Illustration of a hiker and their dog sitting on a rocky outcrop at sunset, looking out over layered mountain ridges and a river winding through the valley below.

Pourquoi CloudHound

Conçu pour les migrations AWS et OCI par des personnes qui les ont menées

CloudHound réunit la découverte, la modélisation des coûts et la migration dans une seule plateforme. Il mesure chaque serveur, le chiffre selon les tarifs AWS et OCI en vigueur, et le déplace vers AWS ou OCI.

Collected data

Each record is a sample from a running server: CPU, memory, disk, processes and network flows. The same data is used to price, plan and migrate the estate.

records collected, and counting

From first install to cutover

A typical engagement, day by day.

Day 0
Install
One command per server, or agentless over SSH and WinRM. Inventory, processes and connections arrive instantly.
Hour 1
First utilization
CPU, memory and disk are sampled every minute, and each server is described within minutes.
Day 3
Costed estate
Every server priced against live AWS and OCI rates: on-demand, reserved and savings plan side by side.
Day 5
Dependency mapping
Five days of network flows are clustered into applications.
Day 7
Wave plan
A wave plan is generated from the dependency map. You can drag applications between waves.
Cutover
Migration
Each wave replicates, verifies block by block, converts and boots on EC2. 125 servers in 12 minutes.

Data residency

You choose one of seventeen AWS regions when you create a portfolio. Agents in any country report to that region over outbound-only TLS, and their data is processed and stored only there.

Agents reporting

0Records received since you opened this page

Right-sizing

The agent records high fidelity utilization at one-minute resolution, continuously: CPU, memory and the operation and byte counters of every drive. Each server is sized on 7 days of those minutes, to its measured use. Here is one week of an illustrative application server, provisioned at 8 vCPU and 32 GiB with nine drives, its CPU and memory shaped on a production server’s week.

CPUPer minute, 7 days

MemoryPer minute, 7 days

Disk throughput9 drives stacked, per minute, 7 days, MiB/s. Legend: p99 per drive

C: System1.5 MiB/s, 151 IOPS

D: Application28.5 MiB/s, 744 IOPS

E: Logs13.8 MiB/s, 514 IOPS

F: Uploads30.9 MiB/s, 367 IOPS

G: Cache32.0 MiB/s, 2,296 IOPS

H: Queue10.1 MiB/s, 1,330 IOPS

I: Reports37.1 MiB/s, 979 IOPS

J: Backups79.6 MiB/s, 366 IOPS

K: Archive15.6 MiB/s, 334 IOPS

Provisioned

8 vCPU, 32 GiB, 9 drives

CPU needed

8 × 63% × 1.2 = 6.04rounded up to 7 vCPU

Memory needed

32 × 65% × 1.1 = 22.9 GiBof 32 GiB provisioned

Disk throughput needed

126 MiB/s = 132 MB/s

Disk IOPS needed

4,650 IOPS

Recommended

m6a.2xlarge8 vCPU, 32 GiB, 312 MB/s and 12,000 IOPS to EBS. The smallest size with enough vCPU and memory. m5a.2xlarge costs within 0.5% of it, so CloudHound picks the newer m6a.

Volumes

9 × gp3Each volume provisioned from its own p99. The busiest drive needs 2,296 IOPS and 80 MiB/s, inside the 3,000 IOPS and 125 MiB/s every gp3 volume includes, so nothing is provisioned above the baseline. io2 is recommended only above 80,000 IOPS or 2,000 MiB/s

One server after a week of discovery

What CloudHound knows about a typical application server after a week of discovery. The server is illustrative; every field on it is one the agent collects.

ORDERS-APP-02

Windows Server 2016 StandardVMware VM8 vCPU, 32 GiB10.20.14.37

7 daysObserved

10,080Samples per metric

289Fields per sample

0Connections mapped

Order management application server

Hosts the customer order portal and its API in IIS, and a Java service under Tomcat that takes orders from the EDI gateway and archives documents to Amazon S3. A local SQL Server instance holds the order databases; reporting and the warehouse system read from it on 1433. Busy 07:00 to 19:00 on weekdays, quiet overnight apart from the 02:00 backup.

Migration effort2 of 5

Lift and shift. Every dependency is reached by name or IP, with no hardware dongles or host-bound licenses.

Linux replatforming3 of 5

The API and the Tomcat service move as they are. The portal is .NET Framework 4.8 and needs porting first.

Containerization3 of 5

order-sync is stateless and container ready. The database stays on EC2 or moves to RDS.

Right sizing

8 vCPU32 GiB today

m6a.2xlarge8 vCPU, 32 GiB

$520.93 a month on demand, $396.59 on a 3-year no-upfront Compute Savings Plan, Windows, US East (Ohio), plus $188.80 for nine right-sized gp3 volumes.

Traffic per process is the sum of the connections that process owns. Presence is the share of samples the process was running in.

DriveIOPS, 7 daysMiB/s, 7 daysVolume on AWS

C:SystemNTFS, SCSI 0:0, 78 of 120 GiB

gp3, 120 GiB, $9.60 a monthKept, the saving is under 10%3,000 IOPS, 125 MiB/s, the baseline

D:ApplicationNTFS, SCSI 0:1, 262 of 400 GiB

gp3, 400 GiB, $32.00 a monthKept, the taper lands on its size3,000 IOPS, 125 MiB/s, the baseline

E:LogsNTFS, SCSI 0:2, 41 of 100 GiB

gp3, 60 GiB, $4.80 a monthRight-sized from 100 GiB3,000 IOPS, 125 MiB/s, the baseline

F:UploadsNTFS, SCSI 0:3, 133 of 250 GiB

gp3, 190 GiB, $15.20 a monthRight-sized from 250 GiB3,000 IOPS, 125 MiB/s, the baseline

G:CacheNTFS, SCSI 0:4, 22 of 60 GiB

gp3, 40 GiB, $3.20 a monthRight-sized from 60 GiB3,000 IOPS, 125 MiB/s, the baseline

H:QueueNTFS, SCSI 0:5, 9 of 40 GiB

gp3, 20 GiB, $1.60 a monthRight-sized from 40 GiB3,000 IOPS, 125 MiB/s, the baseline

I:ReportsNTFS, SCSI 0:6, 57 of 150 GiB

gp3, 80 GiB, $6.40 a monthRight-sized from 150 GiB3,000 IOPS, 125 MiB/s, the baseline

J:BackupsNTFS, SCSI 0:8, 388 of 600 GiB

gp3, 600 GiB, $48.00 a monthKept, the saving is under 10%3,000 IOPS, 125 MiB/s, the baseline

K:ArchiveNTFS, SCSI 0:9, 610 of 1000 GiB

gp3, 850 GiB, $68.00 a monthRight-sized from 1,000 GiB3,000 IOPS, 125 MiB/s, the baseline

Web applications

Background services

Network interfaces

SQL Server Enterprise features

CPUp99 63%

Memoryp99 65%

Networkp99 1.9 MiB/s

Disk IOPS, summedp99 4,649 IOPS

High fidelity utilization: one sample every 60 seconds, 10,080 per metric over the week. Dashed navy line is the 99th percentile, orange is the sizing buffer above it, 20 percent on CPU and 10 percent on memory.

Migration readiness

Reclaimable before migration

Dependency mapping

An illustrative estate, drawn the way CloudHound draws yours from the network flows between your servers: 145 servers, 780 connections, grouped into applications. Drag a server, hover to see what it talks to, click an application to isolate it.

Wave planning

CloudHound sees which servers talk to each other and uses that traffic to group them into applications. Each application is then placed in a migration wave with the databases it depends on, so nothing moves without the systems it needs. This example follows 1,000 servers from discovery to a ten-wave plan. Drag to rotate it.

1 of 6

Illustrative estate of 1,000 servers in 41 applications.

What moves where

CloudHound moves each server to AWS in one of two ways. It can rehost the whole server on Amazon EC2 as an exact, verified copy. It can also replatform the server's databases directly to Amazon RDS for SQL Server, PostgreSQL, MySQL or MariaDB.

Database server

Windows Server 2022, SQL Server 2022

Operating system
Applications and services
Databases
Data files

In use

Amazon EC2

r6a.xlarge

Waiting

Automated QA after launch

When the server boots on Amazon EC2, the CloudHound agent on it checks it against the source: every service, port, database, volume and network path the source used.

Waiting for the server to bootExample: a Windows Server 2019 application server with IIS and SQL Server

0/26

Collecting information from the server

How to fix it

Neither issue is on the server. The route table of its subnet in AWS has no route to two networks on premises. In the VPC console, open that route table and add:

10.20.8.0/24 through the transit gateway, for the 3 app servers
10.20.9.0/24 through the transit gateway, for the mail relay

One disk, from first sync to go live

Block-level migration copies the disk's blocks, so the result is an exact clone that boots as the source did. The source is never modified. Reading raw blocks runs at disk speed: in our benchmark a 1 TiB disk copied in 255 seconds. Every pass after it moves only the blocks that changed, so the final sync carries only recent changes. Before launch, the target volume is read back and compared with the source block by block, then the disk is right-sized. The figures below are from an illustrative disk.

Initial sync-Blocks in use
Delta-Changed blocks only
Final sync-Moments before launch
Verify-Data blocks checked
Rightsize-Free space trimmed

Not yet copiedUnusedChanged on the sourceOn the targetVerifiedTrimmed

125 servers in 12 minutes

A replay of our September 2026 benchmark: 125 servers across 94 operating system variants and 140 disks totaling 9.3 TiB, every one replicated, verified block by block, converted, booted and confirmed on EC2.

0GiB/s0m00s

0 queued0 replicating0 launching0 booted

Ask CH

Ask CH, the CloudHound assistant, works from the same data as the rest of the platform. One message can produce the assessment pack and then start the migration.

PPTXA board readout in your own template: headline figures, end-of-support risk, right-sizing, the business case and the waves.
XLSXA workbook of real tables and live formulas over named assumptions, so the term or discount can be changed.
DOCXA written report with a contents page, numbered sections, figures and tables.
RUNThe migration itself. It lists the servers it will replicate and waits for your approval before anything moves.

The estate, the figures and the documents shown are illustrative.

Most AI assistants

›

‹1,164 servers

That's a database query.

Ask CH

Readiness assessment · 1,164 servers

Cost estimate · 6 pricing strategies

Regional pricing · 7 regions compared

Utilization · 7 days, 95th percentile

End-of-life data · 208 near end of support

Licensing · 226 SQL Server instances

Dependency map · 89 application groups

Wave plan · 7 named waves

Composing the readout, workbook and report

PPTXmap-assess-readout.pptx15 slides

XLSXbusiness-case.xlsxlive formulas

DOCXboard-report.docxyour template

business-case.xlsx · Excel

D6fx=SUM(Servers[Annual cost on AWS]) * (1 - Discount)

	A	B	C	D	E	F
1	Aldermont Health · AWS business case
2	Every figure below is a formula over the Servers and Assumptions sheets
3	Cost line (USD a year)	Today	On AWS	Saving	Share of AWS bill	Assumption
4	Block storage	3,510,000	2,310,000	1,200,000	35%	gp3, used data
5	Backup	2,640,000	1,930,000	710,000	29%	30-day retention
6	EC2 compute	3,100,000	1,180,000	1,920,000	18%	3-yr savings plan
7	Windows and SQL licenses	1,420,000	550,000	870,000	8%	License included
8	Network and support	790,000	650,000	140,000	10%	Business support
9	Total	11,460,000	6,620,000	4,840,000	42% lower	5 years: 24.2M
10
11	Term	3 years	Discount	0%	Region	us-west-2

SummaryServersAssumptionsPricing

board-report.docx · Word

FileHomeInsertDesignLayoutReferencesReviewView

Calibri11Heading 1Normal

Navigation

1. Executive summary

2. Assessment approach

3. Estate overview

4. Platform composition

5. Utilization and right-sizing

6. Licensing

7. The business case

7.1 Where the money goes

7.2 Pricing strategy

8. End-of-support exposure

8.1 Hotspots

8.2 Remediation by wave

9. Region selection

10. Dependency mapping

11. Migration waves

12. Recommendations

Appendix A. Assumptions

Appendix B. Server inventory

ALDERMONT HEALTH · AWS Migration AssessmentBoard report

7. The business case

Running the assessed estate on AWS is estimated at USD 6.62M a year against a current run cost of USD 11.46M, a reduction of 42%. Over five years the difference is USD 24.2M. The estimate uses live AWS price lists for US West (Oregon), a three-year compute savings plan with no upfront payment, and right-sizing on 7 days of high fidelity utilization.

USD 11.46Mcurrent annual run cost

USD 6.62Mannual cost on AWS

USD 24.2Msaved over five years

7.1 Where the money goes

Figure 7.1 Annual cost on AWS by component (USD)

Block storage2.31M

Backup1.93M

EC2 compute1.18M

Windows and SQL licenses0.55M

Network overhead0.31M

AWS Business support0.28M

Storage and backup make up 64% of the AWS bill. Tiering file-server and backup data to S3 and FSx is the largest saving still available after migration.

7.2 Pricing strategy

Table 7.2 Annual cost by commercial strategy

Strategy	Annual cost	vs on-demand
No right-sizing	USD 8.94M	+20%
On-demand	USD 7.48M	-
1-year savings plan	USD 6.84M	-9%
3-year savings plan (selected)	USD 6.62M	-11%

7.3 Five-year view

Figure 7.3 Cumulative saving against the current run cost (USD)

4.8M9.7M14.5M19.4M24.2M

Year 1Year 2Year 3Year 4Year 5

1. Prices from AWS price lists. 2. Right-sizing at the 95th percentile plus 20% CPU and 10% memory headroom.

AWS Migration AssessmentPage 7 of 24

ALDERMONT HEALTH · AWS Migration AssessmentBoard report

8. End-of-support exposure

36 servers already run software past vendor end of support and a further 172 reach it within twelve months, 18% of the estate. Migration is the point at which that risk is retired: each wave replaces or upgrades the unsupported software it carries.

Windows Server 2012 R2 and SQL Server 2012 to 2016 are the most urgent: 50 servers already past support. All of them are scheduled into waves 1 to 3.

8.1 Hotspots

Table 8.1 Software at or near end of support

Software	Servers	Support ends	Risk
Windows Server 2012 R2	14	Oct 2023	Critical
SQL Server 2012 / 2014	10	2022 / 2024	Critical
SQL Server 2016	26	Jul 2026	Critical
CentOS / Red Hat 7	19	Jun 2024	Critical
PostgreSQL 9.3 - 13	17	2018 - 2025	Critical
Windows Server 2016	168	Jan 2027	Warning
SQL Server 2017	27	Oct 2027	Warning
Windows Server 2019	452	Jan 2029	Watch

8.2 Remediation by wave

Wave 1 (50 servers) retires every Windows Server 2012 R2 host and moves its workloads to Windows Server 2022 on EC2.

Waves 2 and 3 carry the SQL Server estate, re-platforming 2012 to 2016 instances to supported versions or to RDS.

8.3 If the estate stays where it is

Staying on premises means buying extended security updates each year from 2027, and leaves 208 servers without vendor patches for part or all of that period.

Option	Security updates	Cost trend
Migrate in waves 1 to 3	Current on arrival	Falls 42%
Extended support	Paid, time-limited	Rises yearly
Do nothing	None	Flat, risk grows

3. Counts are servers except PostgreSQL (instances). 4. Windows Server 2019 extended support ends January 2029.

AWS Migration AssessmentPage 8 of 24

Page 7 of 246,412 wordsEnglish (United States)

Then the migration

Nothing moves until you approve it.

Approve these changes?

Start replication for novadesk-vda-01US West (Oregon) · initial sync

Start replication for novadesk-vda-02US West (Oregon) · initial sync

Start replication for novadesk-vda-03US West (Oregon) · initial sync

47 more servers in wave 1

ApproveCancel

Throughput

Transferred

vda-01

vda-02

vda-03

vda-04

vda-05

+ 45 more

Discovery to board pack to replication, in one conversation.

Ask CH, the CloudHound assistant

Who we are

CloudHound was founded by two people who have sat on both sides of an enterprise migration. Aidan has built software companies for twenty years on AWS, most recently working as a principal solutions architect at an AWS Premier Partner, running assessments and migrations. Alex spent four years at AWS in Seattle on strategic accounts, negotiating the largest enterprise agreements with its biggest customers.

Every migration we ran used the same tooling: discovery tools that were not reliable, assessments that took months and still left questions open, and a migration tool that was a separate product again. So we built one platform for discovery, assessment and migration.

CloudHound is an AWS ISV Partner holding the ISV Migration and Modernization, AI Software and Microsoft Workloads competencies. It is built for partners: you run it inside your clients’ estates, keep the engagement, and deliver the assessment in days.

Illustration of a meeting table by a window over a city skyline, with a laptop and network diagram on one side and a signed agreement and pen on the other.

Statue of a dog sitting on a stepped pedestal surrounded by tall trees and plants in soft sunlight.

Partenaire AWS

« CloudHound nous donne immédiatement une vision claire des environnements complexes - de la découverte à la planification par vagues - et nous permet d'obtenir des résultats plus rapides et plus sûrs pour nos clients. »

Responsable des migrations
AWS Premier Partner

Nous contacter

Conçu pour les migrations AWS et OCI par des personnes qui les ont menées

Collected data

From first install to cutover

Install

First utilization

Costed estate

Dependency mapping

Wave plan

Migration

Data residency

Right-sizing

One server after a week of discovery

Web applications

Background services

Network interfaces

SQL Server Enterprise features

Migration readiness

Reclaimable before migration

Dependency mapping

Wave planning

What moves where

Database server

Amazon EC2

Automated QA after launch

One disk, from first sync to go live

125 servers in 12 minutes

Ask CH

7. The business case

7.1 Where the money goes

7.2 Pricing strategy

7.3 Five-year view

8. End-of-support exposure

8.1 Hotspots

8.2 Remediation by wave

8.3 If the estate stays where it is

Who we are