diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index 7e54236bd..edcc46561 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -28,7 +28,27 @@ How to teach Moto to support a new AWS endpoint:
 * If one doesn't already exist, create a new issue describing what's missing. This is where we'll all talk about the new addition and help you get it done.
 * Create a [pull request](https://help.github.com/articles/using-pull-requests/) and mention the issue # in the PR description.
 * Try to add a failing test case. For example, if you're trying to implement `boto3.client('acm').import_certificate()` you'll want to add a new method called `def test_import_certificate` to `tests/test_acm/test_acm.py`.
-* If you can also implement the code that gets that test passing that's great. If not, just ask the community for a hand and somebody will assist you.
+* Implementing the feature itself can be done by creating a method called `import_certificate` in `moto/acm/responses.py`. It's considered good practice to deal with input/output formatting and validation in `responses.py`, and create a method `import_certificate` in `moto/acm/models.py` that handles the actual import logic.
+* If you can also implement the code that gets that test passing then great! If not, just ask the community for a hand and somebody will assist you.
+
+## Before pushing changes to GitHub
+
+1. Run `black moto/ tests/` over your code to ensure that it is properly formatted
+1. Run `make test` to ensure your tests are passing
+
+## Python versions
+
+moto currently supports both Python 2 and 3, so make sure your tests pass against both major versions of Python.
+
+## Missing services
+
+Implementing a new service from scratch is more work, but still quite straightforward. All the code that intercepts network requests to `*.amazonaws.com` is already handled for you in `moto/core` - all that's necessary for new services to be recognized is to create a new decorator and determine which URLs should be intercepted.
+
+See this PR for an example of what's involved in creating a new service: https://github.com/spulec/moto/pull/2409/files
+
+Note the `urls.py` that redirects all incoming URL requests to a generic `dispatch` method, which in turn will call the appropriate method in `responses.py`. 
+
+If you want more control over incoming requests or their bodies, it is possible to redirect specific requests to a custom method. See this PR for an example: https://github.com/spulec/moto/pull/2957/files
 
 ## Maintainers
 
diff --git a/IMPLEMENTATION_COVERAGE.md b/IMPLEMENTATION_COVERAGE.md
index 43983d912..1d5eb946a 100644
--- a/IMPLEMENTATION_COVERAGE.md
+++ b/IMPLEMENTATION_COVERAGE.md
@@ -459,18 +459,18 @@
 
 ## application-autoscaling
 <details>
-<summary>0% implemented</summary>
+<summary>20% implemented</summary>
 
 - [ ] delete_scaling_policy
 - [ ] delete_scheduled_action
 - [ ] deregister_scalable_target
-- [ ] describe_scalable_targets
+- [x] describe_scalable_targets
 - [ ] describe_scaling_activities
 - [ ] describe_scaling_policies
 - [ ] describe_scheduled_actions
 - [ ] put_scaling_policy
 - [ ] put_scheduled_action
-- [ ] register_scalable_target
+- [x] register_scalable_target - includes enhanced validation support for ECS targets
 </details>
 
 ## application-insights
diff --git a/README.md b/README.md
index 7a2862744..956be5da1 100644
--- a/README.md
+++ b/README.md
@@ -65,6 +65,8 @@ It gets even better! Moto isn't just for Python code and it isn't just for S3. L
 |-------------------------------------------------------------------------------------|                             |
 | API Gateway               | @mock_apigateway      | core endpoints done             |                             |
 |-------------------------------------------------------------------------------------|                             |
+| Application Autoscaling   | @mock_applicationautoscaling | basic endpoints done     |                             |
+|-------------------------------------------------------------------------------------|                             |
 | Autoscaling               | @mock_autoscaling     | core endpoints done             |                             |
 |-------------------------------------------------------------------------------------|                             |
 | Cloudformation            | @mock_cloudformation  | core endpoints done             |                             |
diff --git a/moto/__init__.py b/moto/__init__.py
index 4f8f08eda..b4375bfc6 100644
--- a/moto/__init__.py
+++ b/moto/__init__.py
@@ -15,6 +15,9 @@ mock_acm = lazy_load(".acm", "mock_acm")
 mock_apigateway = lazy_load(".apigateway", "mock_apigateway")
 mock_apigateway_deprecated = lazy_load(".apigateway", "mock_apigateway_deprecated")
 mock_athena = lazy_load(".athena", "mock_athena")
+mock_applicationautoscaling = lazy_load(
+    ".applicationautoscaling", "mock_applicationautoscaling"
+)
 mock_autoscaling = lazy_load(".autoscaling", "mock_autoscaling")
 mock_autoscaling_deprecated = lazy_load(".autoscaling", "mock_autoscaling_deprecated")
 mock_lambda = lazy_load(".awslambda", "mock_lambda")
diff --git a/moto/applicationautoscaling/__init__.py b/moto/applicationautoscaling/__init__.py
new file mode 100644
index 000000000..6e3db1ccf
--- /dev/null
+++ b/moto/applicationautoscaling/__init__.py
@@ -0,0 +1,6 @@
+from __future__ import unicode_literals
+from .models import applicationautoscaling_backends
+from ..core.models import base_decorator
+
+applicationautoscaling_backend = applicationautoscaling_backends["us-east-1"]
+mock_applicationautoscaling = base_decorator(applicationautoscaling_backends)
diff --git a/moto/applicationautoscaling/exceptions.py b/moto/applicationautoscaling/exceptions.py
new file mode 100644
index 000000000..2e2e0ef9f
--- /dev/null
+++ b/moto/applicationautoscaling/exceptions.py
@@ -0,0 +1,22 @@
+from __future__ import unicode_literals
+import json
+
+
+class AWSError(Exception):
+    """ Copied from acm/models.py; this class now exists in >5 locations,
+        maybe this should be centralised for use by any module?
+    """
+
+    TYPE = None
+    STATUS = 400
+
+    def __init__(self, message):
+        self.message = message
+
+    def response(self):
+        resp = {"__type": self.TYPE, "message": self.message}
+        return json.dumps(resp), dict(status=self.STATUS)
+
+
+class AWSValidationException(AWSError):
+    TYPE = "ValidationException"
diff --git a/moto/applicationautoscaling/models.py b/moto/applicationautoscaling/models.py
new file mode 100644
index 000000000..39bb497aa
--- /dev/null
+++ b/moto/applicationautoscaling/models.py
@@ -0,0 +1,179 @@
+from __future__ import unicode_literals
+from moto.core import BaseBackend, BaseModel
+from moto.ecs import ecs_backends
+from .exceptions import AWSValidationException
+from collections import OrderedDict
+from enum import Enum, unique
+import time
+
+
+@unique
+class ServiceNamespaceValueSet(Enum):
+    APPSTREAM = "appstream"
+    RDS = "rds"
+    LAMBDA = "lambda"
+    CASSANDRA = "cassandra"
+    DYNAMODB = "dynamodb"
+    CUSTOM_RESOURCE = "custom-resource"
+    ELASTICMAPREDUCE = "elasticmapreduce"
+    EC2 = "ec2"
+    COMPREHEND = "comprehend"
+    ECS = "ecs"
+    SAGEMAKER = "sagemaker"
+
+
+@unique
+class ScalableDimensionValueSet(Enum):
+    CASSANDRA_TABLE_READ_CAPACITY_UNITS = "cassandra:table:ReadCapacityUnits"
+    CASSANDRA_TABLE_WRITE_CAPACITY_UNITS = "cassandra:table:WriteCapacityUnits"
+    DYNAMODB_INDEX_READ_CAPACITY_UNITS = "dynamodb:index:ReadCapacityUnits"
+    DYNAMODB_INDEX_WRITE_CAPACITY_UNITS = "dynamodb:index:WriteCapacityUnits"
+    DYNAMODB_TABLE_READ_CAPACITY_UNITS = "dynamodb:table:ReadCapacityUnits"
+    DYNAMODB_TABLE_WRITE_CAPACITY_UNITS = "dynamodb:table:WriteCapacityUnits"
+    RDS_CLUSTER_READ_REPLICA_COUNT = "rds:cluster:ReadReplicaCount"
+    RDS_CLUSTER_CAPACITY = "rds:cluster:Capacity"
+    COMPREHEND_DOCUMENT_CLASSIFIER_ENDPOINT_DESIRED_INFERENCE_UNITS = (
+        "comprehend:document-classifier-endpoint:DesiredInferenceUnits"
+    )
+    ELASTICMAPREDUCE_INSTANCE_FLEET_ON_DEMAND_CAPACITY = (
+        "elasticmapreduce:instancefleet:OnDemandCapacity"
+    )
+    ELASTICMAPREDUCE_INSTANCE_FLEET_SPOT_CAPACITY = (
+        "elasticmapreduce:instancefleet:SpotCapacity"
+    )
+    ELASTICMAPREDUCE_INSTANCE_GROUP_INSTANCE_COUNT = (
+        "elasticmapreduce:instancegroup:InstanceCount"
+    )
+    LAMBDA_FUNCTION_PROVISIONED_CONCURRENCY = "lambda:function:ProvisionedConcurrency"
+    APPSTREAM_FLEET_DESIRED_CAPACITY = "appstream:fleet:DesiredCapacity"
+    CUSTOM_RESOURCE_RESOURCE_TYPE_PROPERTY = "custom-resource:ResourceType:Property"
+    SAGEMAKER_VARIANT_DESIRED_INSTANCE_COUNT = "sagemaker:variant:DesiredInstanceCount"
+    EC2_SPOT_FLEET_REQUEST_TARGET_CAPACITY = "ec2:spot-fleet-request:TargetCapacity"
+    ECS_SERVICE_DESIRED_COUNT = "ecs:service:DesiredCount"
+
+
+class ApplicationAutoscalingBackend(BaseBackend):
+    def __init__(self, region, ecs):
+        super(ApplicationAutoscalingBackend, self).__init__()
+        self.region = region
+        self.ecs_backend = ecs
+        self.targets = OrderedDict()
+
+    def reset(self):
+        region = self.region
+        ecs = self.ecs_backend
+        self.__dict__ = {}
+        self.__init__(region, ecs)
+
+    @property
+    def applicationautoscaling_backend(self):
+        return applicationautoscaling_backends[self.region]
+
+    def describe_scalable_targets(
+        self, namespace, r_ids=None, dimension=None,
+    ):
+        """ Describe scalable targets. """
+        if r_ids is None:
+            r_ids = []
+        targets = self._flatten_scalable_targets(namespace)
+        if dimension is not None:
+            targets = [t for t in targets if t.scalable_dimension == dimension]
+        if len(r_ids) > 0:
+            targets = [t for t in targets if t.resource_id in r_ids]
+        return targets
+
+    def _flatten_scalable_targets(self, namespace):
+        """ Flatten scalable targets for a given service namespace down to a list. """
+        targets = []
+        for dimension in self.targets.keys():
+            for resource_id in self.targets[dimension].keys():
+                targets.append(self.targets[dimension][resource_id])
+        targets = [t for t in targets if t.service_namespace == namespace]
+        return targets
+
+    def register_scalable_target(self, namespace, r_id, dimension, **kwargs):
+        """ Registers or updates a scalable target. """
+        _ = _target_params_are_valid(namespace, r_id, dimension)
+        if namespace == ServiceNamespaceValueSet.ECS.value:
+            _ = self._ecs_service_exists_for_target(r_id)
+        if self._scalable_target_exists(r_id, dimension):
+            target = self.targets[dimension][r_id]
+            target.update(kwargs)
+        else:
+            target = FakeScalableTarget(self, namespace, r_id, dimension, **kwargs)
+            self._add_scalable_target(target)
+        return target
+
+    def _scalable_target_exists(self, r_id, dimension):
+        return r_id in self.targets.get(dimension, [])
+
+    def _ecs_service_exists_for_target(self, r_id):
+        """ Raises a ValidationException if an ECS service does not exist
+            for the specified resource ID.
+        """
+        resource_type, cluster, service = r_id.split("/")
+        result = self.ecs_backend.describe_services(cluster, [service])
+        if len(result) != 1:
+            raise AWSValidationException("ECS service doesn't exist: {}".format(r_id))
+        return True
+
+    def _add_scalable_target(self, target):
+        if target.scalable_dimension not in self.targets:
+            self.targets[target.scalable_dimension] = OrderedDict()
+        if target.resource_id not in self.targets[target.scalable_dimension]:
+            self.targets[target.scalable_dimension][target.resource_id] = target
+        return target
+
+
+def _target_params_are_valid(namespace, r_id, dimension):
+    """ Check whether namespace, resource_id and dimension are valid and consistent with each other. """
+    is_valid = True
+    valid_namespaces = [n.value for n in ServiceNamespaceValueSet]
+    if namespace not in valid_namespaces:
+        is_valid = False
+    if dimension is not None:
+        try:
+            valid_dimensions = [d.value for d in ScalableDimensionValueSet]
+            d_namespace, d_resource_type, scaling_property = dimension.split(":")
+            resource_type, cluster, service = r_id.split("/")
+            if (
+                dimension not in valid_dimensions
+                or d_namespace != namespace
+                or resource_type != d_resource_type
+            ):
+                is_valid = False
+        except ValueError:
+            is_valid = False
+    if not is_valid:
+        raise AWSValidationException(
+            "Unsupported service namespace, resource type or scalable dimension"
+        )
+    return is_valid
+
+
+class FakeScalableTarget(BaseModel):
+    def __init__(
+        self, backend, service_namespace, resource_id, scalable_dimension, **kwargs
+    ):
+        self.applicationautoscaling_backend = backend
+        self.service_namespace = service_namespace
+        self.resource_id = resource_id
+        self.scalable_dimension = scalable_dimension
+        self.min_capacity = kwargs["min_capacity"]
+        self.max_capacity = kwargs["max_capacity"]
+        self.role_arn = kwargs["role_arn"]
+        self.suspended_state = kwargs["suspended_state"]
+        self.creation_time = time.time()
+
+    def update(self, **kwargs):
+        if kwargs["min_capacity"] is not None:
+            self.min_capacity = kwargs["min_capacity"]
+        if kwargs["max_capacity"] is not None:
+            self.max_capacity = kwargs["max_capacity"]
+
+
+applicationautoscaling_backends = {}
+for region_name, ecs_backend in ecs_backends.items():
+    applicationautoscaling_backends[region_name] = ApplicationAutoscalingBackend(
+        region_name, ecs_backend
+    )
diff --git a/moto/applicationautoscaling/responses.py b/moto/applicationautoscaling/responses.py
new file mode 100644
index 000000000..9a2905d79
--- /dev/null
+++ b/moto/applicationautoscaling/responses.py
@@ -0,0 +1,97 @@
+from __future__ import unicode_literals
+from moto.core.responses import BaseResponse
+import json
+from .models import (
+    applicationautoscaling_backends,
+    ScalableDimensionValueSet,
+    ServiceNamespaceValueSet,
+)
+from .exceptions import AWSValidationException
+
+
+class ApplicationAutoScalingResponse(BaseResponse):
+    @property
+    def applicationautoscaling_backend(self):
+        return applicationautoscaling_backends[self.region]
+
+    def describe_scalable_targets(self):
+        try:
+            self._validate_params()
+        except AWSValidationException as e:
+            return e.response()
+        service_namespace = self._get_param("ServiceNamespace")
+        resource_ids = self._get_param("ResourceIds")
+        scalable_dimension = self._get_param("ScalableDimension")
+        max_results = self._get_int_param("MaxResults", 50)
+        marker = self._get_param("NextToken")
+        all_scalable_targets = self.applicationautoscaling_backend.describe_scalable_targets(
+            service_namespace, resource_ids, scalable_dimension
+        )
+        start = int(marker) + 1 if marker else 0
+        next_token = None
+        scalable_targets_resp = all_scalable_targets[start : start + max_results]
+        if len(all_scalable_targets) > start + max_results:
+            next_token = str(len(scalable_targets_resp) - 1)
+        targets = [_build_target(t) for t in scalable_targets_resp]
+        return json.dumps({"ScalableTargets": targets, "NextToken": next_token})
+
+    def register_scalable_target(self):
+        """ Registers or updates a scalable target. """
+        try:
+            self._validate_params()
+            self.applicationautoscaling_backend.register_scalable_target(
+                self._get_param("ServiceNamespace"),
+                self._get_param("ResourceId"),
+                self._get_param("ScalableDimension"),
+                min_capacity=self._get_int_param("MinCapacity"),
+                max_capacity=self._get_int_param("MaxCapacity"),
+                role_arn=self._get_param("RoleARN"),
+                suspended_state=self._get_param("SuspendedState"),
+            )
+        except AWSValidationException as e:
+            return e.response()
+        return json.dumps({})
+
+    def _validate_params(self):
+        """ Validate parameters.
+            TODO Integrate this validation with the validation in models.py
+        """
+        namespace = self._get_param("ServiceNamespace")
+        dimension = self._get_param("ScalableDimension")
+        messages = []
+        dimensions = [d.value for d in ScalableDimensionValueSet]
+        message = None
+        if dimension is not None and dimension not in dimensions:
+            messages.append(
+                "Value '{}' at 'scalableDimension' "
+                "failed to satisfy constraint: Member must satisfy enum value set: "
+                "{}".format(dimension, dimensions)
+            )
+        namespaces = [n.value for n in ServiceNamespaceValueSet]
+        if namespace is not None and namespace not in namespaces:
+            messages.append(
+                "Value '{}' at 'serviceNamespace' "
+                "failed to satisfy constraint: Member must satisfy enum value set: "
+                "{}".format(namespace, namespaces)
+            )
+        if len(messages) == 1:
+            message = "1 validation error detected: {}".format(messages[0])
+        elif len(messages) > 1:
+            message = "{} validation errors detected: {}".format(
+                len(messages), "; ".join(messages)
+            )
+        if message:
+            raise AWSValidationException(message)
+
+
+def _build_target(t):
+    return {
+        "CreationTime": t.creation_time,
+        "ServiceNamespace": t.service_namespace,
+        "ResourceId": t.resource_id,
+        "RoleARN": t.role_arn,
+        "ScalableDimension": t.scalable_dimension,
+        "MaxCapacity": t.max_capacity,
+        "MinCapacity": t.min_capacity,
+        "SuspendedState": t.suspended_state,
+    }
diff --git a/moto/applicationautoscaling/urls.py b/moto/applicationautoscaling/urls.py
new file mode 100644
index 000000000..8a608f954
--- /dev/null
+++ b/moto/applicationautoscaling/urls.py
@@ -0,0 +1,8 @@
+from __future__ import unicode_literals
+from .responses import ApplicationAutoScalingResponse
+
+url_bases = ["https?://application-autoscaling.(.+).amazonaws.com"]
+
+url_paths = {
+    "{0}/$": ApplicationAutoScalingResponse.dispatch,
+}
diff --git a/moto/applicationautoscaling/utils.py b/moto/applicationautoscaling/utils.py
new file mode 100644
index 000000000..72330c508
--- /dev/null
+++ b/moto/applicationautoscaling/utils.py
@@ -0,0 +1,10 @@
+from six.moves.urllib.parse import urlparse
+
+
+def region_from_applicationautoscaling_url(url):
+    domain = urlparse(url).netloc
+
+    if "." in domain:
+        return domain.split(".")[1]
+    else:
+        return "us-east-1"
diff --git a/moto/backends.py b/moto/backends.py
index 44534d574..6f612bf1f 100644
--- a/moto/backends.py
+++ b/moto/backends.py
@@ -6,6 +6,10 @@ BACKENDS = {
     "acm": ("acm", "acm_backends"),
     "apigateway": ("apigateway", "apigateway_backends"),
     "athena": ("athena", "athena_backends"),
+    "applicationautoscaling": (
+        "applicationautoscaling",
+        "applicationautoscaling_backends",
+    ),
     "autoscaling": ("autoscaling", "autoscaling_backends"),
     "batch": ("batch", "batch_backends"),
     "cloudformation": ("cloudformation", "cloudformation_backends"),
diff --git a/moto/core/models.py b/moto/core/models.py
index ba4564e4a..26ee1a1f5 100644
--- a/moto/core/models.py
+++ b/moto/core/models.py
@@ -10,6 +10,7 @@ import six
 import types
 from io import BytesIO
 from collections import defaultdict
+from botocore.config import Config
 from botocore.handlers import BUILTIN_HANDLERS
 from botocore.awsrequest import AWSResponse
 from six.moves.urllib.parse import urlparse
@@ -416,6 +417,13 @@ class ServerModeMockAWS(BaseMockAWS):
         import mock
 
         def fake_boto3_client(*args, **kwargs):
+            region = self._get_region(*args, **kwargs)
+            if region:
+                if "config" in kwargs:
+                    kwargs["config"].__dict__["user_agent_extra"] += " region/" + region
+                else:
+                    config = Config(user_agent_extra="region/" + region)
+                    kwargs["config"] = config
             if "endpoint_url" not in kwargs:
                 kwargs["endpoint_url"] = "http://localhost:5000"
             return real_boto3_client(*args, **kwargs)
@@ -463,6 +471,14 @@ class ServerModeMockAWS(BaseMockAWS):
         if six.PY2:
             self._httplib_patcher.start()
 
+    def _get_region(self, *args, **kwargs):
+        if "region_name" in kwargs:
+            return kwargs["region_name"]
+        if type(args) == tuple and len(args) == 2:
+            service, region = args
+            return region
+        return None
+
     def disable_patching(self):
         if self._client_patcher:
             self._client_patcher.stop()
diff --git a/moto/core/responses.py b/moto/core/responses.py
index c52e89898..676d7549d 100644
--- a/moto/core/responses.py
+++ b/moto/core/responses.py
@@ -188,6 +188,9 @@ class BaseResponse(_TemplateEnvironmentMixin, ActionAuthenticatorMixin):
     default_region = "us-east-1"
     # to extract region, use [^.]
     region_regex = re.compile(r"\.(?P<region>[a-z]{2}-[a-z]+-\d{1})\.amazonaws\.com")
+    region_from_useragent_regex = re.compile(
+        r"region/(?P<region>[a-z]{2}-[a-z]+-\d{1})"
+    )
     param_list_regex = re.compile(r"(.*)\.(\d+)\.")
     access_key_regex = re.compile(
         r"AWS.*(?P<access_key>(?<![A-Z0-9])[A-Z0-9]{20}(?![A-Z0-9]))[:/]"
@@ -272,9 +275,14 @@ class BaseResponse(_TemplateEnvironmentMixin, ActionAuthenticatorMixin):
         self.response_headers = {"server": "amazon.com"}
 
     def get_region_from_url(self, request, full_url):
-        match = self.region_regex.search(full_url)
-        if match:
-            region = match.group(1)
+        url_match = self.region_regex.search(full_url)
+        user_agent_match = self.region_from_useragent_regex.search(
+            request.headers.get("User-Agent", "")
+        )
+        if url_match:
+            region = url_match.group(1)
+        elif user_agent_match:
+            region = user_agent_match.group(1)
         elif (
             "Authorization" in request.headers
             and "AWS4" in request.headers["Authorization"]
diff --git a/tests/test_applicationautoscaling/__init__.py b/tests/test_applicationautoscaling/__init__.py
new file mode 100644
index 000000000..baffc4882
--- /dev/null
+++ b/tests/test_applicationautoscaling/__init__.py
@@ -0,0 +1 @@
+from __future__ import unicode_literals
diff --git a/tests/test_applicationautoscaling/test_applicationautoscaling.py b/tests/test_applicationautoscaling/test_applicationautoscaling.py
new file mode 100644
index 000000000..632804992
--- /dev/null
+++ b/tests/test_applicationautoscaling/test_applicationautoscaling.py
@@ -0,0 +1,189 @@
+from __future__ import unicode_literals
+import boto3
+from moto import mock_applicationautoscaling, mock_ecs
+import sure  # noqa
+from nose.tools import with_setup
+
+DEFAULT_REGION = "us-east-1"
+DEFAULT_ECS_CLUSTER = "default"
+DEFAULT_ECS_TASK = "test_ecs_task"
+DEFAULT_ECS_SERVICE = "sample-webapp"
+DEFAULT_SERVICE_NAMESPACE = "ecs"
+DEFAULT_RESOURCE_ID = "service/{}/{}".format(DEFAULT_ECS_CLUSTER, DEFAULT_ECS_SERVICE)
+DEFAULT_SCALABLE_DIMENSION = "ecs:service:DesiredCount"
+DEFAULT_MIN_CAPACITY = 1
+DEFAULT_MAX_CAPACITY = 1
+DEFAULT_ROLE_ARN = "test:arn"
+DEFAULT_SUSPENDED_STATE = {
+    "DynamicScalingInSuspended": True,
+    "DynamicScalingOutSuspended": True,
+    "ScheduledScalingSuspended": True,
+}
+
+
+def _create_ecs_defaults(ecs, create_service=True):
+    _ = ecs.create_cluster(clusterName=DEFAULT_ECS_CLUSTER)
+    _ = ecs.register_task_definition(
+        family=DEFAULT_ECS_TASK,
+        containerDefinitions=[
+            {
+                "name": "hello_world",
+                "image": "docker/hello-world:latest",
+                "cpu": 1024,
+                "memory": 400,
+                "essential": True,
+                "environment": [
+                    {"name": "AWS_ACCESS_KEY_ID", "value": "SOME_ACCESS_KEY"}
+                ],
+                "logConfiguration": {"logDriver": "json-file"},
+            }
+        ],
+    )
+    if create_service:
+        _ = ecs.create_service(
+            cluster=DEFAULT_ECS_CLUSTER,
+            serviceName=DEFAULT_ECS_SERVICE,
+            taskDefinition=DEFAULT_ECS_TASK,
+            desiredCount=2,
+        )
+
+
+@mock_ecs
+@mock_applicationautoscaling
+def test_describe_scalable_targets_one_basic_ecs_success():
+    ecs = boto3.client("ecs", region_name=DEFAULT_REGION)
+    _create_ecs_defaults(ecs)
+    client = boto3.client("application-autoscaling", region_name=DEFAULT_REGION)
+    client.register_scalable_target(
+        ServiceNamespace=DEFAULT_SERVICE_NAMESPACE,
+        ResourceId=DEFAULT_RESOURCE_ID,
+        ScalableDimension=DEFAULT_SCALABLE_DIMENSION,
+    )
+    response = client.describe_scalable_targets(
+        ServiceNamespace=DEFAULT_SERVICE_NAMESPACE
+    )
+    response["ResponseMetadata"]["HTTPStatusCode"].should.equal(200)
+    len(response["ScalableTargets"]).should.equal(1)
+    t = response["ScalableTargets"][0]
+    t.should.have.key("ServiceNamespace").which.should.equal(DEFAULT_SERVICE_NAMESPACE)
+    t.should.have.key("ResourceId").which.should.equal(DEFAULT_RESOURCE_ID)
+    t.should.have.key("ScalableDimension").which.should.equal(
+        DEFAULT_SCALABLE_DIMENSION
+    )
+    t.should.have.key("CreationTime").which.should.be.a("datetime.datetime")
+
+
+@mock_ecs
+@mock_applicationautoscaling
+def test_describe_scalable_targets_one_full_ecs_success():
+    ecs = boto3.client("ecs", region_name=DEFAULT_REGION)
+    _create_ecs_defaults(ecs)
+    client = boto3.client("application-autoscaling", region_name=DEFAULT_REGION)
+    register_scalable_target(client)
+    response = client.describe_scalable_targets(
+        ServiceNamespace=DEFAULT_SERVICE_NAMESPACE
+    )
+    response["ResponseMetadata"]["HTTPStatusCode"].should.equal(200)
+    len(response["ScalableTargets"]).should.equal(1)
+    t = response["ScalableTargets"][0]
+    t.should.have.key("ServiceNamespace").which.should.equal(DEFAULT_SERVICE_NAMESPACE)
+    t.should.have.key("ResourceId").which.should.equal(DEFAULT_RESOURCE_ID)
+    t.should.have.key("ScalableDimension").which.should.equal(
+        DEFAULT_SCALABLE_DIMENSION
+    )
+    t.should.have.key("MinCapacity").which.should.equal(DEFAULT_MIN_CAPACITY)
+    t.should.have.key("MaxCapacity").which.should.equal(DEFAULT_MAX_CAPACITY)
+    t.should.have.key("RoleARN").which.should.equal(DEFAULT_ROLE_ARN)
+    t.should.have.key("CreationTime").which.should.be.a("datetime.datetime")
+    t.should.have.key("SuspendedState")
+    t["SuspendedState"]["DynamicScalingInSuspended"].should.equal(
+        DEFAULT_SUSPENDED_STATE["DynamicScalingInSuspended"]
+    )
+
+
+@mock_ecs
+@mock_applicationautoscaling
+def test_describe_scalable_targets_only_return_ecs_targets():
+    ecs = boto3.client("ecs", region_name=DEFAULT_REGION)
+    _create_ecs_defaults(ecs, create_service=False)
+    _ = ecs.create_service(
+        cluster=DEFAULT_ECS_CLUSTER,
+        serviceName="test1",
+        taskDefinition=DEFAULT_ECS_TASK,
+        desiredCount=2,
+    )
+    _ = ecs.create_service(
+        cluster=DEFAULT_ECS_CLUSTER,
+        serviceName="test2",
+        taskDefinition=DEFAULT_ECS_TASK,
+        desiredCount=2,
+    )
+    client = boto3.client("application-autoscaling", region_name=DEFAULT_REGION)
+    register_scalable_target(
+        client,
+        ServiceNamespace="ecs",
+        ResourceId="service/{}/test1".format(DEFAULT_ECS_CLUSTER),
+    )
+    register_scalable_target(
+        client,
+        ServiceNamespace="ecs",
+        ResourceId="service/{}/test2".format(DEFAULT_ECS_CLUSTER),
+    )
+    register_scalable_target(
+        client,
+        ServiceNamespace="elasticmapreduce",
+        ResourceId="instancegroup/j-2EEZNYKUA1NTV/ig-1791Y4E1L8YI0",
+        ScalableDimension="elasticmapreduce:instancegroup:InstanceCount",
+    )
+    response = client.describe_scalable_targets(
+        ServiceNamespace=DEFAULT_SERVICE_NAMESPACE
+    )
+    response["ResponseMetadata"]["HTTPStatusCode"].should.equal(200)
+    len(response["ScalableTargets"]).should.equal(2)
+
+
+@mock_ecs
+@mock_applicationautoscaling
+def test_describe_scalable_targets_next_token_success():
+    ecs = boto3.client("ecs", region_name=DEFAULT_REGION)
+    _create_ecs_defaults(ecs, create_service=False)
+    client = boto3.client("application-autoscaling", region_name=DEFAULT_REGION)
+    for i in range(0, 100):
+        _ = ecs.create_service(
+            cluster=DEFAULT_ECS_CLUSTER,
+            serviceName=str(i),
+            taskDefinition=DEFAULT_ECS_TASK,
+            desiredCount=2,
+        )
+        register_scalable_target(
+            client,
+            ServiceNamespace="ecs",
+            ResourceId="service/{}/{}".format(DEFAULT_ECS_CLUSTER, i),
+        )
+    response = client.describe_scalable_targets(
+        ServiceNamespace=DEFAULT_SERVICE_NAMESPACE
+    )
+    response["ResponseMetadata"]["HTTPStatusCode"].should.equal(200)
+    len(response["ScalableTargets"]).should.equal(50)
+    response["ScalableTargets"][0]["ResourceId"].should.equal("service/default/0")
+    response.should.have.key("NextToken").which.should.equal("49")
+    response = client.describe_scalable_targets(
+        ServiceNamespace=DEFAULT_SERVICE_NAMESPACE, NextToken=str(response["NextToken"])
+    )
+    response["ResponseMetadata"]["HTTPStatusCode"].should.equal(200)
+    len(response["ScalableTargets"]).should.equal(50)
+    response["ScalableTargets"][0]["ResourceId"].should.equal("service/default/50")
+    response.should_not.have.key("NextToken")
+
+
+def register_scalable_target(client, **kwargs):
+    """ Build a default scalable target object for use in tests. """
+    return client.register_scalable_target(
+        ServiceNamespace=kwargs.get("ServiceNamespace", DEFAULT_SERVICE_NAMESPACE),
+        ResourceId=kwargs.get("ResourceId", DEFAULT_RESOURCE_ID),
+        ScalableDimension=kwargs.get("ScalableDimension", DEFAULT_SCALABLE_DIMENSION),
+        MinCapacity=kwargs.get("MinCapacity", DEFAULT_MIN_CAPACITY),
+        MaxCapacity=kwargs.get("MaxCapacity", DEFAULT_MAX_CAPACITY),
+        RoleARN=kwargs.get("RoleARN", DEFAULT_ROLE_ARN),
+        SuspendedState=kwargs.get("SuspendedState", DEFAULT_SUSPENDED_STATE),
+    )
diff --git a/tests/test_applicationautoscaling/test_validation.py b/tests/test_applicationautoscaling/test_validation.py
new file mode 100644
index 000000000..02281ab05
--- /dev/null
+++ b/tests/test_applicationautoscaling/test_validation.py
@@ -0,0 +1,123 @@
+from __future__ import unicode_literals
+import boto3
+from moto import mock_applicationautoscaling, mock_ecs
+from moto.applicationautoscaling import models
+from moto.applicationautoscaling.exceptions import AWSValidationException
+from botocore.exceptions import ParamValidationError
+from nose.tools import assert_raises
+import sure  # noqa
+from botocore.exceptions import ClientError
+from parameterized import parameterized
+from .test_applicationautoscaling import register_scalable_target
+
+DEFAULT_REGION = "us-east-1"
+DEFAULT_ECS_CLUSTER = "default"
+DEFAULT_ECS_TASK = "test_ecs_task"
+DEFAULT_ECS_SERVICE = "sample-webapp"
+DEFAULT_SERVICE_NAMESPACE = "ecs"
+DEFAULT_RESOURCE_ID = "service/{}/{}".format(DEFAULT_ECS_CLUSTER, DEFAULT_ECS_SERVICE)
+DEFAULT_SCALABLE_DIMENSION = "ecs:service:DesiredCount"
+DEFAULT_MIN_CAPACITY = 1
+DEFAULT_MAX_CAPACITY = 1
+DEFAULT_ROLE_ARN = "test:arn"
+
+
+@mock_applicationautoscaling
+def test_describe_scalable_targets_no_params_should_raise_param_validation_errors():
+    client = boto3.client("application-autoscaling", region_name=DEFAULT_REGION)
+    with assert_raises(ParamValidationError):
+        client.describe_scalable_targets()
+
+
+@mock_applicationautoscaling
+def test_register_scalable_target_no_params_should_raise_param_validation_errors():
+    client = boto3.client("application-autoscaling", region_name=DEFAULT_REGION)
+    with assert_raises(ParamValidationError):
+        client.register_scalable_target()
+
+
+@mock_applicationautoscaling
+def test_register_scalable_target_with_none_service_namespace_should_raise_param_validation_errors():
+    client = boto3.client("application-autoscaling", region_name=DEFAULT_REGION)
+    with assert_raises(ParamValidationError):
+        register_scalable_target(client, ServiceNamespace=None)
+
+
+@mock_applicationautoscaling
+def test_describe_scalable_targets_with_invalid_scalable_dimension_should_return_validation_exception():
+    client = boto3.client("application-autoscaling", region_name=DEFAULT_REGION)
+
+    with assert_raises(ClientError) as err:
+        response = client.describe_scalable_targets(
+            ServiceNamespace=DEFAULT_SERVICE_NAMESPACE, ScalableDimension="foo",
+        )
+        err.response["Error"]["Code"].should.equal("ValidationException")
+        err.response["Error"]["Message"].split(":")[0].should.look_like(
+            "1 validation error detected"
+        )
+        err.response["ResponseMetadata"]["HTTPStatusCode"].should.equal(400)
+
+
+@mock_applicationautoscaling
+def test_describe_scalable_targets_with_invalid_service_namespace_should_return_validation_exception():
+    client = boto3.client("application-autoscaling", region_name=DEFAULT_REGION)
+
+    with assert_raises(ClientError) as err:
+        response = client.describe_scalable_targets(
+            ServiceNamespace="foo", ScalableDimension=DEFAULT_SCALABLE_DIMENSION,
+        )
+        err.response["Error"]["Code"].should.equal("ValidationException")
+        err.response["Error"]["Message"].split(":")[0].should.look_like(
+            "1 validation error detected"
+        )
+        err.response["ResponseMetadata"]["HTTPStatusCode"].should.equal(400)
+
+
+@mock_applicationautoscaling
+def test_describe_scalable_targets_with_multiple_invalid_parameters_should_return_validation_exception():
+    client = boto3.client("application-autoscaling", region_name=DEFAULT_REGION)
+
+    with assert_raises(ClientError) as err:
+        response = client.describe_scalable_targets(
+            ServiceNamespace="foo", ScalableDimension="bar",
+        )
+        err.response["Error"]["Code"].should.equal("ValidationException")
+        err.response["Error"]["Message"].split(":")[0].should.look_like(
+            "2 validation errors detected"
+        )
+        err.response["ResponseMetadata"]["HTTPStatusCode"].should.equal(400)
+
+
+@mock_ecs
+@mock_applicationautoscaling
+def test_register_scalable_target_ecs_with_non_existent_service_should_return_validation_exception():
+    client = boto3.client("application-autoscaling", region_name=DEFAULT_REGION)
+    resource_id = "service/{}/foo".format(DEFAULT_ECS_CLUSTER)
+
+    with assert_raises(ClientError) as err:
+        register_scalable_target(client, ServiceNamespace="ecs", ResourceId=resource_id)
+        err.response["Error"]["Code"].should.equal("ValidationException")
+        err.response["Error"]["Message"].should.equal(
+            "ECS service doesn't exist: {}".format(resource_id)
+        )
+        err.response["ResponseMetadata"]["HTTPStatusCode"].should.equal(400)
+
+
+@parameterized(
+    [
+        ("ecs", "service/default/test-svc", "ecs:service:DesiredCount", True),
+        ("ecs", "banana/default/test-svc", "ecs:service:DesiredCount", False),
+        ("rds", "service/default/test-svc", "ecs:service:DesiredCount", False),
+    ]
+)
+def test_target_params_are_valid_success(namespace, r_id, dimension, expected):
+    if expected is True:
+        models._target_params_are_valid(namespace, r_id, dimension).should.equal(
+            expected
+        )
+    else:
+        with assert_raises(AWSValidationException):
+            models._target_params_are_valid(namespace, r_id, dimension)
+
+
+# TODO add a test for not-supplied MinCapacity or MaxCapacity (ValidationException)
diff --git a/tests/test_cognitoidp/test_cognitoidp.py b/tests/test_cognitoidp/test_cognitoidp.py
index 37e1a56a3..9c4b8de49 100644
--- a/tests/test_cognitoidp/test_cognitoidp.py
+++ b/tests/test_cognitoidp/test_cognitoidp.py
@@ -1243,6 +1243,38 @@ def test_change_password():
     result["AuthenticationResult"].should_not.be.none
 
 
+@mock_cognitoidp
+def test_change_password__using_custom_user_agent_header():
+    # https://github.com/spulec/moto/issues/3098
+    # As the admin_initiate_auth-method is unauthenticated, we use the user-agent header to pass in the region
+    # This test verifies this works, even if we pass in our own user-agent header
+    from botocore.config import Config
+
+    my_config = Config(user_agent_extra="more/info", signature_version="v4")
+    conn = boto3.client("cognito-idp", "us-west-2", config=my_config)
+
+    outputs = authentication_flow(conn)
+
+    # Take this opportunity to test change_password, which requires an access token.
+    newer_password = str(uuid.uuid4())
+    conn.change_password(
+        AccessToken=outputs["access_token"],
+        PreviousPassword=outputs["password"],
+        ProposedPassword=newer_password,
+    )
+
+    # Log in again, which should succeed without a challenge because the user is no
+    # longer in the force-new-password state.
+    result = conn.admin_initiate_auth(
+        UserPoolId=outputs["user_pool_id"],
+        ClientId=outputs["client_id"],
+        AuthFlow="ADMIN_NO_SRP_AUTH",
+        AuthParameters={"USERNAME": outputs["username"], "PASSWORD": newer_password},
+    )
+
+    result["AuthenticationResult"].should_not.be.none
+
+
 @mock_cognitoidp
 def test_forgot_password():
     conn = boto3.client("cognito-idp", "us-west-2")