import sys

sys.setrecursionlimit(10**6)
int1 = lambda x: int(x)-1
p2D = lambda x: print(*x, sep="\n")
def II(): return int(sys.stdin.buffer.readline())
def MI(): return map(int, sys.stdin.buffer.readline().split())
def LI(): return list(map(int, sys.stdin.buffer.readline().split()))
def LLI(rows_number): return [LI() for _ in range(rows_number)]
def MI1(): return map(int1, sys.stdin.buffer.readline().split())
def LI1(): return list(map(int1, sys.stdin.buffer.readline().split()))
def LLI1(rows_number): return [LI1() for _ in range(rows_number)]
def BI(): return sys.stdin.buffer.readline().rstrip()
def SI(): return sys.stdin.buffer.readline().rstrip().decode()
def inval(ni, nj, h, w):
    if ni < 0 or ni >= h or nj < 0 or nj >= w: return True
    return False
dij = [(0, 1), (-1, 0), (0, -1), (1, 0)]
inf = 10**16
# md = 10 ** 9 + 7
md = 998244353

class BitSum:
    def __init__(self, n):
        self.n = n+1
        self.table = [0]*self.n

    def add(self, i, x):
        i += 1
        while i < self.n:
            self.table[i] += x
            i += i & -i

    # [0,i]の和
    def sum(self, i):
        i += 1
        res = 0
        while i > 0:
            res += self.table[i]
            i -= i & -i
        return res

n=II()
aa=LI()
enc= {a:i for i,a in enumerate(sorted(set(aa)))}
bb=[enc[a] for a in aa]
ans=0
cc=[0]*n
low=[0]*n
up=[0]*n


bit1=BitSum(n+3)
bit2=BitSum(n+3)
for i,b in enumerate(bb):
    b=n-b
    c=bit2.sum(b-1)
    cc[i]+=c
    ans%=md
    c=bit1.sum(b-1)
    up[i]=c
    bit2.add(b,c)
    bit1.add(b,1)
# print(cc)

bit1=BitSum(n)
bit2=BitSum(n)
for i,b in enumerate(bb[::-1]):
    c=bit2.sum(b-1)
    cc[n-1-i]+=c
    ans%=md
    c=bit1.sum(b-1)
    low[n-1-i]=c
    bit2.add(b,c)
    bit1.add(b,1)
# print(cc)
# print(up)
# print(low)

for i,(u,l) in enumerate(zip(up,low)):
    cc[i]+=u*l

for a,c in zip(aa,cc):
    ans+=a*c
    ans%=md

print(ans)