What's the ideal hardware for replacing L-2000 and Autoraids?

[Lothar Krueler]

Hello,
we own 2 L-2000 with 2 440 Mhz CPUs and 2 GB Ram and 2 striped Autoraids 12h (each 9*18 GB and 3*36 GB, 2*96MB Controller) in a SG-Cluster.
The most important bottleneck is low performance of the Autoraids, space actually used is nearly 150 GB.
Server performance would become enough by upgrading the servers to 4 CPUs and 4GB.
But the current support-costs let us think about building a high available environment by leasing a new server with a new storage system and using the old servers(out of support) as standby emergency server in an other place.
In this case we would need a new server and a new storage in one place and a new storage in the other place.
The distance between the two places is less than 500 meters and St50/125 fibre-lines are available.
Our Apps would need HP-UX the next three years, the storage should be used longer, perhaps by other systems too.

Can you please give me some advice of what hardware would be the right in my case and what's the best way of mirroring the storages ?

Thanks in advance
Lothar Krueler

[Zinky]

2 x rp3440-4
and
MSA500/1000 Storage


OR

2 x 2-4Way Opteron Servers
and
MSA500/1000 or EVA3000
(you will have to switch to Linux then.

[Lothar Krueler]

Hi Nelson,
thanks for your fast response!

I wonder why you suggest two servers. Sure, for using the L-2000 we have to build in FC-interfaces, but isn't that cheaper than a second new server?
I think of replacing the cluster of 2 L-2000 by one new server in the first location and then use one of the L-2000 with 4 CPU and 4 GB in the other location standby for emergency.

Does the MSAxxx storage support mirroring to an other MSAxxx itself ? Perhaps by copying a snapshot ?

Regards
Lothar

[Zinky · Accepted solution]

I thought you needed total replacement of your L-2000's.

Yes.. the MSA/EVA arrays have in-array replication software -- called Continous Access or CA. If you've a cluster .. I think you can already use VxVM for replication as well..

[Lothar Krueler]

Thank you Nelson,

because i don't want to set the bunny, i gave only 9 pts. Perhaps there will come further opinions.

Regards
Lothar

[Slawomir Gora]

Hi,

MSA is not very good solution for UNIX systems,
it can't use redundant controler when is connected to unix box and is not
supported with MC Service Guard cluster software.

If you want to build campus cluster
(distance limit 300m for 2Gb or 500m for 1Gb MultiMode FO 50)
you should use EVA3000.

As server you can use rp3440 or maby some
itanium box rp4640 or rx2600

[Bernhard Mueller]

Lothar,

you should use one rp3440 and two MSA1000 with the internal FC hub.
(which acutally *is* supported with MC/SG, I asked an HP guy yesterday)

note the distance, over 300m you may have to reduce speed to 1Gb/s as Slawomir said.

I understand that your considerations are focused on $$$$ (or ⠬⠬⠬⠬) So if you need fast reliable failover you still need MC/SG. If you can afford a somewhat longer downtime with manual interaction and just need to protect the data, you could question the cluster solution and instead use ignite to clone the new server as a cold standby using the ohter L2000 as your spare part stock (but then either use 11.11 or consider the non-supported HW list on 11.23!)

Run the cloning on a regualar base with post config scripts for vg imports from the rp3440 (MSAs) and (maybe) all you need to do in case of disaster is edit /etc/rc.config.d/netconf and reboot.

This may not work if you have license manager stuff running bound to the SPU and stuff like that. But that is an issue in MC/SG as well.

Regar

[Ted Buis]

The MSA1000 has some limitations with respect to SG, as it only supports single controller connections presently. The VA7110 or eva3000 are much better but higher priced. You may want to consider two ds2405 with two FC cards in the rp3440 (use two FC/GigE combo cards) and MirrorDisk/UX as a low cost alternative for a two node cluster.
Note this system requires HP-UX 11.11.
Integrity servers are an option if you can
run on 11.23. At 500m your latency will rise somewhat.

[Lothar Krueler]

Hi,
thank you all for your replys, i think all of them will help me. It seems that the question SG or not SG is an important point i have to think about.

Is there anyone, who has experince with non HP storage systems, e.g. IBM or other ?

Greetings,
Lothar

[Ivajlo Yanakiev]

DMX is OK for HPUX 11.11 and 11.23
Do not use Clariion :)