RE: Sizing bandwidth for Disaster Tolerant cluster inter site links

From: Main, Kerry (kerry.main_at_hp.com)
Date: 02/27/04 

Date: Fri, 27 Feb 2004 08:57:35 -0500

> -----Original Message-----
> From: Andrew Rycroft [mailto:andrew.rycroft@intrinsitech.com]
> Sent: February 27, 2004 7:37 AM
> To: Info-VAX@Mvb.Saic.Com
> Subject: Sizing bandwidth for Disaster Tolerant cluster inter
> site links
>
> Hi,
>
> I would really appreciate if anybody can advise me here.
>
> We plan to implement a disaster tolerant cluster with
> OpenVMS. I have now been asked to size the require bandwidth
> to meet the application needs. The 2 sites are about 10km
> apart, and we can put in a managed fibre link, but it is very
> expensive. We have 2 disks that we intend to volume shadow
> across the 2 sites. The average IO rate on the one disk is
> 100 IO/sec, and on the other 50 IO/Sec.
>
> I have seen various presentations where OPenVMS can be
> directed to read from the local disks, but I do not know how
> I can differentiate between reads and writes in the IO rate
> on a per disk level.
>
> Any advice on how to size the correct intersite bandwidth
> required that will not degrade performance would be appreciated ?
>
> Thanks
> Andrew
>

Andrew,

Re: multi-site info..

Wrt to bandwidth, you are correct that with an OpenVMS cluster, one
typically configures shadow sets to read from the local members of the
shadow set. Only writes (updates) are propagated to the remote site,
hence the read-write ratios are a factor to consider in the link speed
requirements.

A few pointers for consideration:

http://h71000.www7.hp.com/openvms/fibre/fc_hbvs_dtc_wp.pdf (multi-site
OpenVMS WP)
http://www2.openvms.org/kparris/ (Keith Parris's presentations)
http://h71000.www7.hp.com/doc/732FINAL/DOCUMENTATION/PDF/aa-pvxmj-te.PDF
(V7.3-2 shadowing doc)

Regards

Kerry Main
Senior Consultant
HP Services Canada
Voice: 613-592-4660
Fax: 613-591-4477
Email: kerryDOTmainAThpDOTcom
(remove the DOT's and AT for email address)

Relevant Pages

  • Re: What Version Of FreeBSD?
    ... but downloading additional 300KB of binary libraries will be more ... than downloading 3MB of these library sources ... So instead of using small disks, which are getting rarer and rarer these ... including those with large disks to burn more bandwidth. ...
    (comp.unix.bsd.freebsd.misc)
  • Re: Use a lot of disks as a large buffer?
    ... Since the total bandwidth of each disk is about 40 ... >> disks bandwidth. ... >> If the requests are random and assuming disk usage is well spread over ... > disadvantage, of course, is the reduced capacity. ...
    (comp.unix.programmer)
  • Re: Re: [9fans] Any one going to change to sata disks?
    ... i had two disks i wanted to use, ... plugged in by itself it will appear as sdC0, if i plugged in a second ... was plugged in to the IDE slots, including just a cdrom, the sata ... would be 5 times slower than the bandwidth to the other. ...
    (comp.os.plan9)
  • Re: Striping data across platters in a single hard disk
    ... Sure it will -- you potentially can get double the bandwidth ... > new SATA disks will do seek sorting for IO optimization. ... Decent computers have been running elevator algorithms for ... PhotoShop may just have ...
    (comp.sys.ibm.pc.hardware.chips)