#pragma GCC optimize("O3,unroll-loops")
#pragma GCC target("avx2,bmi,bmi2,lzcnt,popcnt")
#include <bits/stdc++.h>
using namespace std;
using ll = long long;
template<class T> bool cmin(T &i, T j) { return i > j ? i=j,true:false; }
template<class T> bool cmax(T &i, T j) { return i < j ? i=j,true:false; }
mt19937 mrand(chrono::steady_clock::now().time_since_epoch().count());
uniform_int_distribution<int> ui(0, 1 << 30);
int rng() {
return ui(mrand);
}
const int N=100100;
int n;
int h[N],k[N];
int placed[N];
int order[N];
int st[4*N];
void upd(int p,int x,int u=1,int tl=0,int tr=N-1) {
if (tl == tr) {
st[u]=(x > 0 ? tl : N);
return;
}
int mid=(tl+tr)/2;
if (p <= mid)
upd(p,x,2*u,tl,mid);
else
upd(p,x,2*u+1,mid+1,tr);
st[u]=min(st[2*u],st[2*u+1]);
}
int query(int l,int r,int u=1,int tl=0,int tr=N-1) {
if (l > r) return N;
if (l == tl && r == tr)
return st[u];
int mid=(tl+tr)/2;
return min(query(l,min(mid,r),2*u,tl,mid),query(max(mid+1,l),r,2*u+1,mid+1,tr));
}
void build(int u=1,int tl=0,int tr=N-1) {
st[u]=N;
if (tl==tr) return;
int mid=(tl+tr)/2;
build(2*u,tl,mid);
build(2*u+1,mid+1,tr);
}
int main() {
ios_base::sync_with_stdio(false);
cin.tie(nullptr);
cin >> n;
for (int i=0;i<n;i++)
cin >> h[i] >> k[i];
iota(order,order+n,0);
sort(order,order+n,[&](int i,int j) {
return h[i] < h[j];
});
int lh = 0;
build();
for (int t=0;t<n;t++) {
int i = order[t];
int diff = h[i] - lh;
lh = h[i];
placed[0] += diff;
upd(0,placed[0]);
vector<pair<int,int>> work;
int l=-1;
while(k[i]>0) {
int j=query(l+1,N-1);
assert(j<N);
int now = min(k[i],placed[j]);
placed[j]-=now;
placed[j+1]+=now;
work.emplace_back(j+1,placed[j+1]);
work.emplace_back(j,placed[j]);
k[i] -= now;
l = j;
}
for (auto &[p,np]: work)
upd(p,np);
/*int j = 0;
new_placed[0] = placed[0];
for (;k[i]>0;j++) {
int now = min(k[i],placed[j]);
new_placed[j + 1] = placed[j + 1] + now;
k[i] -= now;
new_placed[j] -= now;
}
for (int z=0;z<=j;z++)
placed[z] = new_placed[z];
*/
}
ll ans = 0;
for (int i=0;i<N;i++) {
ans += 1LL * i * (i - 1) / 2 * placed[i];
}
cout << ans << endl;
}
