the fore­see­able fu­ture, I bet al­most all sys­tems ven­dors will ship both In­tel & AMD sil­i­con. Hearty, hon­est con­grats to In­tel for get­ting back in the game. Both sides of the ar­range­ment look to me like a win for both sides; as Otelli­ni says, So­laris is a big deal in some mar­kets that In­tel wants more of, and if they say they can make it run bet­ter on the chips they build, I’d be in­clined to be­lieve them. Should be fun times in the serv­er biz, hang on tight.
[1 comment]  
Thumper & Friends · We an­nounced a bunch of new box­es this morn­ing (of, course, the damn Regis­ter has had the poop for week­s, I find our leak­i­ness ir­ri­tat­ing). There’s a Real Big Opteron serv­er (per­son­al­ly, I’m more of a scale-out than scale-up kin­da guy, but big iron is a big part of our busi­ness). There’s a blade box. I know noth­ing about blades, nev­er been near one. Then there’s the Thumper oops X4500, it’s in­ter­est­ing. I even have a grainy am­a­teur­ish pho­to of the in­side of a pre-production mod­el ...
Engineering Beauty · This last week, I spent a bunch of time in a room with Andy Bech­tol­sheim. He brought along some of the new box­es he’s work­ing on; some al­most ready, some raw sheet-metal pro­to­type­s. I got some cool pix of a Thumper’s guts but they won’t let me run ’em. Any­how, when you lis­ten to Andy talk about these box­es, you re­al­ize that in their own way they’re beau­ti­ful. You have to watch it around him; if you ask a sim­ple ques­tion like why one server’s I/O is sur­pris­ing­ly faster than another’s, you’re apt to get a 45-minute dis­ser­ta­tion in­volv­ing ten or fif­teen pages’ worth of block di­a­grams (no, I’m not ex­ag­ger­at­ing). Any­how, while they say the X64 box­es are our fastest-growing prod­uct line, I think that group’s feel­ing a lit­tle starved for love in among all the T2000 hoop-la. The mar­ket­ing group sends me notes from time to time point­ing to their news (for ex­am­ple a pow­er cal­cu­la­tor and a vir­tu­al­iza­tion ma­ni­ac), won­der­ing if it’s blog­gable. I keep telling them to grow their own blog­ger­s, but they haven’t yet. Which is a pity, be­cause what Andy’s do­ing makes a pret­ty com­pelling sto­ry.
On Computers · My think­ing about com­put­ers changed this week. I’ve been watch­ing the gen­er­al trends in Web traf­fic and da­ta sizes and so on, and been kind of wor­ried that the servers aren’t go­ing to be up to the loads that we’ll be throw­ing at them. Wel­l... now I’m not wor­ried about that any more. But we have some se­ri­ous soft­ware pain com­ing at us.
Servers In the Right Places · Ear­li­er this month I lament­ed that we didn’t have much of a pro­cess for do­nat­ing com­put­ers to projects that are do­ing good things. We seem to be mak­ing some progress on that, for ex­am­ple the T2000-tryout pro­gram seems to be run­ning a lot smoother. But that’s not al­l; for ex­am­ple, an X2100 showed up Fri­day on the doorstep of Nex­en­ta, as in GNU/So­laris. I think that this kind of thing is a com­plete no-brainer and hope that we man­age to do more of it.
Oracle vs. Niagara · Last week I was in Ed­mon­ton, and spent some time talk­ing to the lo­cal Sun of­fice and some cus­tomer­s. One of the things we talked about was our “through­put com­put­ing” prod­uct line, which is com­ing, uh I be­lieve the par­ty line is “late this year or ear­ly next year”. I like to talk about this stuff be­cause in the Web-centric world where I live, a highly-parallel low-wattage ma­chine hits a bunch of sweet spots at on­ce. (Al­so, it presents in­ter­est­ing soft­ware prob­lems.) Out there in the field, they seem to like it too; then on two sep­a­rate oc­ca­sions I heard “But we wouldn’t be able to use that.” I asked why, and they ex­plained that Oracle’s id­i­ot­ic per-core pric­ing for­mu­la would make it pro­hibitive­ly ex­pen­sive. Hey Or­a­cle, Sun isn’t the on­ly com­pa­ny that’s go­ing to be ship­ping highly-parallel com­put­er­s, and if there’s a tech­nol­o­gy out there that meets a lot of cus­tomer need­s, and you’re stand­ing in the way of them get­ting it, all you’re do­ing is mov­ing the FYO point clos­er and closer.
author · Dad · software · colophon · rights
Random image, linked to its containing fragment

By .

I am an employee
of Amazon.com, but
the opinions expressed here
are my own, and no other party
necessarily agrees with them.

A full disclosure of my
professional interests is
on the author page.