Re: SUN's X64 servers

Well, allegedly, in article <g92dnXIRvMerqAbVnZ2dnUVZ_jednZ2d@xxxxxxxxxxx>, Richard B. Gilbert <rgilbert88@xxxxxxxxxxx> wrote:
Sun System User wrote:
We are looking at buying X4240's series server (only one) with 16
146gb drives and will be running latest Solaris 10 release. We do not
currently have any backup hardware or software we can use to backup
this server.

How reliable X64's servers are compare to Sparc's ?

Electronics are VERY reliable! Where you can expect problems are things
like fans, disk drives, tape drives, tape libraries, and anything else
that has moving parts.

Console on anything but the V20z servers are a real ***, though.
(And the Java console has been unusable in our locked-down network, so
we require serial or ssh->serial consoles that's reliable.)

SPARC has the far better ALOM / OBP interfaces. Still, x86 gear isn't
too bad -- they're nice on the performance angle. I really hate
consoles; the v20z servers were the only ones where console worked
reliably and out of the box.

Sun did something very bad; with V1 and V2 of X2100 firmware, console
defaulted to 115200bps, but then they switched to 9600bps with the V3
firmware. You really can't switch this kind of low-level/critical stuff
mid-stream. We had to deal with quite a few issues related to this change.
Not just with jumpstarting, but with servers in full production all over
the map. Great fun.

I've got boatloads of the X2100/2200/4100/etc. gear and we just couldn't
get console to work after upgrading to the V3 firwmare (required for
security fixes). Sun's currently baffled, too -- inspiring. So we had to
downrev firmware to V2 (e.g. 2.91) and hear about it from the security guys.

Each of our X2100/X2200 server at 2.70 or 2.91 has ~25 identified vulns
_PER SERVER_. JUST for the ELOM interface! Secure servers but unusable
consoles or insecure servers but usable consoles? Great fun dealing with
both Sun and our security guys.

That aside, seems OK as far as reliability goes. I don't like the SATA
disks, though; they just simply don't hold up as long as SCSI does. Both
are rated for very different duty cycles, and it shows up pretty clearly
in operational practice.

SATA's one selling point is: it's cheap. But you get what you pay for.
:-) SAS is OK but these disks keeps occasionally failing in less than a
year from the factory -- we're rather disappointed with operational
reliability. Still, gear's under support, so we don't stress too much
about it, and it's not been horribly bad or anything. Just some raised
eyebrows over the premature failures, but still a low-enough percentage
we're not too concerned... yet.

With SPARC, if disk0 dies, it can auto-boot off disk1 or you just pop
something in and type 'boot disk0' (or 'boot disk1'). But on x86, we've
found that it's less than trivial to deal with mirrored disk
replacement, especially from a booting perspective.

The x86 BIOS comes with a fair amount of legacy crud that makes this
stuff 'interesting'. Too bad Sun hasn't adopted EFI yet for the X*
series, would probably make this kind of heartache a non-issue.

I still strongly prefer the SPARC stuff -- they were much less time
consuming and far easier to get going AND to maintain. Still, the price
vs performance for the x86 gear in various situations makes it rather
attractive.

-Another Sun System User
.