Different step responses depends on poles. Generally due to zero at orign, system gain will be zero. For step or impulse inputs, output will decay exponential; this is because the DC gain for this system is zero, so the total area of the impulse response must be zero. Final value of the step response is always zero; it responds to a change in the input, but sags back to zero if the input is steady.

@ Utkal thanks

Is there any way to calculate response?

The step response can be calculated applying the inverse Laplace transform to the expression [H(s)/s]. H(s) is the systems transfer function.

The question is well thought but has a lot  of directions.

1) if G(s) =  y(s)/R(s) = s then y(t) =  exp(-t) and the steady state error to a reference step will be inifinity.

2) if there are poles in the RHP then no question about stability.

3) if there are poles in the LHP then the step response will be determined by the location of poles. e.g. if one of the poles is very near to the zero and the other poles are far from the origin then the zero may not dominate that much.

Thanks Imran

If the poles are complex and close to zero ?

Not defined neither in practice, nor in therory.

if poles are complex and near to a zero at the origin then imagine the root locus of the system. Both the poles break at the real axis and system's step response eventually becomes zero. (In feedback it can't at all track a step)

Further more emmanuel Delaleau is right no such system exists.

However for visualization you may execute the following MATLAB code:

n =[1 0];

d  = [1 2 2];

sys = tf(n,d)

figure(1);

rlocus(sys)

system = feedback(sys,1)

figure(2);

step(system)

The system will eventually come back to the initial steady-state irrespective of the magnitude of step perturbation. for example, for a system with two poles. Physically, the velocity of a damped vibrator (or analogously current in RLC circuit) exhibits such a behaviour. If we start from rest, the velocity finally will again become zero, irrespective of the magnitude of the step (or even impulse) perturbation.  

An example of such a system in practice is a wash-out filter. To evaluate the step response, as indicated by Utkal, just cancel the zero at s=0 with the input transfer function 1/s and calculate the impulse response of the remainder of the transfer function.