In the previous episode we discussed the basics of using indexes, what they are, how to use them, along with some common problems and solutions. One of the problems you will certainly face as time passes is index fragmentation. While indexes will certainly help your database perform well, they do require care and feeding–meaning there is some administrative overhead.
In this episode, we continue our conversation with Randolph West and discuss what we can do to take care of indexes and keep them well maintained over time. There is an option to use default maintenance plans and to do fragmentation and updating statistics, but should you use them? What are the alternatives? We didn’t forget to consider strategies for deciding which indexes to keep and which can be dropped.
“That’s purely Microsoft’s fault because generally, people are using the maintenance wizard are using it because they don’t know how to do it with [other] tools.“
“You still need to update the statistics after you’ve done a reorganization which is something that people tend to forget”
Listen to Learn
- General maintenance needed to take care of indexes and keep them well structured on an ongoing basis
- Good strategies for deciding which index is get included and which ones we need to ignore
- Should you use index maintenance wizard
- What column store indexes are
About Randolph West
Randolph West solves technology problems with a focus on SQL Server and C#. He is a Microsoft Data Platform MVP who has worked with SQL Server since the late 1990s. When not consulting, he can be seen acting on the stage and screen or doing voices for independent video games.
Transcription: Episode 82
Steve: So then let’s jump a little bit more into the care and feeding side of things, and I know we touched a little bit on rebuild and a little bit on statistics but if we look at what are the general things that we needed to do to take care of indexes and keep them well structured on an ongoing basis?
Randolph: I’d be interested to hear your side because I’ve been speaking a lot. I’ll open if I had any feedback.
Steve: Ok, I mean, so really I think on that there’s the concept of reorganizing versus rebuilding as you get fragmentation over time. And then there’s statistics and I think that, or let me start with a very bad way of doing it that I see quite often that is if you go and use one of the default maintenance plans. And the maintenance plan you just go and click let’s rebuild my indexes, let’s reorganize my indexes, then let’s rebuilt statistics. One of the things you end up doing is rebuilding which basically rebuilds the entire index from scratch without, basically if you’re in Enterprise Edition you can do that on online mode but if you’re not in Enterprise Edition that’s an offline operation. But what the default maintenance plan people sometimes check is that they say, “Let’s rebuild them.” And that gives you the best structured index you’re probably going to end up with and it also rebuilds your statistics. And then they reorganized them and then they rebuilt statistics with like a 5% sampling or something like that. And in that example you end up with a whole lot of work that’s actually giving you sort of a negative approach or negative impact. Whereas if you just rebuilt them to begin with and then not done the reorg, and not done the statistics rebuild afterwards you would be in a lot better shape.
Randolph: So, you know, those defaults that you’re talking about, the reason that happens is that because the maintenance plan wizard has got rebuilt then reorg, then statistics in that order in the list.
Steve: Absolutely, yeah, people just click the checkboxes or select them and that’s what you get.
Randolph: So when I see that I blame Microsoft for that. That’s purely Microsoft’s fault because generally people here using the maintenance wizard are using it because they don’t know how to do it with say, Ola Hallengren or MinionWare tools so just to name two, there are many. So that’s Microsoft’s fault and every time I see that I do not blame the DBA, or the accidental DBA, or whoever is managing it. It’s not their fault. But it is causing a whole bunch of work that is unnecessary.
Steve: Yes, it may not be their fault but it’s their pain they have to deal with. When I see that I generally going a little bit of education to explain why that’s bad and turn those off. And like you said, replace it with MinionWare or Ola Hallengren scripts. Both which are far better solutions at doing at than the default maintenance plans.
Randolph: So the one thing that I wanted to mention, Steve, is that an index rebuild is, as I mentioned before is size of data operations. So if your table is 10GB and you need to rebuild it, if you’re doing a clustered index rebuild it’s going to require 10GB of space in the data file and then it’s going to require at least 10GB in the transaction log. And if you’re rebuilding a clustered index it is going to by default rebuild all of the non-clustered indexes on that table that referred to that clustered index and that means all of them. So if you’ve got a 10GB table and you’ve got 30GB of non-clustered indexes on that table it’s going to require at least 40GB in the data file and 40GB in the transaction log because it’s going to rebuild everything. So if you’ve got a maintenance plan that says let’s do a clustered index rebuild and then before that you had the non-clustered indexes rebuilding it’s going to do them again anyway so it’s a nightmare. This is why I say it is statistic updates more regularly are going to be a lot more effective than doing index rebuilds over the long term. You still do need index maintenance but let’s not do it every night.
Steve: And that’s where I like the third party solutions that we mentioned there for the ability to rebuild those where you have options to say given a certain level of fragmentation I want to choose to reorganize this or if it’s even more fragmented then we’re going to choose to rebuild it. And some of them even have the option of only do this for a certain amount of time.
Randolph: Can you go into detail what’s the difference between a rebuild and a reorg is?
Steve: Ok, so with the rebuild it’s basically, technically it’s like the equivalent if you drop the index then recreate it, and then it’s going to be completely rebuilt from scratch and it will take up the log and disk I/O like you talked about a moment ago. But with the reorg, what that’s going to do is going to the index and basically restructure or compact the index pages as needed. It will go through and shuffle them around to compact them down to a smaller amount of space. And it will do that page by page as it goes through there and that can be done while things are online in Standard Edition. But it’s slow and the end outcome is not as good as a complete rebuild. It might get you real close but it’s not going to be good as complete rebuild but the benefit is you can keep it online.
Randolph: And you still need to update the statistics after you’ve done a reorganization which is something that people tend to forget as well. The statistics need to be up to date so that the query optimizer knows where the data is and how much of it is in there, and what the distribution of the data is in that index so that it can make the right choice about what type of query to create because a lot of queries is coming down to SORTS and JOINS and that’s where you find your biggest performance bottlenecks. A SORT is a blocking operation so everything in the query has to stop until the sort is finished. So if your index is up to date and it’s sorted correctly then it doesn’t have to do that operation it’s already done. Whole bunch of things about loops, we can talk about here as well which I’m not going to. An up to date statistic is much more valuable to the query optimizer than a defragmented index.
Carlos: There’s another point there even that, we’ve talked about current feeding of the indexes, right, rebuilds vs. reorgs. However, we can also have statistics that are outside of the indexes, right, on the table. Those are also providing feedback to SQL Server to the optimizer. And if we neglect those we’re still kind of the same peril because the optimizer maybe using the statistics but we’re not giving them that current feeding.
Randolph: Correct so, it’s valuable to have a, I hate this word, holistic approach to your current feeding is that you need to look at not just indexes, and not just fragmentation, and not just statistics but all of them. And in a way that is effective and efficient for your maintenance window. You don’t want to go and rebuild indexes every night because it doesn’t make any sense. But you also don’t want to stop doing statistics updates.
Steve: Yup. One thing I commonly see with that is rebuilding of the statistics or updating the statistics where somebody will have a job that goes through and blindly update every single statistic in the entire database every night. And that can have a pretty big performance impact on things depending on the size of your database. And a better approach to do that is to look at which tables or indexes actually have changes and only update those that have a fair amount of changes to them on that frequent basis. And then maybe on a regular basis of once a month or once a week you do an update on across all of them.
Randolph: Yeah, that’s a very good point. The amount of data that changes in a table if you’ve got automatic statistics enabled, we’re going to go a little bit of a segway now, by default automatic statistics are enabled. There are other settings that you can choose in 2016. There’s a whole bunch of new stuff per database which you can choose. Let’s talk about everything before 2016 because that’s what most people are still using. Statistics will only update themselves if a certain percentage of the table has changed. The value has changed across versions recently but the fact is that statistics will only update themselves if a certain percentage has been changed. So if you’ve got a really large table with tens of millions of records or rows and you have only one million that have changed recently statistics will not be updated automatically. So it does pay you to use one of the third party tools we’ve mentioned like MinionWare or Ola Hallengren’s maintenance solution where it will go and check, has any number changed and then you can update the statistics there and that will help your query optimizer as well.
Steve: The other side effect you can get with that is that if you do hit that threshold where it decides it’s time to rebuild those statistics that might be in the middle of your peak load during the day. And right when you’ve got a lot of traffic coming to your database and index or statistics rebuild occurred that can have impact on things too.
Randolph: That in fact is one of the recommendations for SharePoint style databases. There’s a lot of things I don’t like about SharePoint but the fact is a lot of people use it so one of their recommendations is to turn off automatic statistics updates on any SharePoint style database. That includes CRM, it includes GreatPlains, all of those even. What’s the one that I used? Whatever it is, there’s a whole bunch of them where you should turn off statistics updates automatically and then include statistics rebuilds in your maintenance plans. So it’s just to keep in mind each database is different.
Steve: It will probably apply to Navision and Dynamics as well.
Randolph: Yes that’s the one. Yeah, Dynamics is the one I’m trying to remember. Thank you! So any SharePoint style database they do recommend turning off statistics updates automatically and to do them in a separate maintenance window. So it pays to do your research to make sure that you’re doing the right kind of maintenance plans for your database and your instance.
Steve: Yup, very good point.
Carlos: So I guess a couple of different things we talked about all of these components. And I guess let’s now talk about some of the benefits right, so all of these things we have to go in, right, kind of the holistic approach, kind of having to know our data, getting more familiar with it. Ultimately to what end are we going to do that? I think, so we’ve talked a little bit about performance. I guess we should probably talk about how that performance gets into the system if you will or the mechanisms that cause the improvements?
Randolph: Ok, so when I do my performance session, what I say is, “You’ve got expensive items like your CPU. You’ve got less expensive items like your RAM, and you’ve got even less expensive items like your hard drives.” So your CPU is the most valuable thing to you because the SQL license is per CPU core and you want to make sure you’re using the most efficiencies of your CPU and memory as you can. What an index does is, we spoke about this before, it is a copy of the data so you want to keep your indexes as efficient as possible so that if you’ve got a large table you don’t want your index to be large as well. You want it to be smaller so that less of it is in memory because that’s what this game is about. SQL Server is all about being in memory as much data as possible in memory. So for Standard Edition up to 2016 even you’ve only got a certain amount of memory that you can access. 2016 Service Pack 1, the limit is still there but that’s your buffer pool that is in memory as opposed to everything. But the fact is that there are limits to the amount of memory you can use for SQL Server specifically with Standard Edition because not everybody can afford Enterprise. So you have to manage the amount of data that’s going into memory as much as you can and that is most effectively done by the right kind of indexes for your queries. And that’s also why you don’t want to have duplicate indexes because it will also be in memory. Also, you don’t want to have wide indexes because they will take up more memory than they need to. And that’s why included columns are very handy way to reduce the size of your indexes that’s why we have filtered indexes. All of these things to try and reduce the amount of data that is in memory so that we can do a lot more with what’s in the buffer pool.
Carlos: So then how do we go about or what’s the best ways to determine, we talked about looking at execution plan that kind of gives the recommended index there. So what are some good strategies to go about deciding which index is get included and which ones we need to ignore?
Steve: One approach I take on that is if I’m looking at an execution plan and it suggests an index. If it’s a development environment one of the things I’ll try initially is just create the index and see how it helps and I’ll drop it, and then go look and see is there a similar index. I mean, if the index that it suggested was beneficial I’ll go and see if there’s a similar index that could be modified or added to that would give the similar and a fact that the suggested index was doing. And sometimes that means you’re taking existing index and just add in an include to it or taking an existing index and add in another column into it.
Carlos: Ok, so that’s kind of a dev environment, one query at a time.
Steve: Yup, and the other approach that I see there is I can even go and look at some of the missing index DMVs to go and figure out. Actually you look at DMVs for a handful of things, one is to figure out what are the missing indexes and figure out the overall cause associated with those missing indexes. And then come up with what are some good ones to add that are not going to be duplicates of other indexes and then see how that improves performance. You can also, using DMVs, go in and look, you can figure out what are your duplicate indexes. If you have duplicate indexes oftentimes you’ll see that one of them might be not being used at all and the other one is being used a lot. You can go in and drop one of those. However, you need to be careful when you’re looking at unique indexes. You don’t want to drop a unique index versus a clustered index or non-clustered index you want to look at and drop the right one there. Another thing to look at is big clustered indexes, that’s another thing that you track down is what are the clustered indexes that are really wide. And that really wide means it’s going to ripple through all the non-clustered indexes with that extra size. You can also track down unused indexes. What are the indexes that are on your system that are not being used? Now when you look at that you need to be very careful because that’s only going to be the ones that haven’t been used since the SQL Server instance restarted. Also, if you rebuild and index I believe it clears those index usage stats.
Steve: Sometimes, exactly. So it’s something that if you see there’s an index that is not being used you might want to track it over a few weeks to a month and confirm it, yeah it really isn’t being used and then go and take a look dropping those.
Randolph: That goes back to one of the best practices is to have a baseline. So know what your system is doing and track that overtime and then you can refer back to that baseline and say, “Well, this is exactly what’s happening. This is different.” And go from there. Yeah, that’s a good point.
Steve: Yup, and that reminds me of a project, Carlos, that you and I worked on. Were we setup a monitoring component that ran for a month and kept track of unused index details, and log them every few hours, and then we went back at the end of the month and reviewed the ones that over the entire month had no utilization and then suggest those as possible candidates to be dropped.
Randolph: That reminded me of Pinal’s comment as well that he’ll only look at the database that’s been running for a week. The one thing I wanted to mention is the DMVs that you’re referring to there’s one in particular, there’s a group of diagnostic queries that Glenn Berry, Glenn Alan Berry, from sqlskills, he uses and maintains them. He’s got a couple of, and so does Brent Ozar and a whole bunch of other people. But Brent Ozar’s one called as sp_BlitzIndex and Glenn Berry’s one is included in his diagnostic scripts. It’s a very similar query and what it does it waits your missing indexes as according to number of scans, number of seeks and something called Average User Impact. Now that Average User Impact number can be, I think it’s unit less, I’m not quite sure how does that number is calculated but if it’s really in a high impact like tens and hundreds of thousands then usually I will go and look at that first and say, “Well, how many times has this index been used in terms of number of seek and scans.” And if it’s a high usage index that is missing or a high impact then I will usually create that without too many modifications if it’s not too wide or doesn’t have too many include columns.
Carlos: I think it’s the number of times requested multiplied by the number of hits.
Steve: Yup, and just to jump in there I think I’d feel a little bit left out I didn’t have the chance to mention Database Health Monitor on that. In Database Health Monitor there are many indexing reports similar to what Randolph has described there.
Randolph: Certainly, from Steve Stedman’s solutions has a magnificent and free tool called Database Health Monitor which I have used. I don’t even get paid for this statement. It’s a good product. It’s free which is even more amazing. This is the great thing about the SQL Server community. There are a lot of free tools that are out there that are adding value all the time. And all that people asked is that you recognized them and I recognized Steve as a valuable member of our community. That ends the sponsorship message.
Steve: Thank you for the shameless plug.
Carlos: So it’s interesting, I guess, and maybe I’m being a little bit paranoid but using, so when I’m looking at that impact and I will take that and also using the ones that I’m interested in adding. Of course I want to do what Steve mentioned looking for duplicates or kind of what’s there, right? Then am I going to get over my threshold of ten or whatever just kind of taking into consideration what additional load am I putting on this table by creating the index. And while you want to test that out mostly the environments that I worked on they just don’t have a good mechanism for creating the similar load as I have in production. So when I go to implement that index one of the first things I’ll start to look at is those usage stats, right. Because I want the usage stats in the index that I just created to be going up because I know that, “Ok well, wait a second what did I do? Is this still a really good candidate?”
Randolph: Yeah, that’s a very good point.
Carlos: So I guess, other that’s how are going to go out and creating them, couple of other things that we didn’t really get into some of the other new indexes like column store or even XML indexes. But I guess other thoughts about when to start looking at these other, we even try to approach that now. Maybe I feel like we should punt on some of those.
Randolph: I can briefly mention about XML indexes that the way they work is they’re going to be larger than your table or your XML column. The reason being is that it will create an internal table according to the structure the XML document or XML field that you’re indexing. So it actually expands out the XML data into an internal table and then indexes that so you could end up with a column that is say 100kb maximum. That’s a small one and you can end up with an index that is 500kb, or gigabyte, or tens of gigabyte because it’s creating an internal table under the covers. A system table that it is then indexing so be wary of XML columns in that if you’re going to index then make sure that there’s sensible indexes that they’re only indexing certain nodes inside that XML document and be aware of that. And also I’d like to add here that if you use the XML data type, the data going into that XML row or column is not going be the same as what you get out. It does modify the XML data going in for whatever reasons.
Carlos: Say that one more time.
Randolph: If you have XML data that you put in for auditing reasons for example. If you pull it out it’s going to have been modified somehow. Either the tags will be slightly different or the spacing will be slightly different so do not use an XML data type to store audited records of XML types. If you want to keep an absolute record of what your XML looked like as it came out of a web service for whatever reason store it in a varchar or nvarchar column instead because in it it is identical. If you put it into an XML data type it will modify the data. It is still the same data but it will slightly modify. The tags will be slight different or whatever.
Steve: So then as far as column store indexes I think that’s something we could probably dedicate an entire episode just to talk about.
Randolph: We definitely can.
Carlos: I think you’re right.
Randolph: A column store index is neither an index nor clustered so it’s so complicated. In fact, there is a series of post. I think it’s Nico who has done it over one hundred posts on how column store indexes work. We could talk for days on that. They are fascinating and completely subverts what you think you know about indexes, and data, and row level storage and all that kind of stuff. It’s fascinating stuff.
Steve: Yeah, and just a few notes on that I think that it’s one of those things that is great for data warehousing or OLAP type things. And may not always be the best option for your OLTP side.
Randolph: At the same time you could have reporting style queries in your database and with 2016’s optimizations for column store you could have which can change and can be clustered and all sorts of stuff. You could have some stuff that in your OLTP environment that could be OLAP style indexes. Yeah, there’s so much to think about there.
Carlos: We saw that in Episode 78 with Brian Carrig. They talked about using their column store index in a transactional environment.
Steve: Yup, and then one of the things that I’ve seen that is one of those misunderstandings around column stores is that if you, because column store does the compression on each column, is that people think of it as I don’t have to have any non-clustered indexes if I have a column store index. And that’s not true at all. And that if you’re looking at a bigger table and you’re using column store you may still need to have some non-clustered indexes on it as well.
Randolph: Oh yeah, so do you research. Read all one hundred and whatever post. Become an expert then implement them.
Carlos: So we’ve talked a little bit about, so implications, how do we know what to use. So space considerations we talked a little bit about more in the rebuild process, right? That we’re going to need additional space in our data file and our log file things like that. I think we’ve had a pretty good conversation there. I guess one last idea I’ll throw out that we can bat around a little bit. We talked a lot about having multiple TempDB files and so one of the thoughts that I’ve seen out there is if you’re potentially, I won’t say, so what’s the value there is that if you’re looking to spread out or move your indexes from where your data might reside. One way to do that is to create another data file potentially on a separate storage and then rebuild those indexes.
Randolph: Ok, there are two things that I can talk about there. The first thing I wanted to mention is if you do need to move your table into a different file or file group all you have to do is do a clustered index rebuild and target that file group as the new location for the clustered index because remember the clustered index is you data. So that’s a handy way if you need to move your table into a different file group that’s how you will do it. The other thing is by virtue of that definition there is that you could have your non-clustered indexes in a different file as your data and you might want that for improved I/O performance or if you’ve got your read-only stuff in a different file group or all sorts of reasons for that. It’s very handy for splitting your load on the I/O level. Less of a problem these days but it’s a nice thing for large tables to split your non-clustered indexes from your clustered index so that’s not reading from the same portion of the disk or the I/O subsystem or both.
Steve: Just to note on that index rebuild for a clustered index to move it to a different file group although that will move the table and the entire index. One of the things that could be left behind when you do that is any of the varchar max or nvarchar max or other large items that are located outside of the data pages for that table.
Randolph: Yeah, off row data will be affected. Steve, I don’t know because I’ve never tried. What happens if you tell it to do the index rebuild with the log compaction enabled.
Steve: You know, that’s a good question. I have not tried that for a long time.
Randolph: I smell a blog post.
Carlos: Ok, very good. Well, awesome. I think great conversation and obviously there’s a lot more that we could talk about indexes but I think this was a noble attempt at covering some of the basics and getting into some of the nitty gritty as well.
Randolph: Yeah, the fact is that indexes don’t stand by themselves. They are a very important part of everything including statistics and everything else so don’t think that once you became an index expert you become an index performance tuning expert because that’s not true. You have to have a very broad knowledge of how things work in a number of different fields used upon to get the best performance out of your system. And there’s nothing wrong with good enough. You don’t have to have 100% defrag indexes. You don’t have to have indexes rebuilt. You can have them reorganized. Don’t have to have them reorganized at all if your statistics are up to date in certain contexts. There are a lot of tradeoffs that you have to think about when doing your maintenance plans and indexes form just a small part of that.
Steve: Very good point.
Carlos: So shall we do SQL Family?
Steve: So how did you first get started with SQL Server, Randolph?
Randolph: Steve, that’s a good question because when I originally got this question I had to think very hard about it. I have been fascinated from a personal level about organizing data. Back in my youth I would create lists and lists of things CDs, books all that kind of stuff and I would do them in the paper equivalent of Microsoft Excel. And then when Lotus 123 came along I started playing with that because my dad worked at the bank and he was a 123 expert. If I’d know about VisiCalc I probably would have used it. And then I got into Microsoft Access because it was there. It was a very old version of Access. I think it was version 2 that I started with and then I started working at a PeopleSoft implementation partner in South Africa where I am from. And that’s where I first work with SQL Server and Oracle at the same time, and I was not a fan of SQL Server. I have to be honest. At that time it was version 6.5 and version 7 that’s around the time I joined so there were issues because as some of your listeners may know SQL Server’s engine was rebuilt around 7 time, 6.5, 7 time. In fact, if you look on MSDN you can download 6.5 and you can download 2000 but you cannot download version 7. It is very difficult to come by. There are maybe good reasons for that. I got exposed to SQL Server that would have been in 1997, around there. So yeah, that’s where I first got exposed but then I didn’t really play with it until I got to the bank. I was more of an Oracle guy. Got to the bank in 2006 so there was a large gap of not playing in SQL Server and then I couldn’t figure out why DESCRIBE, the keyword DESCRIBE wasn’t working in SQL Server. For all of you Oracle office you notice there I prefer SQL Server now because as I said before a lot has changed. It is a much better product than it was in 1997. In fact, I think it has surpassed Oracle. I think it’s better than anything else as well and that’s because I’m trying to keep abreast of all the new stuff. I don’t want to be stuck in the past and have assumptions about the product. I want to play around with the new stuff so. That was a long way of saying 20 years or so.
Steve: Ok. Well I know that in the late 90’s Oracle was a much better product than SQL Server.
Randolph: It really was.
Steve: Yup, and that has changed. That has changed significantly in the last 17 years.
Randolph: Yeah, it has.
Carlos: If you could change one thing about SQL Server what would it be?
Randolph: I had a pot for this but I’m going to change my mind. The pot answer is in-memory processing on SQL Server, in-memory OLTP which was called Hackathon, that was the code name for it. There is a transaction log operation that happens even for stuff that you don’t want kept. It’s still being backed by disks somewhere. It may not be in the transaction log itself but it is still being backed by disk somewhere. I would like to have in-memory objects that do not have anything on disk backing them at all. They must just be there, and if the server goes down, oh well tough luck.
Carlos: Alright. Yes, I remember seeing you in the Bob Ward presentation at SQL PASS talking about the in-memory OLTP.
Randolph: There’s some very very cool stuff with in-memory OLTP that I’ve been playing with. I’ve been blogging about it as well on my blog bornsql.ca. That speaks about some in-memory ops and I found. Interestingly I found another issue that I’ve spoken to Jonathan Kehayias about where Temp tables are slower than Temp variables in certain cases. So I’m investigating that at the moment and it had to do with my in-memory investigations.
Steve: Oh interesting, I’d like to see what you find there.
Randolph: If I could change, that was my pot answer about SQL Server what I would change. What I would change is having a TempDB per user database as a configurable extra.
Steve: That would be nice.
Randolph: In other words it doesn’t have to be default but if I need a TempDB for a particular database I would like to have that as a separate entity so that I could manage it separately and keep track of the objects that are happening for that particular database. The architecture for that is going to be really difficult to do so I don’t know if Microsoft is going to do that but that’s what I would like.
Steve: Alright, so what’s the best piece of career advice that you’ve received.
Randolph: Oh, it’s going to be along answer as well, sorry. “Let your work speak for itself” is the first and foremost one. So it doesn’t matter what people say about you if your work can speak for itself then you don’t have to say anything. That’s the best piece of advice that I had. But the other is “Everything you say and everything you do is being recorded somewhere” so please treat people with respect. Treat people as you would have them treat you and don’t say things on voicemails that can be played back to you at meetings with the CEO and saying Randolph don’t do that.
Steve: I like that too.
Carlos: We won’t ask for the personal experience that led you to that advice.
Randolph: I have at the end of the year or every now and then when it’s a quiet time on my blog I say career limiting moves and that was one of them. So there are a bunch of career limiting moves that I can recommend to you. Don’t do because I’ve done them not because it’s good advice but it’s because I’ve done them so don’t do that because I’ve done it, and I will be one of them.
Carlos: Our last question for you today. If you could have one superhero power what would it be and why do you want it?
Randolph: I thought about this and I honestly don’t know. I liked Pinal’s when where he could touch a server and know what’s wrong with it. The problem is whenever I touch as server if it’s going to break it’s going to break then and there. Maybe I want the reverse of that superpower.
Steve: To just know without touching the server.
Randolph: Yeah, you know per cost of maintenance is one thing. But if I’m in a room and it is going to break it is going to break while I’m there which is good I guess in way because you then you know that we can figure it out then and there. But I’d like the reverse of that that would be cool.
Carlos: Well, awesome. Randolph thanks so much for being with us today.
Randolph: Well, thanks for having me it’s a topic near and dear to my heart indexing and maintenance and stuff like that so I’m glad I got an opportunity to speak to you guys about it.