RE: OT: Sparc not dead yet
- From: "Main, Kerry" <Kerry.Main@xxxxxx>
- Date: Tue, 25 Apr 2006 23:54:25 -0400
-----Original Message-----
From: AlexNOSPAMDaniels@xxxxxxxxxxxxx
[mailto:alexdaniels@xxxxxxxxxxxxx]
Sent: April 25, 2006 6:30 PM
To: Info-VAX@xxxxxxxxxxxx
Subject: Re: OT: Sparc not dead yet
Main, Kerry wrote:
Would you be willing to pay extra for this hot swapcapability vs using
VMS servers on std HW (cheaper)
If your saying the cost of the hot-swap 'PCI' card capability in
hardware, would make hardware more expensive, then maybe you need to
review the facts, that capability is already in VMS supported
hardware.
Marvel's specifically.
From the HP website..
Bottom line is hot swap adds a lot of complexity into overall designs
and testing. Read much higher costs. Imho, if one implements dual
options (NIC's / SAN cards etc), then you can implement a solution with
much more standard hardware. Read much lower costs.
It's also been confirmed to me from other people in HP, or if you like
just open up an IO Drawer on a Marvel and look for the thumb-screw
plates that hold in the 'PCI' cards. The ones that have purple/red
thumb-screw plates, were designed to be hot-swappable in hardware.
And with PCI-E coming along, I suspect a lot of future Integrity boxes
will also support it.
And do you think these features will impact the pricing you will pay for
this capability?
with teamingIf you have this capability, local HW hot-swapping becomesmuch less of
an issue. Especially if FC/NIC adapters are implemented
design. If one adapter fails, simply schedule a time forthe system to
be taken down and have it replaced. Since you do not needto tell end
users about this server going, they will not care.
As per my last mail, thats what I already do. With the increased costs
of the extra cards, to impliment FailSafe on top of LAN
failover (so as
I don't get a performance hit). What I would like as an option however
would be to hot-swap the cards and then just have FailSafe on some
boxes (in the case of NICs).
Well, increased NIC cards are a few hundred $'s, so in most solutions,
this is an insignificant cost.
With that project I worked on, using TCPware, we were able tono loss of
telnet to a server, start a monitor system type command to generate
activity, pull the associated NIC cable and the session continued
running after only about a 2-3 second pause. No errors and
data. With TCPware, we could see that the NIC failed overproperly. This
also supported fail-back as well i.e. when we re-connectedthe NIC cable
after a 2-3 second pause, the connection automaticallyfailed back to
the original port.
As per TCP/IP Services FailSAFE IP, and VMS' LAN Failover.
performance
As I'm sure your aware FailSafe IP also gives you added
subnet. Losingbenfits for your outgoing traffic, if there in the same
workaround is toa card drops your performance. And yes the obvious
if you had adeploy FailSafe on top of LAN Failover, with the added cost.
As you indicated, the loss is in transmit only i.e. perhaps
big FTP background load, there might be some loss, butremember that the
overall load is also split across the other servers in thecluster as
well. I doubt end users would notice the impact of one NICfailing on
one server in the cluster when its mate takes over.
Users don't notice, nor do batch jobs, as I tend to impliment FailSAFE
on top of LAN Failover. If I just had FailSAFE, then I would get the
performance hit. And sure the users wouldn't notice, but large data
transfers being generated by batch jobs in progress would notice.
have not beenFC cards same issue, you lose performance, possible path
changes across
the cluster et al.
Heck, the 2Gb cards I have used for the last 2 years or so
pushed at all - even with high IO loads.
If that were true in all cases, it makes you wonder why VMS
Engineering
are qualifing 4GB HBAs with Integrity boxes?
Part of the overall storage industry .. Have to keep up with storage
vendor options.
And indeed, support in VMS is already latent in F8.3.
I know - I am one of the OpenVMS Ambassadors.
The "QLogic ISP2432", a dual-port 4GB HBA and PCI-Express too, a shame
we can't use the hot-swap features of PCI-Express.
Yes, there might be some path
changes, but that is transparent to applications and end users.
Sure it's transparent in terms of availability. In terms of
performance
it's not in a lot of environments.
If you've balanced for performance reasons a small number
disks on to a
specific HBA, that fails and gets the paths get pushed off to another
HBA, and they stay there (or can only be moved to other busy HBAs)
until you can drop the box, your application can certainly notice.
If your design includes a single HBA for any storage subset, then you
could have issues if that single HBA fails.
Also, while there may be app's out there that can overload a 2GB HBA
before some other bottleneck is hit first, I certainly have not
encountered any - and that includes some very heavy IO intensive SAN
clusters. I usually recommend staying away from dedicating HBA's in this
way.
Of course, the answer is to simply to buy more HBAs, being that you
know you can't hot-swap/add in new ones.
With any single HBA solution, your design and end users get hosed when a
failure occurs. If the failure occurs at 2:00AM, then by the time the
failure is analyzed, parts ordered and replaced (even with hot swap),
your solution has failed those users impacted during this time. Multiple
HBA's / NIC's etc are the best solution imho.
then drop aSo yes one can wait for the users/batch jobs to finish,
puchase morenode, but it's not ideal and customers are left having to
base design.hardware to mitigate for these times.
That's what mission critical Customers do as part of their
In most Cust environments I have seen, normal users tend tologout at
5-8:00pm or they get disconnected for security reasons i.e.forgot to
logout before heading home.
I'm talking about trading systems supporting users and markets all
around the world. 9-5 local time isn't what it's about.
Time is not important in this solution. As I stated earlier, this
solution provides capability to repair servers in prime time with zero
application availability impact. The dual LAN/SAN adapters carry th load
until the connections have moved to another server.
I'm more than familar with running active-active clusters and being
able to drop a node, my issue is the hit while one is waiting to drop
it. More hardware (which costs) helps here and can mostly solve the
problem, but hot-swapping would be a nice option to have too.
I am not saying hot swap PCI is not a good feature - just that imho,
dual PCI options on active-active clustered servers using using standard
HW will likely be a cheaper overall solution than having to move to
higher end servers that have the hot swap PCI capabilities.
Kerry Main
Senior Consultant
HP Services Canada
Voice: 613-592-4660
Fax: 613-591-4477
kerryDOTmainAThpDOTcom
(remove the DOT's and AT)
OpenVMS - the secure, multi-site OS that just works.
.
- Prev by Date: Re: hypothetical question
- Next by Date: EFI and Satellite booting
- Previous by thread: Re: OT: Sparc not dead yet
- Next by thread: RE: OT: Sparc not dead yet
- Index(es):
Relevant Pages
|
