DBAs Are From Saturn,

DBAs Are From Saturn,
and Storage Admins Are From Neptune
Why is it difficult for DBAs and Storage Admins to work
together? From our conversations with numerous IT teams,
we see patterns of miscommunication and misaligned
goals between DBAs and Storage Admins. The common
thread is a lack of understanding of the challenges faced
by the other team. We wrote this eBook to help DBAs
better understand the challenges of the Storage Admin,
and to propose a solution to reduce conflicting goals
between DBAs and Storage Admins.
Why Storage Admins
Frustrate Us
vs
Let’s face it, DBAs and Storage Admins don’t
always see eye-to-eye. You’ve just come out of the
hundredth meeting with your Application Owner
and Business Analyst to justify your architecture
recommendations and finally you’re on the same
page. The last thing you need is to convince some
hardware geek down the hall why you need RAID-10,
replication, or a 100GB cache for your LUNs.
After your meeting, he’ll probably just disappear for
2 weeks only to return with some totally different,
complicated LUN scheme that costs too much! It’s
that kind of frustration that drove Larry Ellison to
build ASM, right?
Most DBAs have a common set of frustrations
when dealing with their Storage Admins:
• Why is provisioning LUNs so slow?
• Why is the storage design so complicated?
• Why does the storage architecture ultimately
fail to provide the consistent performance and
availability that it promises?
To get answers and to solve this problem,
you need to get into your Storage Admin’s head.
You need to understand the rationale behind why
they approach database storage requests the way
they do.
Finally you need to understand that there is a
better way.
What you say, and
what your Storage
Admin hears:
DBA says: I need 10K IOPs
on everything.
Storage Admin hears: I have
champagne tastes and a
Budweiser budget.
DBA says: Let’s keep it simple
and RAID-10 everything.
Storage Admin hears: I’m
pretty sure a bigger number
is better.
DBA says: I need 2 full
snapshots every 15 mins for
reporting and analytics.
Inside the mind of a Storage Admin
Your Storage Admin is constantly reminded that his
infrastructure spend is the single biggest expenditure
across IT. Every year, he needs to show up with ways
to make his services more agile, higher performing
and more resilient; while driving down the cost.
Like his peers, he’s adopted some form of an IT-asa-Service Model deploying a Private Cloud. This
however, forces him to manage a highly complex,
mixed workload environment.
Storage Admin hears: I’m
going to ask for a re-layout
before your vacation to Hawaii!
Slow Disk Tier
Slowest Disk Tier
Even more frustrating, the incumbent storage
technology, the hybrid disk array, wasn’t designed to
economically deliver the maximum I/O, resiliency and
protection requirements for all provisioned storage.
Hybrid arrays require Storage Admins to manually “create” performance service tiers by
selective combinations of striping, I/O balancing, RAID sets, snapshots and replication.
Your simple request
for a primary LUN
may look like this:
But the necessary
architecture will
look like this:
Storage Admin hears: All your
15K disk are now belong to us.
DBA says: I’m not sure how
much this app will grow,
so let’s be conservative.
Flash Tier
Fast VPN
Flash
Tier
SOE1
OS
FC
Tier
Fast VPN
TempDB1
Data
TempDB2
Data
TempDB1
Log
TempDB2
Log
SOE2
Redo
FRA
TEMP
DATA
OLTP 1
Log
OLTP 2
Log
SQL
Server
OLTP1
DATA
Fast VPN
DSS Log
TempDB1
Data
SQL
Server
OLTP2
DATA
TempDB1
Log
SQL
Server
DSS
DATA
SATA
Tier
Oracle
SQL Server OLTP
SQL Server DSS
| | | | | | | | | | |
Unfortunately, that’s not where this story ends. The complicated
architecture your Storage Admin has created will now need to
be managed!
Latency Spike
There has to be a better way
So, let’s take a moment to recap. DBAs are looking for consistent
performance, resilience, and high availability. You’re looking for a
way to get this at a reasonable cost. You care about complexity
because when something goes wrong, you expect your Storage
Admin to be able to restore performance, maintain uptime or
recover data – quickly.
What if you could get the performance, resilience, and high
availability without sacrificing simplicity? What if this cost the same
or less than your existing storage? Why wouldn’t you try it?
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
When your application starts seeing latency spikes 9 months
from now, your Storage Admin only needs to refer to his handy
schema and spreadsheets to sort out what esoteric process
associated with another application, is conflicting with your
application. This kind of trouble-shooting takes hours or days,
if it happens at all. That’s why Storage Admins always ask you
to buy storage when performance dips. It’s too hard to fix the
existing storage!
All-Flash Arrays solve the DBA - Storage Admin conflict
by delivering:
• 100’s of thousands of IOPs with consistent sub-millisecond latency.
Worse yet, what if you have a catastrophic failure in your
stack? This is when you realize that your application’s storage
architecture didn’t allow for granular enough snapshots
because the storage wasn’t available or the cost was too high.
Now a lengthy rebuild from logs is required. While you didn’t
lose data, your RTO is blown.
• Better data resilience and protection.
• Cost per GB, at or below the cost of disk. Compression,
de-duplication and sophisticated management of consumer
grade flash have dropped the price of All-Flash Arrays to at or
below the cost of disk arrays.
• Consistent, predictable performance for all database workloads.
• Better data protection. The best All-Flash Arrays will offer
space-optimized snapshots that support tighter RPOs and RTOs
without the overhead. Again, no need to reduce protection to
cut costs!
Send Pure Storage your AWR report, and we’ll schedule a free
consultation with our Oracle performance experts to talk about
how Pure Storage’s FlashArray can improve the performance and
resiliency of your Oracle environment while actually reducing the
cost of your storage architecture (that should make the Storage
Admins happy, right?).
Storage Admins are from Neptune,
and DBAs are from Saturn
Why is the relationship with your DBA so hard to get
right? How do you collaborate with someone who is
used to working autonomously, and with specialized
infrastructure? From our work with the most successful
IT teams in the industry, we’ve found patterns of how the
most effective Storage Admins and DBAs align and solve
infrastructure problems. If you’re a Storage Admin, we
wrote this eBook for you!
DBAs are impossible to please,
so why bother?
vs
You’re in a meeting to talk about a new order
processing application that requires back-end
storage, which, as the storage admin, you will be
responsible for setting up and maintaining. You’ve
had conversations with the DBA about the storage
requirements for the project, which at the time,
were reasonable – the new application will need
to support 1000 users, driving 20,000 transactions
per minute at peak load. You and the DBA settled
on the storage required to support this – 5 TB,
5,000 IOPS, 5 ms latency and RAID-5 protection.
A complicated proposition, but you’ll get there with
some creative caching and tiering so it won’t break
the bank.
Six weeks later, the other shoe drops. Surprise,
surprise, the requirements have totally changed.
The project will now be required to support
5,000 users driving 100,000 transactions per
minute. The storage requirements have sky
rocketed – 20 TB, 20,000 IOPS, 0.5 ms latency
and RAID 10. The storage architecture you’ve
just spent weeks putting together is completely
useless. And there is no budget to purchase the
latest and greatest gear…
So how can you solve the problem? First, you need
to get into your DBA’s head to understand where
these requirements and last-minute requests are
coming from. What is most important to him/her?
Finally you need to understand that there is a
better way.
Storage Admin says: With the
current storage infrastructure,
I can’t support RAID 10. We just
don’t have enough capacity.
DBA hears: You are an enemy
of efficient data management.
Storage Admin says: To
support your IOPS and latency
requirements, we are going to
need performance HDD’s and
cache. We don’t have the budget
this quarter.
DBA hears: I am overcomplicating your simple
request for a moderate amount
of performance storage,
despite the fact that you fund
much of my storage budget.
DBA hears: ALERT! CODE RED!
Shoot first, ask questions later.
An Oracle DBA lives and dies by the performance and availability of enterprise
data. The DBA guards data as a lioness guards her cubs. Anything that threatens
the smooth functioning of an Oracle DBA’s systems is to be treated with extreme
prejudice. Unfortunately, that sometimes includes the storage infrastructure and the
people who manage it. Only the very best will do for an Oracle database – RAID
10, sub millisecond response time, oceans of capacity – these are a few of a DBA’s
favorite things.
If high availability and performance are bacon and Bon Jovi (things you love) then
downtime and high latency are Brussels sprouts and waiting in line at the post office –
to be eliminated at all cost.
While you would love to shower them with flash cache, short-stroked LUNs, and
snapshots – you can’t because that is way too expensive. And they just don’t
understand that trade-off.
What’s more, DBAs view your storage maintenance windows with suspicion.
They remember that time in 2011 when the maintenance window spilled over into
production hours due to a faulty .config file.
medium
IT
high
Storage Admin says: We have
a storage maintenance window
next week. Unfortunately we
may have to bring down the
database for a couple of hours.
Inside the mind of a DBA
low
What you say, and
what your DBA hears:
DB Performance
They remember 2011 because nothing is worse than the database being down –
and the CIO suddenly knows their name as “that guy responsible for the outage.”
There has to be a better way
If you had a wish list, it would probably look a little like this:
••
••
• •••••
$
All-Flash Arrays deliver:
• A simple, easy way to manage storage infrastructure, without
tiers/tears!
• Consistent, predictable performance across all workloads.
• A resilient, highly available storage platform that is nondisruptive to the rest of the data center’s infrastructure.
• Reduction in overall spend on storage infrastructure, without
compromising on any of the above.
What if you could have everything that’s on this list? Pure
Storage’s FlashArray can help dramatically simplify your storage
infrastructure, deliver sub millisecond latencies and be completely
non-disruptive to the rest of the infrastructure. And all for a price
that is less than the cost of traditional arrays.
20
DB Latency (ms)
Storage Admins are constantly battling to find a balance between
supporting the needs of the business and staying within budget
– often by designing tiered storage architectures. Yet, we have
to deal with managing this increasingly complex environment,
while trying to be agile and move at speeds that are dictated by
the business. When we aren’t agile or cheap, we can’t be good
partners to the business.
10
0 1
Pure Storage
100% Flash
Tier 1
Legacy Disk
5
10
15
20
Time
25
30
35
40
Don’t take our word for it, try it out for yourselves. Join forces
with your DBA to do a POC on the FlashArray for their Oracle
database. It could be the start of a beautiful relationship.
Pure Storage.
A better way.
Now you are
in a better space.