Re: RAC Geographical Architecture
- From: "K Gopalakrishnan" <kaygopal@xxxxxxxxx>
- To: "Michael Fontana" <MFontana@xxxxxxxxx>
- Date: Mon, 16 Apr 2007 16:07:17 -0700
Mike, It does not imply 100km limit, the reason is differnt and we call as 'speed of light factor'. Quoting from my 'Oracle 10g RAC Handbook' Chapter 15, ***BEGIN*** In vacuum the light travels at 186, 282 miles per second. To make the math simpler, we round that off to 200,000 miles per second or 200 miles per millisecond. We require confirmation of any message, so we must use round-trip distances. Therefore, light can travel up to 100 miles away and back in 1 millisecond. But the actual transit time is longer, in part because the speed of light is slower by about 30% in an optical fiber and straight lines rarely occur in global communications situations, but also because delays are created when the signal passes through an electronic switch or signal regenerator. While considering all the factors like the inevitable switch delays, the conservative rule of thumb is that it takes 1 millisecond for every 50-mile round trip. Therefore, 500 miles adds 10 milliseconds to the latency of a disk access. Given normal disk access latency of 10-15 milliseconds, this merely doubles the latency, however the shadowing software layer can cope with that. But if the latency is more than that, the software might think the disk at the other end has gone off-line and will incorrectly break the shadow set. So the speed of light becomes the limiting factor because a latency of 1 millisecond (ms) is introduced for every 50miles of distance between the data centers, even with the use of dark fiber. ***END*** On 4/16/07, Michael Fontana <MFontana@xxxxxxxxx> wrote:
Someone else wrote back with a link to a good whitepaper from Oracle published in 2006. It seemed to imply a 100km limit, between the servers, in geographical terms. In addition to the added cost for the DWDM (hundreds of thousands$$$$), a site would also need at least four mirrored (remote) copies of disk, with clustering technology (which ain't cheap). You seem to imply this has been do-able since 2004? And that there are least 10 such sites running it? Amazing.
The number is quite higher than 10. To my knowledge it is around 50+ and 10 of them are direclty known to be . -Gopal
I would imagine 11g may have even more capability. While I don't normally like to discuss or even consider vaporware, this time my question is based upon putting together a 3-5 year plan for my executives. Any information on future Oracle direction would be helpful. Do you have a link from Oracle?
I am neither qualified nor allowed to talk more on this !!! But extended RAC is increasingly common now a days..
-----Original Message----- From: K Gopalakrishnan [mailto:kaygopal@xxxxxxxxx] Sent: Monday, April 16, 2007 4:27 PM To: Michael Fontana Cc: oracle-l@xxxxxxxxxxxxx Subject: Re: RAC Geographical Architecture Mike, Your instructor is right. There is no limitation from RAC side. There are various technologies support this and the most common is DWDM (dark fiber) and there are lot of customers (I have personally participated/known more than 10+ of such) successfully deployed this architecture and running without any problems. In case if you have my RAC book, I have a dedicated chapter for "Extended RAC". Have a look in near by bookstore -Gopal BTW 11g RAC/ASM has some additional features for extended RAC (like preferred read for asm failure groups), however it would be too premature to discuss that now !
-- Best Regards, K Gopalakrishnan Co-Author: Oracle Wait Interface, Oracle Press 2004 http://www.amazon.com/exec/obidos/tg/detail/-/007222729X/ Author: Oracle Database 10g RAC Handbook, Oracle Press 2006 http://www.amazon.com/gp/product/007146509X/ -- http://www.freelists.org/webpage/oracle-l