The best commercially available oscillators I've seen have frequency tolerance of 20 ppm. I suspect that if you can hold the temperature stabile to within 1 degree Celsius, these won't drift around too much, so you can have them selected to your 1 ppm requirement if you can control the temperature to 20 degrees Celsius +/- 1 degree. I suspect that you need either to have numerous oscillators at precisely the same frequency, or you need to match a standard frequency within that tolerance. Which you really need will greatly affect your cost.

You're probably asking too much of an oscillator if you want it to perform that well over the entire industrial temperature range.

Your problem, now, is a systems analysis problem, and not a hardware design problem. Why do you require so tight a control over frequency? Why must it be 1 ppm and not 2, or even 10? Will 1 ppm be enough? Why must the oscillator tolerate so wide a temperature range? Can a TCXO work?

Vectron, among others,
(see
http://www.vectron.com/products/..._index.htm
for examples)
produce temperature controlled oscillators that allow you to tune them over a 10 ppm range, for example, and you might be able to get "close enough" once their temperature is stabilized to ensure they stay within your requirement over the entire operating range.

They also offer others that are "close" to your asserted requirement (1.5 ppm). http://www.vectron.com/products/tcxo/c2560.pdf
These can be tweaked to hit less than 1.0 ppm with some coaxing.

RE